This action might not be possible to undo. Are you sure you want to continue?
POWER FLOW INTO A POWER SYSTEM SIMULATION
ENVIRONMENT
BY
JAMES DANIEL WEBER
B.S., University of Wisconsin  Platteville, 1995
THESIS
Submitted in partial fulfillment of the requirements
for the degree of Master of Science in Electrical Engineering
in the Graduate College of the
University of Illinois at UrbanaChampaign, 1997
Urbana, Illinois
ii
REDBORDERED FORM
iii
ABSTRACT
In this thesis, a Newtonbased optimal power flow (OPF) is developed for implementation
into a power system simulation environment. The OPF performs all system control while
maintaining system security. System controls include generator megawatt outputs, transformer
taps, and transformer phase shifts, while maintenance of system security ensures that no power
system component’s limits are violated. Special attention is paid to the heuristics important to
creating an OPF which achieves solution in a rapid manner. Finally, sample applications of the
OPF are discussed. These include transmission line overload removal, transmission system
control, available transfer capability calculation (ATC), real and reactive power pricing,
transmission component valuation, and transmission system marginal pricing.
iv
ACKNOWLEDGEMENTS
I would like to thank Professor Thomas Overbye for his knowledge, guidance and support
throughout my graduate studies. I would also like to thank the University of Illinois Power
Affiliates Program and the University of Illinois Graduate College for their generous financial
support.
I would also like to thank my professors and friends from my undergraduate studies while at
the University of Wisconsin  Platteville. The preparation and experience I gained there were
invaluable. Special thanks go to Dr. Mesut Muslu and Dr. Richard Shultz.
Finally, I would like to thank my family. My brothers Brian and Scott continue to be there
for support and friendship. Most of all I thank my mother and father, Jan and Gene. They
continue to supply the unconditional love and support which allow me and my brothers to
achieve what we have and will.
v
TABLE OF CONTENTS
Page
1. INTRODUCTION................................................................................................................... 1
1.1 Motivation ........................................................................................................................... 1
1.2 Literature Survey................................................................................................................. 3
1.3 Goals of the OPF................................................................................................................. 4
1.4 Overview............................................................................................................................. 5
2. DEVELOPMENT OF NEWTONBASED OPTIMAL POWER FLOW............................... 7
2.1 Background on Newton’s Method....................................................................................... 7
2.1.1 Problem statement ....................................................................................................... 7
2.1.2 Development of Lagrangian, gradient and Hessian..................................................... 8
2.1.3 Application of inequality constraints........................................................................... 9
2.1.4 Solution method .......................................................................................................... 9
2.2 Application of Newton’s Method to OPF ......................................................................... 11
2.2.1 The objective function............................................................................................... 11
2.2.2 Equality constraints ................................................................................................... 12
2.2.3 Inequality constraints................................................................................................. 13
2.2.4 Soft constraints by using penalty functions............................................................... 14
2.2.5 Treatment of discrete variables ................................................................................. 17
2.2.6 Summary of optimal power flow problem................................................................ 17
2.2.7 Summary of Lagrangian terms .................................................................................. 18
2.2.8 Calculation of gradient and Hessian.......................................................................... 19
2.2.9 Solution of the optimal power flow........................................................................... 20
2.3 A Sample Case Illustrating OPF Algorithm Process......................................................... 21
2.4 Information Gained from the OPF Solution...................................................................... 22
3. HEURISTICS OF THE OPTIMAL POWER FLOW SOLUTION ...................................... 24
3.1 Classification of OPF Variables ........................................................................................ 24
3.2 Implementation of Sparse Matrix Techniques .................................................................. 25
3.3 Determination of the Set of Binding Inequality Constraints ............................................. 28
3.4 Searching Algorithms........................................................................................................ 30
3.5 Solution of an OPF Repeatedly Over Time....................................................................... 31
4. USES OF AN OPTIMAL POWER FLOW IN A POWER SYSTEM SIMULATION
ENVIRONMENT.......................................................................................................................... 33
4.1 Example Line Overload Removal ..................................................................................... 33
4.2 Use for Bus Real and Reactive Power Pricing.................................................................. 34
4.3 Use for Area Real Power Pricing ...................................................................................... 36
4.4 Example Transformer Tap Control ................................................................................... 38
4.5 System MVAR Control Using Transformer Taps [26] ................................................... 41
4.6 Transmission Line Valuation by TimeDomain Simulation ............................................. 45
4.7 Capacitor Bank Valuation by TimeDomain Simulation .................................................. 47
4.8 Limit on Available Transfer Capability (ATC) Due to a Voltage Constraint [26] .......... 48
vi
4.9 Transmission System Pricing Through ShortRun Marginal Costing (SRMC)................ 52
5. CONCLUSION..................................................................................................................... 56
APPENDIX A. ECONOMIC INTERPRETATION OF THE LAGRANGE MULTIPLIERS.... 58
APPENDIX B. CALCULATION OF THE GRADIENT OF THE LAGRANGIAN................... 61
APPENDIX C. CALCULATION OF THE HESSIAN OF THE LAGRANGIAN ...................... 64
APPENDIX D. SUMMARY OF DERIVATIVE CALCULATIONS .......................................... 68
APPENDIX E. SIXBUS SAMPLE POWER SYSTEM.............................................................. 82
APPENDIX F. TWENTYTHREE BUS SAMPLE POWER SYSTEM...................................... 84
APPENDIX G. SEVENBUS SAMPLE POWER SYSTEM....................................................... 87
APPENDIX H. THREEBUS SAMPLE POWER SYSTEM....................................................... 89
REFERENCES.............................................................................................................................. 90
vii
LIST OF TABLES
Table Page
Table 3.1 OPF problem variables................................................................................................. 24
Table 4.1 Sample transaction scenarios ....................................................................................... 37
Table 4.1 Summary of tap ratio control experiment.................................................................... 41
Table 4.1 Comparison of two systems ......................................................................................... 42
Table 4.2 Results for various interchanges with transformer taps left inactive ........................... 44
Table 4.3 Results for various interchanges with transformer taps active..................................... 44
Table 4.1 Simulation data for sixbus system with transmission lines removed ......................... 46
Table 4.1 Simulation data for sixbus system with transmission lines removed ......................... 48
Table 4.1 SRMC calculation for sixbus, twoarea system........................................................... 53
Table 4.2 SRMC calculation for sixbus, twoarea system where no limit is encountered.......... 54
Table E.1 Line characteristics for sixbus system........................................................................ 82
Table E.2 Bus characteristics for sixbus system......................................................................... 83
Table E.3 Economic information for sixbus system................................................................... 83
Table F.1 Line characteristics for twentythree bus system......................................................... 85
Table F.2 Bus characteristics for twentythree bus system.......................................................... 86
Table F.3 Economic information for sixbus system................................................................... 86
Table G.1 Line characteristics for sevenbus system................................................................. 87
Table G.2 Bus characteristics for sevenbus system.................................................................. 88
Table G.3 Economic information for sevenbus system.......................................................... 88
Table H.1 Line characteristics for threebus system.................................................................... 89
Table H.2 Bus characteristics for threebus system ..................................................................... 90
Table H.3 Economic information for threebus system............................................................... 90
viii
LIST OF FIGURES
Figure Page
Figure 2.1 Newton’s Method Flowchart ....................................................................................... 10
Figure 2.1 Multiarea System with Scheduled Interchanges ......................................................... 13
Figure 2.1 Penalty Function for Bus Voltage ............................................................................. 16
Figure 2.2 Penalty Function for Line MVA Flow Limit............................................................. 16
Figure 3.1 Sample Binary Tree..................................................................................................... 30
Figure 4.1 Sixbus, SingleArea System Not on OPF Control..................................................... 33
Figure 4.2 Sixbus, SingleArea System on OPF Control............................................................ 34
Figure 4.1 Line Limit from Bus 5 to 4 Raised to 100 MVA....................................................... 35
Figure 4.1 Sixbus, TwoArea System on OPF Control.............................................................. 36
Figure 4.2 Transaction of 65.5 MW Undertaken ......................................................................... 37
Figure 4.1 TwentyThree Bus System on OPF Control with Tap Control Off ........................... 38
Figure 4.2 TwentyThree Bus System with Tap Control On........................................................ 41
Figure 4.1 SevenBus, OneArea System with Taps Inactive ..................................................... 41
Figure 4.2. SevenBus, OneArea, System with Taps Active ..................................................... 42
Figure 4.3. SevenBus, TwoArea System with TapChanging Transformers ............................ 43
Figure 4.1 Load Factor Curve for sixbus System........................................................................ 46
Figure 4.1 Load Factor Curve for TwentyThree Bus System..................................................... 47
Figure 4.1 ThreeBus Base Case with No Area Power Transfer.................................................. 49
Figure 4.2 ThreeBus Example at Maximum Power Transfer (15 MVAR Capacitor Support). 50
Figure 4.3 ThreeBus Example at Maximum Power Transfer (30 MVAR Capacitor Support). 51
Figure 4.4 ThreeBus Example at Maximum Power Transfer (45 MVAR Capacitor Support). 51
Figure 4.1 Sixbus, TwoArea System Undergoing Transaction with Line Limit Doubled ........ 54
Figure D.1 Transformer Model .................................................................................................... 74
Figure E.1 OneLine Diagram of SixBus System...................................................................... 82
Figure F.1 OneLine Diagram of the TwentyThree Bus System ............................................. 84
Figure G.1 OneLine Diagram of the SevenBus System............................................................ 87
Figure H.1 OneLine Diagram of the ThreeBus System............................................................ 89
ix
NOTATION
Throughout this thesis, variables that do not have a subscript next to them may be considered
vectors or matrices, while all variables that refer to a scalar will have a subscript next to them.
V
k
Magnitude of voltage at bus k.
δ
k
Angle of voltage at bus k.
t
km
Transformer tap ratio between buses k and m.
α
km
Transformer phase shift between buses k and m.
P
Gk
The real power generated at bus k.
b
km
Element of the imaginary part of the network admittance matrix.
g
km
Element of the real part of network admittance matrix.
y
km
Magnitude of an element of the network admittance matrix.
δ
km
Phasor angle of an element of the network admittance matrix.
P
k
Real power injection at bus k.
Q
k
Reactive power injection at bus k.
P
km
The real power flow from bus k to bus m.
Q
km
The reactive power flow from bus k to bus m.
S
km
The MVA flow from bus k to bus m.
P
int
The real power interchange for an area.
P
sched
The scheduled real power interchange for an area.
f(• ) The objective function.
g(• ) Equality constraints.
h(• ) Inequality constraints.
L(•) The Lagrange function or Lagrangian.
H(•) The Hessian of the Lagrangian.
∇L(•) The gradient of the Lagrangian.
a
i
, b
i
, and c
i
, Coefficients for the quadratic cost curve of a generator
X
max
Signifies a maximum bound on a variable.
x
X
min
Signifies a minimum bound on a variable.
µ In general, this is a Lagrange multiplier for an equality constraint.
λ In general, this is a Lagrange multiplier for an inequality constraint.
List of Lagrange multipliers
µ
Pk
For P
k
, the real power injection at bus k.
µ
Qk
For Q
k
, the reactive power injection at bus k.
µ
viset
For generator voltage set point.
µ
int
For generator voltage set point.
λ
Skm
For MVA constraint on line from bus k to m.
λ
PGih
For generator maximum power output constraint at bus i.
λ
PGil
For generator minimum power output constraint at bus i.
λ
Vih
For maximum bus voltage constraint at bus i.
λ
Vil
For minimum bus voltage constraint at bus i.
λ
tkm max
For maximum transformer tap ratio constraint .
λ
tkm min
For minimum transformer tap ratio constraint.
λ
α
km max
For maximum transformer phase shift constraint.
λ
α
km min
For minimum transformer phase shift constraint.
1
1. INTRODUCTION
1.1 Motivation
Throughout the entire world, the electric power industry has undergone a considerable change
in the past decade and will continue to do so for the next several decades. In the past the electric
power industry has been either a governmentcontrolled or a governmentregulated industry
which existed as a monopoly in its service region. All people, businesses, and industries were
required to purchase their power from the local monopolistic power company. This was not only
a legal requirement, but a physical engineering requirement as well. It just didn’t appear feasible
to duplicate the resources required to connect everyone to the power grid.
Over the past decade, however, countries have begun to split up these monopolies in favor of
the free market. Numerous papers and articles have been written on this topic with a good
overview of the topic found in a series of articles written for IEEE Spectrum in July and August
of 1996 [1  6]. In the United States, the change from the regulated monopoly to the free market
system has become known as restructuring. For the remainder of this thesis, it will be referred to
as restructuring.
One of the cornerstones of any restructuring plan is the ability to operate the transmission
system in a manner which is fair to all participants in the industry. In the United States, the
Federal Energy Regulatory Commission (FERC) oversees issues involving the transmission
system. FERC presently believes that the only manner in which everyone will be on an equal
playing field is to create open access to all. As stated in [7], “participants in wholesale power
markets will have nondiscriminatory open access to the transmission systems of public utilities.”
2
In order to achieve the ideal of open access, many outstanding engineering problems will need to
be investigated and tools created for their solution.
It is very important that these problems be addressed early in the restructuring process. If
these engineering problems become overshadowed by short term economic concerns, then the
result could be decreased electricity reliability. In the past year, the western United States has
seen the consequences of pushing the transmission system too hard on two separate occasions.
The two multistate blackouts in the Western States Coordinating Council (WSCC) system in the
last several months may be destined to repeat themselves [8].
In a presently unpublished report from a joint PSERC/EPRI workshop, many of these tools
were identified [9]. A partial list of problems that can be addressed from the work done in this
thesis follows:
Control Problems
• Computation of realtime available transfer capability (ATC)
• Realtime control of power flows
• Tools to relieve congestion in fair, justifiable and economic manner
• Tools for congestion management (including congestion pricing)
• Tools for determining the ISO action before the contingency occurs
Economic Problems
• Realtime pricing and price risk management services
• Tools for operating the power system in the most economical manner
• Transaction evaluation tools that enable players to evaluate their own costs
• Methodologies for determining the value (cost) of ancillary services in improving
efficiency and flexibility
UserInterface and Simulation Problems
• Market simulation models
• Better tools for communication and display of information will permit the better
operation of the min ISO scenario
3
The work presented in this thesis utilizes an optimal power flow program, OPF, as the tool
for solving these problems. The OPF is a natural choice for addressing these concerns because it
is basically an optimal control problem. The OPF utilizes all control variables to help minimize
the costs of the power system operation. It also yields valuable economic information and insight
into the power system. In these ways, the OPF very adeptly addresses both the control and
economic problems.
After creating the OPF program, the userinterface and simulation problems may also be
addressed by implementing the OPF into a power system simulator. In this way, the results of
the economic and control operations of the OPF can easily be utilized by the user of the program.
1.2 Literature Survey
The optimal power flow problem has been discussed since its introduction by Carpentier in
1962 [10]. Because the OPF is a very large, nonlinear mathematical programming problem, it
has taken decades to develop efficient algorithms for its solution. Many different mathematical
techniques have been employed for its solution. The majority of the techniques discussed in the
literature use one of the following five methods [11, p.517]:
• Lambda iteration method  Also called the equal incremental cost criterion (EICC)
method. This method has its roots in the common method of economic dispatch used
since the 1930s. See [11, p. 39 ]
• Gradient method  See paper written by Dommel and Tinney [12]
• Newton’s method  See paper by Sun et al. [13]
• Linear programming method  See paper by Alsac et al. [14]
• Interior point method  See paper by Wu, Debs, and Marsten [15]
4
An excellent literature survey of the different techniques can be found in a paper by Huneault and
Galieana published in 1991 [16]. Though it does not discuss the interior point method, it does
make reference to over 150 papers on the optimal power flow problem covering all the other
methods for solving the OPF.
This thesis will explore the application of Newton’s method to the OPF problem.
Specifically, it will explore the implementation of a Newton’s method based OPF in the power
system simulator POWERWORLD [17].
1.3 Goals of the OPF
Before beginning the creation of an OPF, it is useful to consider the goals that the OPF will
need to accomplish. The primary goal of a generic OPF is to minimize the costs of meeting the
load demand for a power system while maintaining the security of the system. The costs
associated with the power system may depend on the situation, but in general they can be
attributed to the cost of generating power (megawatts) at each generator. From the viewpoint of
an OPF, the maintenance of system security requires keeping each device in the power system
within its desired operation range at steadystate. This will include maximum and minimum
outputs for generators, maximum MVA flows on transmission lines and transformers, as well as
keeping system bus voltages within specified ranges. It should be noted that the OPF only
addresses steadystate operation of the power system. Topics such as transient stability, dynamic
stability, and steadystate contingency analysis are not addressed.
5
To achieve these goals, the OPF will perform all the steadystate control functions of the
power system. These functions may include generator control and transmission system control.
For generators, the OPF will control generator MW outputs as well as generator voltage. For the
transmission system, the OPF may control the tap ratio or phase shift angle for variable
transformers, switched shunt control, and all other flexible ac transmission system (FACTS)
devices.
A secondary goal of an OPF is the determination of system marginal cost data. This marginal
cost data can aid in the pricing of MW transactions as well as the pricing ancillary services such
as voltage support through MVAR support. In solving the OPF using Newton’s method, the
marginal cost data are determined as a byproduct of the solution technique. This will be
discussed later in Section 2.4 on page 22.
1.4 Overview
The OPF program written in conjunction with this thesis uses Newton’s method as its
solution algorithm. It will tackle all of the goals set forth for an OPF except the control of
switched shunts and other FACTS devices. The control of these may be added at a later time as
desired.
The remainder of this thesis will discuss the development of the OPF. Chapter 2 of this
thesis will discuss the theory of the Newtonbased optimal power flow. It will lay a framework
for the mathematics and engineering behind the OPF computer source code. Chapter 3 will
discuss some special heuristics important to creating an OPF which achieves solution in a rapid
manner. Chapter 4 will show several sample applications of the OPF. The sample applications
6
discussed will include transmission line overload removal, transmission system control, available
transfer capability (ATC) calculations, real and reactive power pricing, transmission component
valuation, and transmission system marginal pricing. Finally, Chapter 5 will conclude with a
summary and several improvements which would aid in creating a truly useful OPF.
7
2. DEVELOPMENT OF NEWTONBASED OPTIMAL POWER FLOW
2.1 Background on Newton’s Method
Newton’s method is wellknown in the area of power systems. It has been the standard
solution algorithm for the power flow problem for decades [18]. A good reference for the theory
of Newton’s method is a book by Luenberger [19], which describes Newton’s method as well as
its quadratic convergence properties. The detailed explanation in [19] is left to the interested
reader. This thesis will only cover the process of applying Newton’s method to a minimization
problem such as the OPF.
Newton’s method is a very powerful solution algorithm because of its rapid convergence near
the solution. This property is especially useful for power system applications because an initial
guess near the solution is easily attained. System voltages will be near rated system values,
generator outputs can be estimated from historical data, and transformer tap ratios will be near
1.0 p.u.
2.1.1 Problem statement
A general minimization problem can be written in the following form.
Minimize (the objective function)
subject to: (equality constraints)
(inequality constraints)
f(x)
h (x) = , i = , , . . . , m
g (x) , j = , , . . . , n
i
j
0 1 2
0 1 2 ≤
There are m equality constraints and n inequality constraints and the number of variables is equal
to the dimension of the vector x.
8
2.1.2 Development of Lagrangian, gradient and Hessian
The solution of this problem by Newton’s method requires the creation of the Lagrangian as
shown below.
( ) ( ) ( ) ( ) L z f x h x g x
T T
· + + · µ λ the Lagrangian
where [ ] z x
T
· µ λ , µ and λ are vectors of the Lagrange multipliers, and g(x) only
includes the active (or binding) inequality constraints.
A gradient and Hessian of the Lagrangian may then be defined.
Gradient = ∇L(z) =
∂
∂
¸
1
]
1
L z
z
i
( )
= a vector of the first partial derivatives of the Lagrangian
Hessian = ∇
2
L(z) = H =
∂
∂ ∂
¸
1
]
1
1
2
L z
z z
i j
( )
=
∂
∂ ∂
∂
∂ ∂
∂
∂ ∂
∂
∂ ∂
∂
∂ ∂
¸
1
]
1
1
1
1
1
1
1
2 2 2
2
2
0 0
0 0
L z
x x
L z
x
L z
x
L z
x
L z
x
i j i j i j
i j
i j
( ) ( ) ( )
( )
( )
µ λ
µ
λ
=
Note the extremely sparse structure of the Hessian matrix shown. This sparsity will be exploited
in the solution algorithm.
From this, according to optimization theory, the KuhnTucker necessary conditions of
optimality are [19, p. 314]:
∇
x
L(z
*
) = ∇
x
L([x
*
, λ
*
, µ
*
]) = 0;
∇
λ
L(z
*
) = ∇
λ
L([x
*
, λ
*
, µ
*
]) = 0;
∇
µ
L(z
*
) = ∇
µ
L([x
*
, λ
*
, µ
*
]) = 0;
λ
i
*
≥ 0 if g(x
*
) = 0 (i.e., the inequality constraint is active)
λ
i
*
= 0 if g(x
*
) ≤ 0 (i.e., the inequality constraint is not active)
µ
i
*
= Real
where z
*
= [x
*
, λ
*
, µ
*
] is the optimal solution.
Thus solving the equation ∇
z
L(z
*
) = 0 will yield the optimal problem solution.
a matrix of the
second partial
derivatives of
the Lagrangian
9
2.1.3 Application of inequality constraints
It should be noted that special attention must be paid to the inequality constraints of this
problem. As noted, the Lagrangian only includes those inequalities that are being enforced. For
example, if a bus voltage is within the desired operating range, then there is no need to activate
the inequality constraint associated with that bus voltage. For this Newton’s method
formulation, the inequality constraints will be handled by separating them into two sets: active
and inactive [19, p. 326]. For efficient algorithms, the determination of those inequality
constraints that are active is of utmost importance. While an inequality constraint is being
enforced, the sign of its associated Lagrange multiplier at solution determines whether continued
enforcement of the constraint is necessary. Essentially the Lagrange multiplier is the negative of
the derivative of the function that is being minimized with respect to the enforced constraint (see
Appendix A for a derivation of this fact). Therefore, if the multiplier is positive, continued
enforcement will result in a decrease of the function, and enforcement is thus maintained. If it is
negative, then enforcement will result in an increase of the function, and enforcement is thus
stopped. The outer loop of the flow chart in Figure 2.1 performs this search for the binding or
active constraints.
2.1.4 Solution method
Considering the issues discussed above, the solution of the minimization problem can be
found by applying Newton’s method to ∇
z
L(z) = 0. A flowchart of this process is shown in
Figure 2.1. This flowchart will be useful for any generic minimization problem. A more detailed
10
discussion of this flowchart will be reserved for the following section. There the application of
Newton’s method to the optimal power flow problem will be discussed.
Make initial guess of
vector z = [ x µ λ ]
T
and which inequality
constraints to enforce.
Calculate the Hessian
and gradient of the
Lagrangian
Create the Lagrangian
given the active
inequality constraints.
Solve the equation
[H] ∆z = ∇L(z)
for ∆z
Calculate new z
z
new
= z
old
 ∆z
Check
tolerance
∆z< ε
Are correct
inequalities
enforced?
Problem Completed
YES
NO
YES
Determine new set of
inequalities to
enforce using
Lagrange multipliers
NO
Figure 2.1 Newton's Method Flowchart
11
2.2 Application of Newton’s Method to OPF
As discussed in Section 1.3, the goal of the OPF is to minimize the costs of meeting the load
demand for a power system while maintaining the security of the system. This section of the
thesis will discuss the application of Newton’s method in a manner that will achieve this desired
goal.
First, the objective function, f(x), will be introduced. It will reflect the desire to minimize the
costs of the system. Then the equality and inequality constraints will be discussed. These
constraints model the physical laws of the power system as well as the need to maintain system
security. At this point, the concept of soft constraints using penalty functions will be introduced.
Penalty functions also model the need to maintain system security. Finally, all terms in the
Lagrangian, gradient, and Hessian will be summarized.
2.2.1 The objective function
The objective function for the OPF reflects the costs associated with generating power in the
system. The quadratic cost model for generation of power will be utilized:
C a b P c P
P i i Gi i Gi
Gi
· + +
2
where P
Gi
is the amount of generation in megawatts at generator i. The objective function for the
entire power system can then be written as the sum of the quadratic cost model at each generator.
( ) f x a b P c P
i i Gi i Gi
i
( ) · + +
∑
2
This objective function will minimize the total system costs, and does not necessarily minimize
the costs for a particular area within the power system.
12
2.2.2 Equality constraints
The equality constraints of the OPF reflect the physics of the power system as well as the
desired voltage set points throughout the system. The physics of the power system are enforced
through the power flow equations which require that the net injection of real and reactive power
at each bus sum to zero.
( ) ( )
[ ] [ ]
( ) ( )
[ ] [ ]
P V V g b P P
Q V V g b Q Q
k k m km k m km k m
m
N
Gk Lk
k k m km k m km k m
m
N
Gk Lk
· · − + − − +
· · − − − − +
·
·
∑
∑
0
0
1
1
cos sin
sin cos
δ δ δ δ
δ δ δ δ
For a derivation of the power flow equations see Grainger and Stevenson [20, p. 330]. For an
explanation of forming the network admittance matrix see Grainger and Stevenson [20, p. 33].
It is also common for the power system operators to have voltage set points for each
generator. In this case, an equality constraint for each generator is added.
V V
Gi Gi setpo
− ·
int
0
Finally, for multiarea power systems, a contractual constraint requires that the net power
interchange be equal to the scheduled power interchange. This adds an equality constraint for all
but one area.
[ ] P P P P
km interchange sceduled interchange
tie lines
sceduled interchange
− · − ·
∑
0
This last area must not have the equality constraint and essentially becomes a “slack area.”
Consider a simple threearea system with scheduled interchanges as shown in Figure 2.1. With
the loads given and the interchanges constrained for two areas, the last area will be forced to have
its interchanges set as planned. Thus, adding another constraint is not needed. Indeed, it will
13
cause numerical problems if added because the equations associated with this constraint are
dependent on the other interchange constraints and would thus lead to a singular Hessian matrix.
AREA
ONE
AREA
TWO
AREA
THREE
P
12int
P
13int
P
23int
Load
1
Load
2
Load
3
Figure 2.1 Multiarea System with Scheduled Interchanges
2.2.3 Inequality constraints
The inequality constraints of the OPF reflect the limits on physical devices in the power
system as well as the limits created to ensure system security. Physical devices that require
enforcement of limits include generators, tap changing transformers, and phase shifting
transformers. This section will lay out all the necessary inequality constraints needed for the
OPF implemented in this thesis.
Generators have maximum and minimum output powers and reactive powers which add
inequality constraints.
P P P
Q Q Q
Gi Gi Gi
Gi Gi Gi
min max
min max
≤ ≤
≤ ≤
14
Load tap changing transformers have a maximum and a minimum tap ratio which can be
achieved and phase shifting transformers have a maximum and a minimum phase shift, which
can be achieved. Both of these create inequality constraints.
t t t
km km km
km km km
min max
min max
≤ ≤
≤ ≤ α α α
For the maintenance of system security, power systems have transmission line as well as
transformer MVA ratings. These ratings may come from thermal ratings (current ratings) of
conductors, or they may be set to a level due to system stability concerns. The determination of
these MVA ratings will not be of concern in this thesis. It is assumed that they are given.
Regardless, these MVA ratings will result in another inequality constraint. To make the
mathematics less complex, the constraint used in the OPF will limit the square of the MVA flow
on a transformer or transmission line.
S S
km km
2
2
0 − ≤
max
To maintain the quality of electrical service and system security, bus voltages usually have
maximum and minimum magnitudes. These limits again require the addition of inequality
constraints.
V V V
i i i min max
≤ ≤
2.2.4 Soft constraints by using penalty functions
One issue sometimes encountered when trying to solve a minimization problem is the non
existence of a feasible solution. Essentially this means that too many constraints have been
added to the problem and no solution exists which obeys all of the constraints. One way to avoid
15
this issue is to implement soft inequality constraints in the form of penalty functions. The word
“soft” signifies that the constraint is not absolutely enforced. The soft constraint only encourages
the solution to meet the constraint by enforcing a penalty if the constraint is not met. In the OPF
problem, soft equality constraints are not used, because of the nature of the equality constraints in
the OPF problem. The power flow equations can not be violated as they are imposed by physics,
and the generator set points of a power system are normally not moved around frequently. For
the inequality constraints, the penalty functions offer a viable option.
Penalty functions are added to the objective function of the minimization problem. Ideally, a
penalty function will be very small near a limit and increase rapidly as the limit is violated more.
A wellsuited penalty function for use in Newton’s method is the quadratic penalty function [19,
p. 443] which meets the requirements of a penalty function and is also easily differentiated for
use by Newton’s method. In the OPF presented in this thesis, soft inequality constraints are
available for transmission line MVA limits as well as bus voltage limits.
( )
( )
Penalty Functions
W k S S
W
k V V V V
V V V
k V V V V
km km km
i
i i i i
i i i
i i i i
· −
¸
¸
_
,
·
− <
≤ ≤
− >
¹
'
¹
¹
¹
¹
¹
¹
'
¹
¹
¹
¹
¹
¹
¹
¹
¹
2
2
2
2
2
0
max
min min
min max
max max
;
;
;
Figure 2.1 and Figure 2.2 show the graphs of these penalty functions used in the OPF.
16
V
i min
V
i max
V
i
Voltage Penalty
Function Value
Figure 2.1 Penalty Function for Bus Voltage
S
km max

2
S
km

2
Transmission Line
MVA Flow Penalty
Function Value
Figure 2.2 Penalty Function for Line MVA Flow Limit
Note that these penalty functions fit the requirements perfectly. While the inequality
constraint is not violated, the penalty function has a value of zero. As the constraint begins to be
violated, the penalty function quickly increases. Another advantage of the quadratic penalty
function is the ability to control how hard or soft to make the constraint. For very large values of
k, the quadratic penalty function behaves much like a hard constraint. By adjusting it to smaller
values, one can control the importance given to the limit.
After doing trial and error experimentation, values for k were chosen for use in the OPF of
this thesis. For soft bus voltage constraints, k was chosen to be $200/V
2
. For soft transmission
line constraints, k was chosen to be $100/MVA
4
.
17
2.2.5 Treatment of discrete variables
In all the discussions thus far, all variables have been assumed to be continuous. For
example, the constraint V V V
i i i min max
≤ ≤ allows V
i
to take on all values within the specified
range. The OPF algorithm as presented in this thesis also assumes this for the tap ratios and
phase shift angles of variable transformers, although this is not true for them. Variable
transformers have a fixed number of discrete tap positions at which they may operate. For
example, a typical tap changing transformer has 33 discrete positions (nominal, 16 above and 16
below). This problem should be addressed in future work.
One possible solution for this problem is to round the optimal setting found assuming a
continuous tap to the nearest discrete tap. This could be done for all transformers. However,
three problems arise from this methodology. First, there is no guarantee that the rounded
solution is the optimal solution. Second the solution may become infeasible after rounding, i.e.,
some constraints may be violated. Finally, this methodology will not work well for discrete
variables that have very large step sizes such as switched capacitor banks. This final problem is
addressed in references [21, 22].
2.2.6 Summary of optimal power flow problem
In summary, the optimal power flow problem can be written in the following form.
18
( ) minimize:
subject to:
penalties generators
a b P c P
P
Q
V V
P P
h x
S S
P P
P P
V V
V V
t t
t t
i i Gi i Gi i
k
k
i i set
scheduled
km km
Gi Gi
Gi Gi
i i
i i
km km
km km
km km
km km
+ + +
·
·
− ·
− ·
¹
;
¹
¹
¹
¹
¹
¹
¹
·
− ≤
− ≤
− ≤
− ≤
− ≤
− ≤
− ≤
− ≤
− ≤
¹
;
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
∑ ∑
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
α
α α
α α
int
max
max
min
max
min
max
min
max
min
( )
¹
¹
¹
¹
¹
¹
¹
≤ g x ( ) 0
It should be noted that the constraints on the reactive power at each generator are not included in
the problem as stated above. These constraints will be taken care of by treating a generator bus at
a Q limit as a load bus. This is commonly done in a power system when modeling generator
reactive power limits [23, p. 228].
2.2.7 Summary of Lagrangian terms
Given the problem statement for the OPF as shown in Section 2.2.5, the Lagrangian can be
written as a summation of several terms. These terms are summarized as follows:
19
( )
( )
( )
( )
( )
( )
( )
( )
λ
λ
λ
λ
λ
λ
λ
λ α α
λ α α
α
α
Skm km km
Pgih Gi Gi
PGil Gi Gi
Vih i i
Vil i i
tkmh km km
tkml km km
km km km
km km km
S S
P P
P P
V V
V V
t t
t t
2
2
−
¸
¸
_
,
−
−
−
−
−
−
−
−
¹
;
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
¹
max
max
min
max
min
max
min
max max
min min
*
*
*
*
*
*
*
*
*
inequality constraints
( )
( )
( )
( )
µ
µ
µ
µ
Pk k
Qk k
Viset i i set
scheduled
P
Q
V V
P P
−
−
¹
;
¹
¹
¹
¹
¹
¹
¹
int int
equality constraints
( )
( )
( )
a b P c P
k S S
k V V
k V V
i i Gi i Gi
generators
km km
i i
i i
+ +
¹
;
¹
−
¸
¸
_
,
−
−
¹
;
¹
¹
¹
¹
¹
¹
¹
∑
2
2
2
2
2
2
generator cost functions
penalty functions
max
min
max
*
*
*
Note: The terms with an asterisk (*) next to them are only included when the corresponding
constraints are being enforced
2.2.8 Calculation of gradient and Hessian
Given the terms of the Lagrangian, calculation of the gradient and Hessian is a very
straightforward process, albeit an extremely tedious one. For completeness, details on
calculating the gradient and Hessian are found in Appendices B, C, and D.
20
2.2.9 Solution of the optimal power flow
Once an understanding of the calculation of the Hessian and gradient is attained, the solution
of the OPF can be achieved by using the Newton’s method algorithm. The Newton’s method
algorithm is summarized in the flowchart in Figure 2.1 on page 10. The application of Newton’s
method to the OPF algorithm used in this thesis is summarized as follows:
Step 1. Initialize the OPF solution.
a. Initial guess at which inequalities are violated.
b. Initial guess z vector (bus voltages and angles, generator output power,
transformer tap ratios and phase shifts, all Lagrange multipliers).
Step 2. Evaluate those inequalities that have to be added or removed using the
information from Lagrange multipliers for hard constraints and direct evaluation
for soft constraints.
Step 3. Determine viability of the OPF solution. Presently this ensures that at least one
generator is not at a limit.
Step 4. Calculate the gradient and Hessian of the Lagrangian.
Step 5. Solve the equation [H]∆z = ∇L(z).
Step 6. Update solution z
new
= z
old
 ∆z.
Step 7. Check whether ∆z < ε. If not, go to Step 4, otherwise continue.
Step 8. Check whether correct inequalities have been enforced. If not go to Step 2. If so,
problem is solved.
21
2.3 A Sample Case Illustrating OPF Algorithm Process
In order to gain insight into the process which the OPF algorithm undergoes while
determining the optimum solution, a test was run on a 118bus system. The details of this system
are not included in this thesis because this case is only shown to demonstrate the solution process
described in the flowchart of Figure 2.1 on page 10.
The output to a message file during the solution of the 118bus system is shown in the
following.
1. Case OPF Case initialized
2. Starting OPF solution at 15:15:19
3. OPFitr 0 MaxMis: Controls: 25.7240 Voltages: 3113.4475 Angles: 3216.6133 Constraints: 5.8894
4. OPFitr 1 MaxMis: Controls: 0.0000 Voltages: 73.2151 Angles: 29.0626 Constraints: 0.1888
5. OPFitr 2 MaxMis: Controls: 0.0000 Voltages: 0.3480 Angles: 0.1868 Constraints: 0.0067
6. OPFitr 3 MaxMis: Controls: 0.0000 Voltages: 0.0025 Angles: 0.0007 Constraints: 0.0000
7. OPFitr 4 MaxMis: Controls: 0.0000 Voltages: 0.0001 Angles: 0.0002 Constraints: 0.0000
8. Adding Hard MVA Constraint from bus 64 to 65
9. Adding Soft Voltage (high) Constraint at bus 9
10. Adding BusQ Constraint for bus 12
11. Adding Soft Voltage (low) Constraint at bus 53
12. Bus 40 Gen at Max MW Limit
13. OPFitr 5 MaxMis: Controls: 10.0000 Voltages: 571.4009 Angles: 896.2391 Constraints: 1.6602
14. OPFitr 6 MaxMis: Controls: 0.0000 Voltages: 21.5599 Angles: 183.7967 Constraints: 0.3324
15. OPFitr 7 MaxMis: Controls: 0.0000 Voltages: 1.5178 Angles: 19.1469 Constraints: 0.0393
16. OPFitr 8 MaxMis: Controls: 0.0000 Voltages: 0.0233 Angles: 0.3324 Constraints: 0.0009
17. OPFitr 9 MaxMis: Controls: 0.0000 Voltages: 0.0005 Angles: 0.0003 Constraints: 0.0000
18. OPFitr 10 MaxMis: Controls: 0.0000 Voltages: 0.0001 Angles: 0.0004 Constraints: 0.0000
19. OPFitr 11 MaxMis: Controls: 0.0000 Voltages: 0.0001 Angles: 0.0003 Constraints: 0.0000
20. Successful OPF  Final Cost 51324.70 at 15:15:20
This output file can help in the understanding of the Newton’s method solution process.
Lines 1 and 2 initialize the OPF variables and assume a Lagrangian that contains no constraints.
Lines 3 to 7 then perform the inner loop of Newton’s method, which continuously solves [H]∆z
= ∇L(z) until convergence. After having reached convergence here, the inequality constraints are
checked to ensure that none are violated. As can be seen from lines 8 to 12, several inequality
constraints are found to be violated: a transmission line has exceeded its MVA capacity, a
22
generator has exceeded its MVAR limit, a generator has exceeded its MW limit, and two buses
are outside their desired ranges. With these violations noted, the Lagrangian is recalculated and
the inner loop is begun again. Lines 13 to 19 show the inner loop process proceeding
successfully again. After convergence is achieved again, the inequality constraints are again
checked. Line 20 shows that new violations were found and the successful final cost is shown.
2.4 Information Gained from the OPF Solution
The solution of the OPF, while difficult, has many great advantages over the classical
economic dispatch [11] of a power system. The OPF is capable of performing all of the control
functions necessary for the power system. While the economic dispatch of a power system does
control generator MW output, the OPF controls transformer tap ratios and phase shift angles as
well. The OPF also is able to monitor system security issues including line overloads and low or
high voltage problems. If any security problems occur, the OPF will modify its controls to fix
them, i.e., remove a transmission line overload.
Besides performing these enhanced engineering functions, probably the greatest advantage of
the OPF is the great wealth of knowledge it yields concerning the economics of the power
system. In studying the Lagrange multipliers associated with each constraint, one can show that
they can be interpreted as the marginal costs associated with meeting the constraint. A derivation
of this fact can be found in Appendix A. Therefore, the Lagrange multipliers, µ
PK
and µ
QK
,
can be seen as the marginal cost of real and reactive power generation at bus k in
$
MW hr
¸
1
]
1
and
$
MVAR hr
¸
1
]
1
, respectively. These prices could then be used to determine electricity prices at bus
23
k. On a larger level, the Lagrange multiplier, µ
int
, associated with the area interchange
constraint, can be seen as the marginal cost of allowing an area to break its interchange. If this
cost is positive, then the area would benefit from buying electricity, while if it is negative, the
area would benefit from selling electricity. These costs may be of use in determining the price
which an area would charge for a megawatt transaction with another area.
While these economic data are very helpful for realtime pricing algorithms, other data could
be used to help with transmission system planning. Consider the Lagrange multiplier associated
with a hard transmission line MVA constraint, λ
Skm
. This can be interpreted as the cost savings
per hour for each additional 1 MVA increase in a line’s rating. This information may be of use in
planning where new transmission lines will have the greatest economic impact on the power
system.
Much more information can be gained from a single OPF solution, but a great potential also
lies in the simulation of a power system over time using a power system simulator with an OPF
as its solution engine. This will be discussed in Chapter 4.
24
3. HEURISTICS OF THE OPTIMAL POWER FLOW SOLUTION
3.1 Classification of OPF Variables
While writing software to perform an OPF solution, a primary concern is identification of
variables during the process. Because of this, in order to handle the variables in the OPF
problem efficiently, it is convenient to separate them into three categories: controls, states, and
constraints. The control variables correspond to quantities that can be arbitrarily manipulated,
within their limits, in order to minimize the costs. These include generator MW outputs,
transformer tap ratios, and transformer phase shift angles. The states correspond to quantities
that are set as a result of the controls, but must be monitored. They are also of interest at the
solution. The states include all system voltages and angles. Finally, the constraint variables are
variables associated with the constraints. These include all the Lagrange multipliers. The
variables in the OPF problem are summarized in Table 3.1.
Table 3.1 OPF problem variables
Variable Classification Variables in Classification
Control P
Gk
, t
km
, and α
km
State V
k
and δ
k
Constraint µ
Pk
, µ
Qk
, µ
viset
, µ
int
, λ
Skm
, λ
PGih
,
λ
PGil
, λ
Vih
,
λ
Vil
, λ
tkm max
, λ
tkm min
, λ
α
km max
, and
λ
α
km min
In addition to the OPF variables, it is also important to keep track of any added soft constraints.
For hard constraints, the Lagrange multiplier is monitored, but for soft constraints, a fourth
25
variable classification is added: penalty. No variable data are stored for this classification. It
simply serves as a place holder for the penalty function.
3.2 Implementation of Sparse Matrix Techniques
At the heart of the solution of the OPF using Newton’s method is the solution of the linear
system of equations, [H]∆z = ∇L(z). During the Newton’s method solution process, this system
of equations is solved repeatedly. Because of this, and because one of the primary objectives of
an OPF is to find its solution in a short amount of time, the speed of the solution of this linear
system of equations is essential to a successful OPF solution. Fortunately, as is the case with
many power system matrices, the Hessian matrix is extremely sparse. By implementing sparse
matrix routines, the equations can be quickly solved.
Much of the structure of the Hessian matrix for the OPF problem is known prior to any
calculation. In general, for any Newton’s method routine, the Hessian takes the following form
as discussed in Section 2.1.2:
H =
W J
J
T
0
¸
1
]
1
Already, it is seen that the lower right quadrant of the matrix has entries of zero. By a closer
examination of the Hessian structure for the OPF problem, advantage can be taken of this
sparsity.
The first step was to reorder the variables to include the generator megawatt controls first,
followed by the remaining controls, states, and then constraints. This resulted in the following
structure.
26
H =
W n
W J
n J
PG PG
T
remaining remaining
T
PG remaining
0
0
0
¸
1
]
1
1
1
where W
PG
is a purely diagonal matrix with entries of 2c
i
, n
PG
is a matrix with a (1) entry in each
column corresponding to the net power injection constraint, and possibly an entry of (t1)
in a column corresponding to a generator with a maximum or a minimum MW constraint
being enforced, and W
remaining
and J
remaining
are the remaining parts of the Hessian matrix.
Due to this structure, the sparse LU factorization up to the rows of W
remaining
, using Gauss’s
method [24], will result in the following matrix.
H
modified
=
W W n
W J
n J n W n
PG PG PG
T
remaining remaining
T
PG remaining PG PG PG
T
0
0
1
1
−
−
−
¸
1
]
1
1
1
W
remaining
and J
remaining
are unaffected by these steps and fills are only created by the term
n W n
PG PG PG
T −1
. In order to determine the fills created by this matrix, consider only its zero/non
zero structure. One will find that most fills created by this term are actually desired.
Because W
PG
is a diagonal matrix, it will not affect the zero/nonzero structure of the
resulting matrix; therefore, only consider the matrix n n
PG PG
T
. If no generators are at a megawatt
limit, then each column of n
PG
will have only a single entry of (1). In this situation, the matrix
will result in only diagonal entries in the rows corresponding to the Lagrange multipliers µ
Pk
.
The diagonal entries are needed for factorization anyway, so this is a desired event. If, however,
some generators are at a megawatt limit, then the matrix n n
PG PG
T
will have some offdiagonal
entries. These entries will occur in offdiagonal pairs such as (row λ
PGih
,column µ
Pk
), and
(row µ
Pk
, column λ
PGih
) .
27
Because generators will not be normally at their limits, and because only two nondiagonal
fills are created for each generator at a limit, this ordering can help speed the solution. This
ordering is what is presently done in the OPF source code written for this thesis.
In studying the Hessian matrix structure further, one can see that the Hessian rows
corresponding to the Lagrange multipliers corresponding to the voltage set points contain only
one entry, a (+1). In order to exploit this structure, consider ordering the variables in the
following manner: generator MW outputs first, voltage states corresponding to generators
second, Lagrange multipliers corresponding to generator voltages (in same order as the states)
third, and finally the remaining states and constraints. This results in the following structure for
the Hessian matrix
H =
W n
W I W J
I
W W J
n J J
PG PG
T
vset rvset
T
rvset
T
rvset r rr
T
PG rvset rr
0 0 0
0
0 0 0 0
0 0
0 0
¸
1
]
1
1
1
1
1
1
where I is the identity matrix.
Given this ordering, switching the second and thirdrow partitions around yields the following.
H =
W n
I
W I W J
W W J
n J J
PG PG
T
vset rvset
T
rvset
T
rvset r rr
T
PG rvset rr
0 0 0
0 0 0 0
0
0 0
0 0
¸
1
]
1
1
1
1
1
1
Given this ordering, the sparse LU factorization up to the rows of W
r
, using Gauss’s method [24],
will result in the following matrix.
28
H
modified
=
W W n
I
W I W J
W W J
n J J n W n
PG PG PG
T
vset rvset
T
rvset
T
rvset r rr
T
PG rvset rr PG PG PG
T
0 0 0
0 0 0 0
0
0 0
0
1
1
−
−
−
¸
1
]
1
1
1
1
1
1
The fills created by this method would be identical to those created by the previous ordering.
However, in addition, the sparse matrix routines could also take advantage of simply “skipping”
over the processing of the second and thirdrow partitions because the only operations required
by Gauss’s method would be divisions by 1.
This ordering could be implemented and would result in increased solution speed; however, it
does require row pivoting to be done by the sparse matrix solution routines. At present, the
implementation of this ability into the OPF source code has not been done, but it can be added at
a later time. Presently, the OPF source code does ordering with generator megawatt output
controls first.
3.3 Determination of the Set of Binding Inequality Constraints
When reading any paper written about the application of Newton’s method to the OPF
problem, one runs into discussions of finding the binding set of inequalities [13]. The reason for
this emphasis is the role this process has in determining the speed of an OPF solution, and thus
ultimately its usefulness. In referring to Figure 2.1 showing the Newton’s method solution
algorithm, one sees that this determination of the binding set makes up the outer loop of the
flowchart. In a typical OPF problem, this outer loop will be executed about three or four times.
Therefore, even removing one of these iterations could save 25  33% of the execution time
assuming that the inner loop time remains relatively constant.
29
For the OPF source code written for this thesis, the following process was used to determine
if the correct inequalities had been enforced.
Step 1. Check if any inequality constraints can be removed. An inequality constraint can
be removed if the Lagrange multiplier associated with it is negative.
a. Hard line MVA constraints
b. Hard bus voltage constraints
c. Maximum and minimum generator megawatt outputs
d. Maximum and minimum transformer tap ratios
e. Maximum and minimum transformer phase shifts.
Step 2. Determine if any line MVA limits are violated. If they are, add a hard constraint
or soft penalty function depending on user’s preference.
Step 3. Determine if any generator reactive power limits are being violated. For those
that are violated, change the generator bus to a load bus [23, p. 228].
Step 4. Check if any bus voltage limits are being violated. If they are, add a hard
constraint or soft penalty function depending on user’s preference.
Step 5. Check all control variables to see if they are operating within their limits. If they
are not, add a constraint.
a. Maximum and minimum generator MW outputs
b. Maximum and minimum transformer tap ratios
c. Maximum and minimum transformer phase shifts.
Using this process, the OPF will determine which inequality constraints to enforce.
30
3.4 Searching Algorithms
While performing the OPF solution, it is often necessary to search through the list of controls,
states, and constraints for a particular variable. While for small systems the use of a simple
linked list is adequate, when the system becomes large, this searching can begin to dominate the
CPU time spent, because with a linked list, the average search time is proportional to half of the
number of elements in the list (N/2). It is therefore very useful to take advantage of a data
structure suited for repeated searches. One such data structure is the binary tree. A sample
binary tree is shown in Figure 3.1.
8
5 12
3 10 7 13
1 4 6
Figure 3.1 Sample Binary Tree
A binary tree obeys very simple rules which make searching for an individual element of the
tree a very quick process. At each node, all values to the left of the node are smaller, and all
values to the right of the node are larger. In this way, it can be shown that the maximum search
time for an element of the tree is proportional to the base two logarithm of the number of
elements (log
2
N). The saving in time is substantial when the number of nodes becomes very
31
large. Consider a system which has 1024 elements: 1024/2 = 512 and log
2
1024 = 10. The
binary tree results in an average search which is 50 times faster.
Applying this thinking to OPF variables requires that they be given an ordering so that
“smaller than” and “larger than” comparisons can be made. In the OPF source code, the
variables are separated, as discussed in Section 3.1, into controls, states, and constraints. For the
ranking in this thesis, precedence order is controls, states, and then constraints. Within each
variable type precedence is defined as follows:
Controls
1. Bus number (from bus for transformer tap or phase shift)
2. (To bus number for transformer tap or phase shift)
3. Type of control with ranking: P
Gk
, t
km
, and α
km
.
States
1. Bus number
2. Type of state with ranking: V
k
and δ
k
.
Constraints
1. Bus number (or area number for interchange constraints)
2. (To bus number for line constraints)
3. Type of constraint with ranking: µ
Pk
, µ
Qk
, µ
viset
, µ
int
, λ
Skm
, λ
Pgih
, λ
PGil
,
λ
Vih
, λ
Vil
, λ
tkm max
, λ
tkm min
, λ
α
km max
, and λ
α
km min
.
The binary tree structure can then be applied to these structures with this ranking. Using the
binary tree data structure greatly aids in reducing the amount of time spent searching.
3.5 Solution of an OPF Repeatedly Over Time
One of the great opportunities of an OPF solution is the ability to simulate a power system
over time while keeping it at its optimal solution. This will be discussed in the next section, but
heuristics for increasing the computational speed of this process will be discussed here.
32
As mentioned in Section 2.1, the convergence of Newton’s method algorithm is very rapid
near the solution. Because of this, the initial guess for the variables is very important. It was
also mentioned in the previous section that the speed of the OPF is greatly influenced by how
quickly the binding set of inequalities is found. Both of these important parameters for solving
the OPF quickly can be met when simulating a power system over time.
By assuming that the power system will not undergo drastic change over the next time step,
the output of one OPF solution (the Hessian, gradient, voltages, angles, Lagrange multipliers, as
well as which inequalities are binding) can be fed as the initial guess for the next OPF solution.
At present, however, the Hessian and gradient of the Lagrangian are recalculated from scratch at
each time step of the simulation. The search for the binding inequalities is also done from
scratch for each time step. By keeping the solved OPF solution, the speed of solution in the
simulation environment will be greatly increased. These modifications will greatly enhance the
power system simulation over time.
In calculating the Hessian a time step, further advantage could be gained by implementing
partial refactorization schemes for the Hessian as seen in [25]. These schemes only update those
elements of the L and U factors of the Hessian that are affected by changes to the Hessian. While
this holds great promise for future work, these schemes have not yet been implemented into the
OPF written for this thesis.
33
4. USES OF AN OPTIMAL POWER FLOW IN A POWER SYSTEM
SIMULATION ENVIRONMENT
While a single OPF solution yields valuable information regarding a power system, the
implementation of the OPF into a power system simulation environment holds even greater
promise. In this environment, simulation of a system over time can be done while maintaining it
at its optimal condition. In this way, a vast amount of economic data can be gleaned from the
simulation. This chapter will give several examples of the use of the OPF code as implemented
into the power system simulator, POWERWORLD [17].
4.1 Example Line Overload Removal
A simple power system not operating under the OPF control is shown in Figure 4.1. For
details concerning this power system see Appendix E.
91 MW
92 MVR
100 MW
100 MW
20 MVR
196 MW
27 MVR
1.00 PU
1.00 PU
1.00 PU
Bus 3
Bus 2
Bus 4
100 MW
20 MVR
20 MVR
20 MVR
100 MW
86 MW
237 MW
23 MVR
60 MVR
1.00 PU
Bus 5
Bus 6
0.97 PU
0.97 PU
100 MW
50 MVR
Bus 1
100 MW
10 MVR
0.00 $/MWH
0.00 $/MWH
0.00 $/MWH
0.00 $/MWH
0.00 $/MWH
0.00 $/MWH
0.00 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
147%
Figure 4.1 Sixbus, SingleArea System Not on OPF Control
34
In order to remove the line constraint, the OPF control is turned on, and the line overload is
removed as shown in Figure 4.2.
122 MW
76 MVR
100 MW
100 MW
20 MVR
190 MW
25 MVR
1.00 PU
1.00 PU
1.00 PU
Bus 3
Bus 2
Bus 4
100 MW
20 MVR
20 MVR
20 MVR
100 MW
116 MW
179 MW
48 MVR
43 MVR
1.00 PU
Bus 5
Bus 6
0.97 PU
0.97 PU
100 MW
50 MVR
Bus 1
100 MW
10 MVR
17.23 $/MWH
14.93 $/MWH
13.25 $/MWH
12.77 $/MWH
16.03 $/MWH
15.02 $/MWH
0.00 $/MVRH
0.65 $/MVRH
0.00 $/MVRH
0.12 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
100%
Figure 4.2 Sixbus, SingleArea System on OPF Control
As can be seen, the generators have redispatched themselves in order to remove the line
overload. Further analysis of this process from an economic viewpoint will be discussed in the
next section.
4.2 Use for Bus Real and Reactive Power Pricing
To illustrate the real and reactive power pricing potentials of the OPF solution, the same
system shown in Figure 4.1 is placed on OPF control with its line limit raised. See Figure 4.1.
Note that the OPF results in the same dispatch seen in Figure 4.1 now that the line limit has been
removed.
35
91 MW
92 MVR
100 MW
100 MW
20 MVR
196 MW
27 MVR
1.00 PU
1.00 PU
1.00 PU
Bus 3
Bus 2
Bus 4
100 MW
20 MVR
20 MVR
20 MVR
100 MW
86 MW
237 MW
23 MVR
60 MVR
1.00 PU
Bus 5
Bus 6
0.97 PU
0.97 PU
100 MW
50 MVR
Bus 1
100 MW
10 MVR
14.85 $/MWH
14.19 $/MWH
13.37 $/MWH
13.85 $/MWH
15.25 $/MWH
15.18 $/MWH
0.00 $/MVRH
0.23 $/MVRH
0.00 $/MVRH
0.11 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
Figure 4.1 Line Limit from Bus 5 to 4 Raised to 100 MVA
Comparison of the OPF solutions in Figure 4.2 and Figure 4.1 yield valuable insight. The
total system cost without the line limited as in Figure 4.1 is $7824/hr. With the line limited as in
Figure 4.2 this costs increases, as would be expected, to $7895/hr. Also note the differences
between the bus MW marginal costs in Figure 4.2 and Figure 4.1. Because the generators at
buses 2 and 4 were forced to decrease their output in order to remove the overload, their bus MW
marginal costs also decreased. Conversely, the bus MW marginal costs at buses 1, 3, and 5
increased. As might be expected, the largest changes occurred at the ends of the limited line,
buses 4 and 5.
36
4.3 Use for Area Real Power Pricing
The OPF solution method may also be used with multiarea power systems. The OPF will
enforce the scheduled area interchange in these systems. In Figure 4.1, the simple sixbus system
from before is split into two areas as shown.
161 MW
62 MVR
100 MW
100 MW
20 MVR
133 MW
2 MVR
1.00 PU
1.00 PU
1.00 PU
Bus 3
Bus 2
Bus 4
100 MW
20 MVR
20 MVR
20 MVR
100 MW
143 MW
169 MW
50 MVR
23 MVR
1.00 PU
Bus 5
Bus 6
0.97 PU
0.98 PU
100 MW
50 MVR
Bus 1
100 MW
10 MVR
16.75 $/MWH
15.86 $/MWH
12.15 $/MWH
12.58 $/MWH
16.71 $/MWH
13.39 $/MWH
0.00 $/MVRH
0.24 $/MVRH
0.00 $/MVRH
0.09 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
Area Two
Area One
Figure 4.1 Sixbus, TwoArea System on OPF Control
As can be seen, for this case the generation in Area Two is less expensive than in Area One;
therefore, it would be advantageous for both areas if Area One were to purchase some power
from Area Two. Using the capabilities of the POWERWORLD Area Transactions/Information
display [17], scheduled transactions can be set up between the two areas to optimize their costs.
Table 4.1 summarizes several possibilities.
37
Table 4.1 Sample transaction scenarios
Transaction
[MW]
AreaOne Cost
[$/hr]
AreaTwo Cost
[$/hr]
Sum of Both Areas
[$/hr]
None 4564 3496 8060
50.0 4489 3423 7912
65.5 4482 3413 7895
70.0 4481 3415 7896
80.0 4481 3428 7909
As can be seen, the least expensive scenario for the sum of the areas is when an interchange of
65.5 MW is undertaken. This transaction scenario is shown in Figure 4.2.
122 MW
76 MVR
100 MW
100 MW
20 MVR
190 MW
25 MVR
1.00 PU
1.00 PU
1.00 PU
Bus 3
Bus 2
Bus 4
100 MW
20 MVR
20 MVR
20 MVR
100 MW
116 MW
179 MW
48 MVR
43 MVR
1.00 PU
Bus 5
Bus 6
0.97 PU
0.97 PU
100 MW
50 MVR
Bus 1
100 MW
10 MVR
17.23 $/MWH
14.93 $/MWH
13.25 $/MWH
12.77 $/MWH
16.03 $/MWH
15.02 $/MWH
0.00 $/MVRH
0.65 $/MVRH
0.00 $/MVRH
0.12 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
Area Two
Area One
100%
Figure 4.2 Transaction of 65.5 MW Undertaken
Comparing the OPF solution in Figure 4.2 and Figure 4.2 one can see that they show the
same solution. This is of course not unexpected, because the twoarea economic negotiations
should yield the same solution that an OPF solution disregarding areas would. This must happen
so that all areas are at their optimal solution.
38
4.4 Example Transformer Tap Control
In order to test the transformer tap control, a more complex power system was used as in
Figure 4.1. This system has 23 buses, 10 generators, 23 transmission lines, and 5 transformers.
Three of these transformers are load tap changing (LTC) transformers: buses 5 to 25, 204 to 224,
and 102 to 122. For further details, see Appendix F. The system as shown in Figure 4.1 is on
OPF control, but the ability to control the tap ratios is turned off and all tap ratios are set at 1 p.u.
The system as in Figure 4.2 has activated the tap ratio control functions of the OPF, and the tap
ratios have been manipulated so as to minimize the objective function.
Bus 1
Bus 2
Bus 3
Bus 4
Bus 5
Bus 6
Bus 7
Bus 8
Bus 9
Bus 23
Bus 25
Bus 28
Bus 101
Bus 102
Bus 103
Bus 104
Bus 105
Bus 122
Bus 201
Bus 202
Bus 203
Bus 204
Bus 224
1.03 PU
700 MW
464 MW
1.03 PU
200 MW
763 MW
450 MW
0.98 PU
1.02 PU
200 MW
0.99 PU
350 MW
543 MW
0.97 PU
300 MW 0.96 PU
200 MW
1.03 PU
300 MW 200 MW
657 MW
1.03 PU
1.01 PU
1.01 PU
359 MW
1.01 PU
250 MW
0.98 PU
75 MW
1.01 PU
350 MW
1.03 PU
430 MW
0.99 PU
100 MW
1.00 PU
150 MW
340 MW
1.01 PU
0.99 PU
225 MW
1.02 PU
175 MW
400 MW
300 MW
0.98 PU
1.02 PU
432 MW
125 MW
1.03 PU
97 MVR
106 MVR
106 MVR
98 MVR
106 MVR
97 MVR
16.55 $/MWH
13.90 $/MWH
13.57 $/MWH
14.52 $/MWH
17.25 $/MWH
16.64 $/MWH
16.51 $/MWH
15.58 $/MWH
16.50 $/MWH
15.22 $/MWH
13.39 $/MWH
13.49 $/MWH
13.16 $/MWH
13.68 $/MWH
13.53 $/MWH
13.44 $/MWH
13.51 $/MWH
13.49 $/MWH
13.61 $/MWH
12.86 $/MWH
13.27 $/MWH
12.48 $/MWH
12.86 $/MWH
1.000 tap
1.000 tap
1.000 tap
Figure 4.1 TwentyThree Bus System on OPF Control with Tap Control Off
39
Bus 1
Bus 2
Bus 3
Bus 4
Bus 5
Bus 6
Bus 7
Bus 8
Bus 9
Bus 23
Bus 25
Bus 28
Bus 101
Bus 102
Bus 103
Bus 104
Bus 105
Bus 122
Bus 201
Bus 202
Bus 203
Bus 204
Bus 224
1.03 PU
700 MW
463 MW
1.03 PU
200 MW
761 MW
450 MW
0.98 PU
1.02 PU
200 MW
0.99 PU
350 MW
542 MW
0.97 PU
300 MW 0.96 PU
200 MW
1.03 PU
300 MW 200 MW
657 MW
1.03 PU
1.01 PU
1.01 PU
360 MW
1.01 PU
250 MW
0.98 PU
75 MW
1.01 PU
350 MW
1.03 PU
430 MW
0.99 PU
100 MW
1.00 PU
150 MW
340 MW
1.01 PU
0.98 PU
225 MW
1.02 PU
175 MW
401 MW
300 MW
0.98 PU
1.02 PU
433 MW
125 MW
1.05 PU
97 MVR
106 MVR
106 MVR
98 MVR
106 MVR
97 MVR
16.54 $/MWH
13.91 $/MWH
13.58 $/MWH
14.51 $/MWH
17.24 $/MWH
16.63 $/MWH
16.51 $/MWH
15.58 $/MWH
16.50 $/MWH
15.22 $/MWH
13.39 $/MWH
13.50 $/MWH
13.17 $/MWH
13.69 $/MWH
13.54 $/MWH
13.44 $/MWH
13.52 $/MWH
13.50 $/MWH
13.62 $/MWH
12.86 $/MWH
13.28 $/MWH
12.49 $/MWH
12.86 $/MWH
0.965 tap
1.012 tap
1.010 tap
Figure 4.2 TwentyThree Bus System with Tap Control On
A summary of the tap ratios and resulting total system costs are shown in Table 4.1.
Table 4.1 Summary of tap ratio control experiment
With tap ratio control OFF With tap ratio control ON
Tap ratio from bus 5 to 25 1.000 p.u. 1.012 p.u.
Tap ratio from bus 102 to 122 1.000 p.u. 0.965 p.u.
Tap ratio from bus 204 to 224 1.000 p.u. 1.010 p.u.
Total System Cost $ 50333.25 per hour $ 50330.55 per hour
As expected, the ability to control tap ratios has lowered the total system cost. It should be
noted, however, that the control of the tap ratios will normally not drastically reduce the system
costs. In reality, tap ratio control allows control of the reactive power flow thus reducing losses.
Even given perfect reactive power control, system losses will only be reduced a small amount;
therefore, tap ratio control will not have a huge effect.
40
4.5 System MVAR Control Using Transformer Taps [26]
As previously mentioned, a tap changing transformer is a reactive power controller. It is this
ability which allows it to help control system voltages. Because it controls reactive power, it is
also able to reduce system losses by directing the reactive power in a manner which reduces
system currents. Because reactive power and real power are not completely decoupled,
controlling reactive power can also help with real power control. In this way, a tap changing
transformer can help lower generation costs both by reducing system losses and by allowing less
expensive generation to operate when it could not have without tap changing control. In this
thesis, the OPF has been configured to allow continuous tap values although in reality one
encounters discrete tap values (often 33 steps). While modelling the tap as continuous is much
easier to implement, it also is better for analyzing the potential abilities of tap changing
transformers without the complication of the discrete jumps which would normally occur.
In order to demonstrate a transformer’s ability to control the system, consider the sevenbus,
onearea system shown in Figure 4.1 (data in Appendix G).
Bus 2
Bus 3
Bus 4 Bus 5
Bus 7
Bus 1
Bus 6
0.99 PU
1.00 PU
1.00 PU 1.04 PU
1.01 PU
1.01 PU
1.04 PU
50 MW
10 MVR
50 MW
10 MVR
100 MW
30 MVR
100 MW
30 MVR
50 MVR
50 MW
10 MVR
50 MW
10 MVR
150 MW
60 MVR
1.0000 tap
1.0000 tap
136 MW
81 MVR
431 MW
59 MVR
20.11 $/MWH
0.02 $/MVRH
18.51 $/MWH
0.36 $/MVRH
18.50 $/MWH
0.28 $/MVRH
18.18 $/MWH
0.21 $/MVRH
19.07 $/MWH
0.00 $/MVRH
16.75 $/MWH
0.00 $/MVRH
18.19 $/MWH
0.32 $/MVRH
Figure 4.1 SevenBus, OneArea System with Taps Inactive
41
The tap changing ability of the transformers in this system has been deactivated and the OPF
solution is shown. This system has a total system cost of $8,059/hr and system losses of 16.22
MW.
Now consider the system with the tap changing abilities activated. The OPF solution is
shown in Figure 4.2, and a comparison of the two systems is shown in Table 4.1. Comparing the
data in this table, it can be seen that by utilizing the variable tap settings, the system is able to
Bus 2
Bus 3
Bus 4 Bus 5
Bus 7
Bus 1
Bus 6
0.98 PU
1.04 PU
1.02 PU 1.04 PU
1.05 PU
0.99 PU
1.04 PU
50 MW
10 MVR
50 MW
10 MVR
100 MW
30 MVR
100 MW
30 MVR
52 MVR
50 MW
10 MVR
50 MW
10 MVR
150 MW
60 MVR
0.9447 tap
0.9497 tap
127 MW
32 MVR
439 MW
105 MVR
19.93 $/MWH
0.08 $/MVRH
18.61 $/MWH
0.25 $/MVRH
18.60 $/MWH
0.25 $/MVRH
18.23 $/MWH
0.21 $/MVRH
18.80 $/MWH
0.00 $/MVRH
16.90 $/MWH
0.00 $/MVRH
18.23 $/MWH
0.21 $/MVRH
Figure 4.2. SevenBus, OneArea, System with Taps Active
Table 4.1 Comparison of two systems
Without Taps With Taps
Bus 1 Generation 431 MW
59 MVAR
439 MW
105 MVAR
Bus 7 Generation 136 MW
81 MVAR
127 MW
32 MVAR
System Costs $8,059/hr $8,032/hr.
System Losses 16.22 MW 15.73 MW
Tap at bus 24 1.000 p.u. 0.9447 p.u.
Tap at bus 35 1.000 p.u. 0.9497 p.u.
42
increase the cheaper generation at bus 1 and ship it towards the bottom of the system while
decreasing system losses. This has been done at the expense of greater variability of system
voltages. However, all voltages are still within 5% of nominal.
While this example displays the use of the tap changing transformer as a reactive power
controller, greater insight can be gained by considering the same system divided into two areas as
shown in Figure 4.3. The generation in area one is much less expensive than generation in area
two. Therefore it would be beneficial to both parties to undergo a megawatt transaction. Various
transactions were undertaken both with the TCUL ability active and with it inactive. Summaries
of these results are shown in Table 4.2 and Table 4.3.
Bus 2
Bus 3
Bus 4 Bus 5
Bus 7
Bus 1
Bus 6
1.00 PU
0.99 PU
0.99 PU 1.04 PU
1.00 PU
1.01 PU
1.04 PU
50 MW
10 MVR
50 MW
10 MVR
100 MW
30 MVR
100 MW
30 MVR
49 MVR
50 MW
10 MVR
50 MW
10 MVR
150 MW
60 MVR
1.0000 tap
1.0000 tap
314 MW
4 MVR
253 MW
145 MVR
28.39 $/MWH
0.14 $/MVRH
14.22 $/MWH
0.04 $/MVRH
14.20 $/MWH
0.11 $/MVRH
28.10 $/MWH
0.10 $/MVRH
24.42 $/MWH
0.00 $/MVRH
13.56 $/MWH
0.00 $/MVRH
28.09 $/MWH
0.02 $/MVRH
Area Two
Area One
Figure 4.3. SevenBus, TwoArea System with TapChanging Transformers
Table 4.2 Results for various interchanges with transformer taps left inactive
Interchange
Area 1 2 [MW
at $/MWhr]
Bus 1
Gen
[MW]
Bus 1
Gen
[MVAR]
Bus 7
Gen.
[MW]
Bus 1
Gen
[MVAR]
Tap
Bus 24
[p.u.]
Tap
Bus 35
[p.u.]
Cost at
Area 1
[$/hr]
Cost at
Area 2
[$/hr]
Total
Cost
[$/hr]
Losses
Area 1
[MW]
Losses
Area 2
[MW]
Total
Losses
[MW]
none 253 145 314 4 1.0000 1.0000 2,949 6,308 9,257 3.45 14.16 17.61
+50 at 18.84 304 114 258 17 1.0000 1.0000 2,718 5,923 8,641 4.19 7.88 12.07
+50 at 18.46 357 88 204 42 1.0000 1.0000 2,582 5,667 8,249 6.80 4.14 10.94
+50 at 18.12 411 66 153 71 1.0000 1.0000 2,543 5,529 8,072 11.27 2.84 14.11
+19 at 17.84 432 58 134 82 1.0000 1.0000 2,555 5,504 8,059 13.46 2.98 16.44
+31 at 17.84 468 48 104 103 1.0000 1.0000 2,603 5,500 8,103 17.60 3.96 21.56
+50 at 17.37 526 34 58 139 1.0000 1.0000 2,779 5,559 8,338 25.83 7.51 33.34
Table 4.3 Results for various interchanges with transformer taps active
Interchange
Area 1 2 [MW
at $/MWhr]
Bus 1
Gen
[MW]
Bus 1
Gen
[MVAR]
Bus 7
Gen.
[MW]
Bus 1
Gen
[MVAR]
Tap
Bus 24
[p.u.]
Tap
Bus 35
[p.u.]
Cost at
Area 1
[$/hr]
Cost at
Area 2
[$/hr]
Total
Cost
[$/hr]
Losses
Area 1
[MW]
Losses
Area 2
[MW]
Total
Losses
[MW]
none 253 104 314 36 1.0432 1.0466 2,938 6,299 9,237 2.60 13.79 16.39
+50 at 18.83 304 98 258 34 1.0137 1.0209 2,716 5,921 8,637 4.01 7.84 11.85
+50 at 18.45 357 97 204 32 0.9851 0.9933 2,584 5,664 8,248 6.86 4.04 10.90
+50 at 18.12 411 101 152 32 0.9578 0.9645 2,542 5,514 8,056 11.15 2.14 13.29
+25 at 17.83 439 105 127 32 0.9448 0.9499 2,557 5,475 8,032 13.84 1.86 15.70
+25 at 17.83 467 105 102 38 0.9340 0.9411 2,593 5,462 8,055 16.91 1.99 18.90
+50 at 17.31 524 99 54 60 0.9265 0.9280 2,754 5,491 8,245 24.19 3.73 27.92
44
A comparison of Table 4.2 and Table 4.3 shows reduced costs at every step for the system
with the tap changing transformer active. From a total system cost point of view, the two areas
should stop undergoing transactions after having reached the transactions shown in bold type
(minimum total cost). This is after a transaction of 169 MW for the system with taps inactive,
and a transaction of 175 MW for the system with taps active. Note that after undergoing these
transactions, the systems are at the same state found by running the OPF with no area constraints
enforced.
The information in these tables illustrates the concept that the tap changing transformer is a
reactive power controller. With taps inactive as in Table 4.2, the only elements of the system that
have control over the reactive power flows are the two generators. As a result, the variation in
the injected reactive power from the generators is extremely large. With the taps active as in
Table 4.3, the transformers take over as the reactive power controllers, and as a result, the
generators’ reactive power injections are able to remain relatively constant.
It should be noted that if a generator were to be treated as an MVA limited machine, as
opposed to an MW and MVAR (separately) limited machine, then this reactive power control by
the tap changing transformers would free up the generator for more megawatt production.
4.6 Transmission Line Valuation by TimeDomain Simulation
One of the most intriguing potential uses of an OPF is its use as a pricing tool by doing time
domain simulations of a power system. As a simple example, consider again the sixbus, one
area power system of Appendix E again. Assume that the loads of the system vary as shown in
45
Figure 4.1. The load at a given bus is then equal to its base value shown in Appendix D
multiplied by the load factor.
Load Factor vs. Time
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0:00 4:00 8:00 12:00 16:00 20:00 0:00
Time
L
o
a
d
F
a
c
t
o
r
Figure 4.1 Load Factor Curve for sixbus System
Entering this load factor curve into the POWERWORLD simulation software and simulating the
power system for a full 24 hours yield a total cost in dollars for operating the power system over
that period of time. By resimulating the system repeatedly with variations in system structure,
one can gain useful economic insights from the comparison of total system costs. Note that the
case was simulated at 600 times real time for the first 23 hours and at 60 times real time for the
last hour. The POWERWORLD simulation software utilizes trapezoidal rule integration to
calculate the total system cost from the incremental costs at each time point.
Using this technique, the sixbus system was simulated with various transmission lines
removed. The results are summarized in Table 4.1.
Table 4.1 Simulation data for sixbus system with transmission lines removed
System Configuration
Total Cost
for 24 Hour
Simulation
Difference from
No Change
Configuration
Percent Increase
from No Change
Configuration
No Changes $ 148,240 N/A N/A
Line from Bus 3 to Bus 5 Removed $ 148,377 $137.00 0.09%
Line from Bus 3 to Bus 4 Removed $ 148,632 $392.00 0.26%
Line from Bus 1 to Bus 2 Removed $ 149,311 $1,071.00 0.72%
Line from Bus 4 to Bus 5 Removed $ 150,309 $2,069.00 1.40%
Time Load
Factor
Time Load
Factor
0:00 0.6 13:15 1.1
2:30 0.4 16:30 1.2
5:30 0.4 19:15 1.1
8:00 0.6 22:00 0.8
10:30 0.8 23:59 0.6
46
By comparing the total system cost with all lines connected to the cost with a line removed, one
can find the approximate cost a company would incur per day while taking a line out for
maintenance. It is very important to note, however, that this does in no way take into account the
possibility that another line may unexpectedly go out, in which case all lines may be needed to
take up the slack. In other words, the pricing of redundancies in the transmission system is not
considered in this methodology. Regardless of this, the information is very valuable and could
be used to help with maintenance scheduling, or even possibly in helping to determine where to
build new transmission lines in the future.
4.7 Capacitor Bank Valuation by TimeDomain Simulation
The technique discussed in Section 4.5 can also be applied to the valuation of any
transmission system component. As an example, consider capacitor bank valuation for the
twentythree bus, onearea power system of Appendix E. Assume that the loads of the system
vary as shown in Figure 4.1.
Load Factor vs Time
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0:00 4:00 8:00 12:00 16:00 20:00 0:00
Time
L
o
a
d
F
a
c
t
o
r
Figure 4.1 Load Factor Curve for TwentyThree Bus System
Time Load
Factor
Time Load
Factor
0:00 1.000 13:15 1.275
2:30 0.800 16:30 1.350
5:30 0.800 19:15 1.275
8:30 0.950 22:00 1.100
10:30 1.100 23:59 1.000
47
Again, entering this load factor curve into the POWERWORLD simulation software and
simulating the power system for a full 24 hours yield a total cost in dollars for operating the
power system over that period of time. Following the technique as shown before, the twenty
three bus system was simulated with various capacitor banks removed. The results are
summarized in Table 4.1. This simulation was run at 60 times real time for the entire simulation.
Table 4.1 Simulation data for sixbus system with transmission lines removed
System Configuration
Total Cost
for 24 Hour
Simulation
Difference from
No Change
Configuration
Percent Increase
from No Change
Configuration
No Changes $ 1,358,423 N/A N/A
Capacitor Bank at bus 103 Removed $ 1,359,609 $1,186.00 0.0873%
Capacitor Bank at bus 2 Removed $ 1,359,222 $799.00 0.0588%
Capacitor Bank at bus 104 Removed $ 1,359,626 $1,203.00 0.0886%
Again, from this information, the approximate cost a company would incur per day while
removing the capacitor bank can be inferred from these calculations.
4.8 Limit on Available Transfer Capability (ATC) Due to a Voltage Constraint [26]
Another use for the OPF algorithm is for the calculation of the available transfer capability
for a system. The OPF is able to take into account system security concerns including voltage
limits and transmission line limits while calculating the ATC.
The threebus system of Figure 4.1 (data in Appendix H) can be used to illustrate and
examine the value of reactive power support at a system load bus. Area one consists only of the
generator and load at bus 1 plus two tie lines. Area two consists of the generator and load at bus
3 plus the load at bus 2, and the line from bus 2 to bus 3. Area one has expensive generation,
48
while area two has cheaper generation. For the indicated loads and no area transfers the cost of
operation for the total system is $25,799/hr. For this study, the loads at all buses remain fixed at
the values shown.
The base case shown in Figure 4.1 has each area providing its own load plus a portion of the
system losses. The unfilled pie charts on each line indicate that the lines are loaded at less than
their MVA ratings. This base case has zero scheduled megawatt transfer between the two areas,
but 73 MW of area two’s generation flows through area one.
1.04 PU
0.9710 PU
504 MW
193 MVR
411 MW
70 MVR
300 MW
75 MW
Bus 1
Bus 2
100 MW
1.04 PU
Bus 3
100 MVR
Area Two
Area One
14 MVR 500 MW
100 MVR
30 MVR
50.18 $/MWH
0.00 $/MVRH
20.22 $/MWH
0.00 $/MVRH
22.85 $/MWH
0.94 $/MVRH
72 MW
21 MVR
72 MVR
73 MW
15 MVR
238 MW
55 MVR
73 MW
64 MVR
227 MW
22 MVR
Figure 4.1 ThreeBus Base Case with No Area Power Transfer
Note that the incremental cost for reactive power at a generator bus is $0.00. This is true as
long as no MVAR limits are reached because no cost is associated with generation of MVARs by
a generator in the OPF formulation of this thesis.
Consider the situation where area one wants to purchase power at considerable savings from
area two. Similarly, area two wants to sell power to area one. We first examine the available
49
transfer capability of this system for transfer from area two to area one. An OPF solution with
bus voltage constraints is used to find the maximum power which can be transferred from area
two to area one. The voltage constraint for power quality used in this example is
0.96 ≤ V
i
≤ 1.04.
Both generators have their excitations set to give 1.04 p.u. voltage for serving their local area
load within the voltage constraints. A global OPF solution which minimizes the total cost of
providing the total load will attempt to increase areatwo generation and decrease areaone
generation. This power transfer will change the voltage at bus 2. When the transfer is such that
the voltage reaches its lower limit (0.96), the OPF solution will stop at that constraint. The
solution and associated marginal costs for both real and reactive power at each bus are shown in
Figure 4.2 for the case in which a nominal 15 MVAR capacitor bank is installed at bus 2 for
reactive power support. The available transfer capability for these constraints is 244 MW. The
cost of operation for the total system is $21,341/hr. Thus this maximum power transfer has
resulted in a reduction of total cost by $4,458/hr. Additional transfers are not possible without
violation of the voltage constraint at bus 2.
Note that the incremental cost for MVARs is
almost the same as for MWs at bus 2.
When the capacitor bank at bus 2 is
increased from a nominal 15 MVAR rating to
a nominal 30 MVAR rating, the OPF solution
is given in Figure 4.3. The available transfer
capability for these constraints is increased to
1.04 PU
0.9600 PU
271 MW
290 MVR
668 MW
59 MVR
300 MW
28 MW
Bus 1
Bus 2
100 MW
1.04 PU
Bus 3
100 MVR
Area Two
Area One
14 MVR 500 MW
100 MVR
30 MVR
38.55 $/MWH
0.00 $/MVRH
25.36 $/MWH
0.00 $/MVRH
55.09 $/MWH
54.40 $/MVRH
201 MW
78 MVR
112 MVR
214 MW
26 MVR
354 MW
55 MVR
30 MW
102 MVR
330 MW
16 MVR
Figure 4.2 ThreeBus Example at Maximum
Power Transfer (15 MVAR Capacitor
Support)
50
325 MW. The cost of operation for the total
system is $20,890/hr. Thus the increase in
available power transfer capability has
resulted in an additional reduction of total
cost by $451/hr. Again, additional transfers
are not possible without violation of the
voltage constraint at bus 2. Note that the
incremental cost for MVARs has reduced as
the OPF approaches a more economical
dispatch.
When the capacitor bank at bus 2 is
increased from a nominal 30 MVAR rating to
a nominal 45 MVAR rating, the OPF solution
is given in Figure 4.4. The available transfer
capability for these constraints is increased to
355 MW. The cost of operation for the total
system is $20,849/hr. Thus the increase in available power transfer capability has resulted in a
further reduction of total cost by $41/hr. Note that the OPF has reached an optimal dispatch
schedule without hitting the bus 2 voltage constraint (0.963 > 0.96). Therefore, no additional
power transfers can lower the total system costs.
The savings between 15 and 30 MVAR capacitors was $451 while the savings between
30 and 45 MVAR capacitors was only $41, because there is a greater change in system megawatt
1.04 PU
0.9600 PU
198 MW
323 MVR
754 MW
56 MVR
300 MW
60 MW
Bus 1
Bus 2
100 MW
1.04 PU
Bus 3
100 MVR
Area Two
Area One
28 MVR 500 MW
100 MVR
30 MVR
34.89 $/MWH
0.00 $/MVRH
27.08 $/MWH
0.00 $/MVRH
38.00 $/MWH
13.37 $/MVRH
242 MW
101 MVR
122 MVR
261 MW
25 MVR
393 MW
51 MVR
64 MW
109 MVR
364 MW
36 MVR
Figure 4.3 ThreeBus Example at Maximum
Power Transfer (30 MVAR Capacitor
Support)
1.04 PU
0.9631 PU
170 MW
332 MVR
786 MW
51 MVR
300 MW
73 MW
Bus 1
Bus 2
100 MW
1.04 PU
Bus 3
100 MVR
Area Two
Area One
42 MVR 500 MW
100 MVR
30 MVR
33.52 $/MWH
0.00 $/MVRH
27.72 $/MWH
0.00 $/MVRH
32.82 $/MWH
1.05 $/MVRH
257 MW
110 MVR
122 MVR
278 MW
24 MVR
408 MW
44 MVR
77 MW
107 MVR
377 MW
49 MVR
Figure 4.4 ThreeBus Example at Maximum
Power Transfer (45 MVAR Capacitor
Support)
51
dispatch when moving from 15 to 30 MVAR than from 30 to 45 MVAR. This is also reflected
in the incremental costs of MVAR at bus 2 for the various cases.
4.9 Transmission System Pricing Through ShortRun Marginal Costing (SRMC)
In the restructured environment of the future, it may be necessary to determine the cost
incurred to the transmission system due to a power transaction between two companies. This
cost would be charged to the two companies undergoing the transaction and would be paid to the
operator of the transmission system. Two methods proposed for the determination of this cost is
through shortrun marginal costing (SRMC) and longrun marginal costing (LRMC). These two
topics are extensively discussed in both [27] and [28]. A brief, albeit simple, explanation of
these follows.
The SRMC takes into account only the operating costs of the power system. The operating
costs include fuel costs, losses, as well as system constraints such as transmission line limits.
The LRMC, however, includes both the operating costs and the capital costs of the power system.
The capital costs include the costs of future expansion to the transmission system as well as the
generation capacity.
Through the use of the OPF, the SRMC technique can easily be implemented as discussed in
[28]. The SRMC of a power transaction can be estimated by solving the OPF for both the system
with the transaction and the system without the transaction in place. After these solutions are
obtained, the SRMC can be found as follows.
Define BMC
i
= (marginal cost of power at bus i after the transaction)
P
i,transaction
= the net power injection at bus i due to the transaction
= ( power generation at bus i before transaction
+ power generation at bus i after transaction)
From these definitions a definition for SRMC can then be made [28].
52
SRMC P
transaction i transaction
·
∑
BMC
i
all buses
in transaction
*
,
As a simple example consider the sixbus, twoarea system from Section 4.3. Figure 4.1 on page
36 shows this system with no transactions, and Figure 4.2 on page 37 shows the system with a
transaction of 65.5 MW undertaken between the two areas. Taking the data from these two
figures, a calculation of the SRMC for the transaction can be made. This calculation is
summarized in Table 4.1.
Table 4.1 SRMC calculation for sixbus, twoarea system
Bus
Nu
m
Generation
Before
Transaction
[MW]
BMC
Before
Transaction
[$/MWhr]
Generation
After
Transaction
[MW]
BMC After
Transaction
[$/MWhr]
P
i,trans
[MW]
BMC
i
*P
i,trans
[$/hr]
1 161 15.86 122 14.93 39 582.27
2 133 12.15 190 13.25 57 755.25
3 143 16.71 116 16.03 27 432.81
4 169 12.58 179 12.77 10 127.70
SRMC P
transaction i transaction
· ·
∑
BMC
i
all buses
in transaction
*
,
132.13
Thus, the SRMC for this transaction is ( $132.13)/hour. As mentioned in [27], SRMC can
be a negative value.
For comparison, consider the same system undergoing the same 65.5 MW transaction, but
with the line between buses 4 and 5 doubled to 100 MVA from 50 MVA. The simulation of this
system is shown in Figure 4.1.
53
125 MW
77 MVR
100 MW
100 MW
20 MVR
165 MW
13 MVR
1.00 PU
1.00 PU
1.00 PU
Bus 3
Bus 2
Bus 4
100 MW
20 MVR
20 MVR
20 MVR
100 MW
113 MW
204 MW
35 MVR
41 MVR
1.00 PU
Bus 5
Bus 6
0.97 PU
0.97 PU
100 MW
50 MVR
Bus 1
100 MW
10 MVR
15.76 $/MWH
15.00 $/MWH
12.77 $/MWH
13.23 $/MWH
15.95 $/MWH
14.30 $/MWH
0.00 $/MVRH
0.23 $/MVRH
0.00 $/MVRH
0.10 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
Area Two
Area One
Figure 4.1 Sixbus, TwoArea System Undergoing Transaction with Line Limit Doubled
Again, Figure 4.1 shows this system with no transactions, and Figure 4.2 shows the system
with a transaction of 65.5 MW undertaken between the two areas. Taking the data from these
two figures, a calculation of the SRMC for the transaction can be made for the system where no
line limit will be encountered. This calculation is summarized in Table 4.2.
Table 4.2 SRMC calculation for sixbus, twoarea system where no limit is encountered
Bus
Num
Generation
Before
Transaction
[MW]
BMC
Before
Transaction
[$/MWhr]
Generation
Aafter
Transaction
[MW]
BMC After
Transaction
[$/MWhr]
P
i,trans
[MW]
BMC
i
*P
i,trans
[$/hr]
1 161 15.86 125 15.00 36 540.00
2 133 12.15 165 12.77 32 408.64
3 143 16.71 113 15.95 30 478.50
4 169 12.58 204 13.23 35 463.05
SRMC P
transaction i transaction
· ·
∑
BMC
i
all buses
in transaction
*
,
146.81
54
Thus, the SRMC for this transaction is ( $146.81)/hour. This is not unexpected. The
previous system encountered a transmission line constraint, while this one has not; therefore, the
incentive to do the transaction should be large, i.e., the SRMC should be less.
While this discussion is not entirely complete, it does display the basic use of the OPF to
calculate the SRMC. More study will be done on this topic in the future.
55
5. CONCLUSION
The OFP written for this thesis has been very successful in achieving the goals set forth for
an OPF. Minimization of system costs, while maintaining system security, was accomplished
through the implementation of Newton’s method to the OPF problem. Newton’s method has
proven to be very adept at solving the OPF problem.
Many applications of the OPF have also been shown as in Chapter 4. The OPF performs
generator control and transmission system control while taking into account system limits. The
marginal cost data from the OPF were shown to aid in the available transfer capability (ATC)
calculation, real and reactive power pricing, transmission system pricing, and transmission
system component valuation.
At present, the OPF works very well for small systems (less than 200 buses). Improvements
will need to be made to this in order for the software to be truly useful. This can and will be
done with relatively minor modifications. As discussed in Section 3.5, by keeping the previous
OPF solution from time step to time step, the speed of the OPF solution will be increased. This
will allow larger systems to be simulated. The author also plans to implement the partial re
factorization schemes for the Hessian as seen in [25]. These improvements to the software will
allow the simulation of systems with thousands of buses. With this ability, the use of the OPF in
a power system simulation environment will be truly useful.
Another problem which should be addressed is the modeling of discrete variables within the
OPF framework. At present, variables such as transformer tap ratios and phase shift angles are
treated as continuous variables.
56
As a final note for this thesis, the author would like to reiterate that while the applications of
the OPF as discussed in this thesis are extremely valuable, they do not take into account the
possibility of random outages of transmission line components or generators. Also, the pricing
schemes discussed do not take into account the need for future expansion of the power system.
57
APPENDIX A. ECONOMIC INTERPRETATION OF THE LAGRANGE
MULTIPLIERS
The Lagrange multipliers found in the optimal solution can be given an economic
interpretation. One can show that the Lagrange multipliers are the negative of the derivative of
the function which is being minimized with respect to the enforced constraint. In other words,
they are the marginal costs associated with enforcing the constraint. A proof of this as taken
from [29] follows.
The nonlinear programming problem that is being solved can be expressed in the following
form (excluding inequality constraints).
Minimize x (the objective function)
subject to: x u (equality constraints)
i
f ( )
h ( ) = , i = , , . . . , m
i
1 2
Therefore, the Lagrange function may be written:
( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )
[ ]
( )
L f h
T
x u u u x u u x u u
with u is a vector of Lagrange multipliers
, , λ λ
λ
· + −
The optimal solution for u = 0 is (x
*
, λ
*
). In this problem formulation, the values of u
i
are
constraints on some resource. One can see that if this constraint value were varied, the optimal
solution (x
*
, λ
*
) would change. The sensitivity to a change in u is desired. This sensitivity can
be expressed as the derivative of the Lagrange function with respect to the variable u. Evaluating
this derivative yields
( )
( ) ( )
( ) ( )
( )
( )
( ) ( )
[ ]
( ) ∇ ·
∂
∂
∇ +
∂
∂
∂
∂
+
∂
∂
− −
u x
x u
x
u
x u
x
u
x u
x
u
u
u
x u u u L f
h
h , , λ λ
λ
λ
Grouping terms of this functions gives
58
( ) ( ) ( )
( ) ( )
( )
( )
( ) ( )
[ ]
( ) ∇ ·
∂
∂
∇ +
∂
∂
¸
1
]
1
+
∂
∂
− −
u x
x u
x
u
x u
x u
x
u
First Term
u
u
x u u
Second Term
u L f
h
h , , λ λ
λ
λ
1 2 444444 3 444444
1 2 444 3 444
Now, study the first two terms of this more closely.
( ) ( )
( ) ( )
( )
( ) ∇ +
∂
∂
¸
1
]
1
· ∇
x x
x u
x u
x
u x u f
h
L λ λ , ,
( ) ( )
[ ] ( ) h L x u u x u − · ∇
λ
λ , ,
Thus the following is written
( ) ( )
( )
( )
( ) ∇ ·
∂
∂
∇ +
∂
∂
∇ −
u x
x u
x
u
x u
First Term
u
u
x u
Second Term
u L L L , , , , , , λ λ
λ
λ λ
λ
1 2 44 3 44 1 2 444 3 444
By the KuhnTucker optimality conditions [20, p. 284], both ( ) ∇
x
x u L , , λ and ( ) ∇
λ
λ L x u , ,
must be zero at the optimal solution. Therefore, both the first and second terms must be zero at
the optimal solution, and the following is left at the optimal solution
( ) ∇ · −
u
* * *
x L ,λ λ
We also know that at the optimal solution,
( ) ( ) ∇ · ∇
u
* *
u
*
x x L f ,λ
Combining these two equations, it is seen that at the optimal solution, the following is true.
( ) λ
*
u
*
x · −∇ f
Thus the Lagrange multipliers are the negatives of the derivative of the function which is being
minimized with respect to the enforced constraint. If the objective function is considered a cost
function, then the λ
*
can be interpreted as the cost per unit of the resources associated with each
constraint. These Lagrange multipliers are thus often referred to as the shadow prices.
59
Although the previous derivation has not included inequality constraints, the derivation with
them included is essentially the same. A simple explanation will be proved here. For all
inequality constraints not being enforced (i.e., they are not active), the derivative of the objective
function with respect to them will be zero, because the problem solution will not be affected by
changing the constraint (u
i
) since it is not being enforced anyway. As for the inequality
constraints that are being enforced, they will simply behave like equality constraints and the
previous derivation can be applied to them.
60
APPENDIX B. CALCULATION OF THE GRADIENT OF THE
LAGRANGIAN
Appendix B contains a list of all the terms that make up the gradient of the Lagrangian for the
OPF. The listing is organized by those equations or constraints which result in the addition of
the gradient term.
Note: The terms that are shown surrounded by parenthesis, (), are not always part of the
gradient. They are dependent on the inequality constraints being enforced.
Note: Many terms are surrounded by brackets, []. Appendix D contains a further
explanation of the calculation of these terms.
Gradient term due to generator cost curves (the objective function)
∂
∂
· +
L
P
b c P
Gi
i i Gi
2
Gradient terms due to the power flow equations
∂
∂
·
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
¹
'
¹
¹
;
¹
∑
L
V
P
V
Q
V
i
Pk
k
i
Qk
k
i k
µ µ
∂
∂
·
∂
∂
¸
1
]
1
1
+
∂
∂
¸
1
]
1
1
¹
'
¹
¹
¹
¹
;
¹
¹
¹
∑
L P Q
i
Pk
k
i
Qk
k
i k
δ
µ
δ
µ
δ
∂
∂
·
L
P
Pk
k
µ
∂
∂
·
L
Q
Qk
k
µ
∂
∂
· −
L
P
Gi
Pk
µ
Gradient terms due to generator voltage set points
∂
∂
·
L
V
i
viset
µ
∂
∂
· −
L
V V
viset
i Gi set
µ
61
Gradient terms due to maximum and minimum generator power output
( )
∂
∂
· − +
L
P P
PGil
Gi Gi
λ
min
( )
∂
∂
· + −
L
P P
PGih
Gi Gi
λ
max
( ) ( )
∂
∂
· − +
L
P
Gi
PGil PGih
λ λ
Gradient terms due to enforcement of hard transmission line constraints
∂
∂
· −
¸
¸
_
,
L
S S
Skm
km km
λ
2
2
max
∂
∂
·
∂
∂
¸
1
]
1
1
¸
¸
_
,
L
x
S
x
x V V
Skm
km
k m k m
λ
δ δ
2
where = or , , ,
Gradient terms due to enforcement of soft transmission line constraints
∂
∂
· −
¸
¸
_
,
∂
∂
¸
1
]
1
1
¸
¸
_
,
L
x
k S S
S
x
x V V
km km
km
k m k m
2
2
2
2
max
where = , , , or δ δ
Gradient terms due to enforcement of area interchange constraints
{ }
∂
∂
· −
∑
L
P P
km sched
µ
int tie lines
∂
∂
·
∂
∂
¸
1
]
1
∑
L
x
P
x
x V V
km
k m k m
µ
δ δ
int
, , ,
tie lines
where = or
Gradient terms due to enforcement of hard voltage constraints
( ) ( )
∂
∂
· + −
L
V
i
vih vil
λ λ
( )
∂
∂
· − +
L
V V
vil
i i
λ
min
( )
∂
∂
· + −
L
V V
vih
i i
λ
max
Gradient terms due to enforcement of soft voltage constraints
( ) ( )
∂
∂
· − − +
L
V
k V V
i
i i
2
min
( ) ( )
∂
∂
· + −
L
V
k V V
i
i i
2
max
Gradient terms due to tap changing transformers
62
( )
∂
∂
· −
L
t t
tkm
km km
λ
max
max
( )
∂
∂
· − +
L
t t
tkm
km km
λ
min
min
( ) ( )
∂
∂
·
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
1
¸
¸
_
,
+
∂
∂
¸
1
]
1
1
¸
¸
_
,
+ + −
L
t
P
t
Q
t
P
t
Q
t
S
t
k
S
t
km
Pm
m
km
Qm
m
km
Pk
k
km
Qk
k
km
Skm
km
km
km
km
tkm tkm
µ µ µ µ
λ λ λ
2 2
max min
Gradient terms due to phase shifting transformers
( )
∂
∂
· −
L
km
km km
λ
α α
α max
max
( )
∂
∂
· − +
L
km
km km
λ
α α
α min
min
( ) ( )
∂
∂
·
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
1
¸
¸
_
,
+
∂
∂
¸
1
]
1
1
¸
¸
_
,
+ + −
L
P Q P Q
S
k
S
km
Pm
m
km
Qm
m
km
Pk
k
km
Qk
k
km
Skm
km
km
km
km
km km
α
µ
α
µ
α
µ
α
µ
α
λ
α α
λ λ
α α
2 2
max min
63
APPENDIX C. CALCULATION OF THE HESSIAN OF THE
LAGRANGIAN
Appendix C contains a list of all the terms that make up the Hessian of the Lagrangian for the
OPF. The listing is organized by those equations or constraints which result in the addition of
the Hessian term.
Note: The terms that are shown surrounded by parenthesis, (), are not always part of the
Hessian. They are dependent on those inequality constraints being enforced.
Note: Many terms are surrounded by brackets, []. Appendix E contains a further
explanation of the calculation of these terms.
Hessian term due to generator cost curves (the objective function)
∂
∂
·
2
2
2
L
P
c
Gi
i
Hessian terms due to the power flow equations
∂
∂ ∂
·
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
¹
'
¹
¹
;
¹
∑
2 2 2
L
x y
P
x y
Q
x y
x y V V
Pk
k
Qk
k
k
i j i j
µ µ δ δ where and = , , , or
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L
V
P
V
i Pk
k
i
µ
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L
V
Q
V
i Qk
k
i
µ
∂
∂ ∂
·
∂
∂
¸
1
]
1
1
2
L P
i Pk
k
i
δ µ δ
∂
∂ ∂
·
∂
∂
¸
1
]
1
1
2
L Q
i Qk
k
i
δ µ δ
∂
∂ ∂
· −
2
1
L
P
Gk Pk
µ
64
Hessian terms due to generator voltage set points
∂
∂ ∂
· +
2
1
L
V
i viset
µ
Hessian terms due to maximum and minimum generator power output
( )
∂
∂ ∂
· −
2
1
L
P
Gi PGil
λ
( )
∂
∂ ∂
· +
2
1
L
P
Gi PGih
λ
Hessian terms due to hard transmission line constraints
∂
∂ ∂
·
∂
∂
¸
1
]
1
1
¸
¸
_
,
2
2
L
x
S
x
x V V
Skm
km
k m k m
λ
δ δ where = or , , ,
∂
∂ ∂
·
∂
∂ ∂
¸
1
]
1
1
¸
¸
_
,
2
2
2
L
x y
S
x y
x y V V
Skm
km
k m k m
λ δ δ where and = or , , ,
Hessian terms due to soft transmission line constraints
∂
∂ ∂
· −
¸
¸
_
,
∂
∂ ∂
¸
1
]
1
1
+
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
1
¸
¸
_
,
2
2
2
2
2 2 2
2 2
L
x y
k S S
S
x y
k
S
x
S
y
x V V
km km
km km km
k m k m
max
where = , , , or δ δ
Hessian terms due to area interchange constraints
∂
∂ ∂
·
∂
∂
¸
1
]
1
∑
2
L
x
P
x
km
µ
int tie lines
∂
∂ ∂
·
∂
∂ ∂
¸
1
]
1 ∑
2 2
L
x y
P
x y
x y V V
km
k m k m
µ
δ δ
int
, , ,
tie lines
where and = or
Hessian terms due to hard voltage constraints
65
( )
∂
∂ ∂
· +
2
1
L
V
i vih
λ
( )
∂
∂ ∂
· −
2
1
L
V
i vil
λ
( )
∂
∂
·
2
2
0
L
V
i
Hessian terms due to soft voltage constraints
( )
∂
∂
·
2
2
2
L
V
k
i
Hessian terms due to tap changing transformers
∂
∂ ∂
·
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
1
¸
¸
_
,
+
∂
∂ ∂
¸
1
]
1
1
¸
¸
_
,
2 2 2 2 2
2
2
2
2
L
t x
P
t x
Q
t x
P
t x
Q
t x
S
t x
k
S
t x
x V V
km
Pm
m
km
Qm
m
km
Pk
k
km
Qk
k
km
Skm
km
km
km
km
k m k
µ µ µ µ
λ δ where = , , , or δ
m
( )
∂
∂ ∂
· +
2
1
L
t
tkm km
λ
max
( )
∂
∂ ∂
· −
2
1
L
t
tkm km
λ
min
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L
t
P
t
km Pm
m
km
µ
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L
t
Q
t
km Qm
m
km
µ
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L
t
P
t
km Pk
k
km
µ
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L
t
Q
t
km Qk
k
km
µ
∂
∂ ∂
·
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
1
¸
¸
_
,
2
2 2
L
t
S
t
S
t
km Skm
km
km
mk
km
λ
or ; depending on which end of the line is metered
∂
∂ ∂
·
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
2
L
t
P
t
P
t
P
t
km km
km
km
km
km
µ
int
int
or ; depending on which end of the line is metered
Hessian terms due to phase shifting transformers
66
∂
∂ ∂
·
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
+
∂
∂ ∂
¸
1
]
1
1
¸
¸
_
,
+
∂
∂ ∂
¸
1
]
1
1
¸
¸
_
,
2 2 2 2 2
2
2
2
2
L
x
P
x
Q
x
P
x
Q
x
S
x
k
S
x
x V V
km
Pm
m
km
Qm
m
km
Pk
k
km
Qk
k
km
Skm
km
km
km
km
k m k
α
µ
α
µ
α
µ
α
µ
α
λ
α α
δ where = , , , or δ
m
( )
∂
∂ ∂
· +
2
1
L
km km
λ α
α max
( )
∂
∂ ∂
· −
2
1
L
km km
λ α
α min
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L P
km Pm
m
km
α µ α
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L Q
km Qm
m
km
α µ α
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L P
km Pk
k
km
α µ α
∂
∂ ∂
·
∂
∂
¸
1
]
1
2
L Q
km Qk
k
km
α µ α
∂
∂ ∂
·
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
1
¸
¸
_
,
2
2 2
L
S S
km Skm
km
km
mk
km
α λ α α
or ;
depending on which end
of the line is metered
∂
∂ ∂
·
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
2
L P P P
km km
km
km
km
km
α µ α α α
int
int
or ;
depending on which end
of the line is metered
67
APPENDIX D. SUMMARY OF DERIVATIVE CALCULATIONS
In Appendices B and C, many partial derivatives left uncalculated. These partial derivatives
are summarized in Appendix D. The listing is organized by the equations being differentiated.
Calculation of partial derivatives of the net power and reactive power injections.
( ) ( )
[ ] [ ]
( ) ( )
[ ] [ ]
P V V g b P P
Q V V g b Q Q
g b
k k m km k m km k m
m
N
Gk Lk
k k m km k m km k m
m
N
Gk Lk
km km
· − + − − +
· − − − − +
·
·
∑
∑
cos sin
sin cos
δ δ δ δ
δ δ δ δ
1
1
where and are elements of the real and
imaginary parts of the network admittance matrix
(D.1)
From Equation (D.1), the following partial derivatives can be found
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
P Q P
V
Q
V
k
i
k
i
k
i
k
i
δ δ
, , ,and
are elements of the familiar power
system Jacobian matrix
∂
∂
¸
1
]
1
·
2
2
2
P
V
g
k
k
kk
∂
∂
¸
1
]
1
·
2
2
0
P
V
k
m
( ) ( )
∂
∂ ∂
¸
1
]
1
· − + −
2
P
V V
g b
k
k m
km k m km k m
cos sin δ δ δ δ
( ) ( )
[ ] [ ]
∂
∂
¸
1
]
1
· − − + −
·
∑
2
2
1
P
V V g b
k
k
k n kn k n kn k n
n
δ
δ δ δ δ cos sin
all
busses
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − + −
2
P
V V g b
k
k m
k m km k m km k m
δ δ
δ δ δ δ cos sin
68
( ) ( )
[ ]
∂
∂
¸
1
]
1
· − − − −
2
2
P
V V g b
k
m
k m km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ] { }
∂
∂ ∂
¸
1
]
1
1
· − − + −
∑
2
P
V
V g b
k
k k
n kn k n kn k n
δ
δ δ δ δ sin cos
n
all busses
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − −
2
P
V
V g b
k
k m
m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − + −
2
P
V
V g b
k
m k
k km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − −
2
P
V
V g b
k
m m
k km k m km k m
δ
δ δ δ δ sin cos
∂
∂
¸
1
]
1
· −
2
2
2
Q
V
b
k
k
kk
∂
∂
¸
1
]
1
·
2
2
0
Q
V
k
m
( ) ( )
∂
∂ ∂
¸
1
]
1
· − − −
2
Q
V V
g b
k
k m
km k m km k m
sin cos δ δ δ δ
( ) ( )
[ ] [ ]
∂
∂
¸
1
]
1
· − − − −
·
∑
2
2
1
Q
V V g b
k
k
k n kn k n kn k n
n
δ
δ δ δ δ sin cos
all busses
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − −
2
Q
V V g b
k
k m
k m km k m km k m
δ δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂
¸
1
]
1
· − − + −
2
2
Q
V V g b
k
m
k m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ] { }
∂
∂ ∂
¸
1
]
1
1
· − + −
∑
2
Q
V
V g b
k
k k
n kn k n kn k n
δ
δ δ δ δ cos sin
n
all busses
69
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − + −
2
Q
V
V g b
k
k m
m km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − + −
2
Q
V
V g b
k
m k
k km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − + −
2
Q
V
V g b
k
m m
k km k m km k m
δ
δ δ δ δ cos sin
Calculation of the partial derivatives of the power and reactive power flow between buses.
( ) ( )
[ ]
( ) ( )
[ ]
P V g V V g b
Q V b V V g b
g , b , g , b
km k kk k m km k m km k m
km k kk k m km k m km k m
kk kk km km
· + + − + −
· − + − − −
2
2
cos sin
sin cos
δ δ δ δ
δ δ δ δ
where and are the contributions to the network
admittance matrix from the branch between bus k and bus m.
(D.2)
From Equation (D.2), the following partial derivatives can be found
( ) ( )
[ ]
∂
∂
¸
1
]
1
· + − + −
P
V
V g V g b
km
k
k kk m km k m km k m
2 cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂
¸
1
]
1
· − + −
P
V
V g b
km
m
k km k m km k m
cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − − + −
P
V V g b
km
k
k m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − − −
P
V V g b
km
m
k m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂
¸
1
]
1
· − + − − −
Q
V
V b V g b
km
k
k kk m km k m km k m
2 sin cos δ δ δ δ
70
( ) ( )
[ ]
∂
∂
¸
1
]
1
· − − −
Q
V
V g b
km
m
k km k m km k m
sin cos δ δ δ δ
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − + −
Q
V V g b
km
k
k m km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − − − −
Q
V V g b
km
m
k m km k m km k m
δ
δ δ δ δ cos sin
∂
∂
¸
1
]
1
·
2
2
2
P
V
g
km
k
kk
( ) ( )
∂
∂ ∂
¸
1
]
1
· − + −
2
P
V V
g b
km
k m
km k m km k m
cos sin δ δ δ δ
∂
∂
¸
1
]
1
·
2
2
0
P
V
km
m
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − − + −
2
2
P
V V g b
km
k
k m km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − + −
2
P
V V g b
km
k m
k m km k m km k m
δ δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − − + −
2
2
P
V V g b
km
m
k m km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − + −
2
P
V
V g b
km
k k
m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − −
2
P
V
V g b
km
k m
m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − + −
2
P
V
V g b
km
m k
k km k m km k m
δ
δ δ δ δ sin cos
71
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − −
2
P
V
V g b
km
m m
k km k m km k m
δ
δ δ δ δ sin cos
∂
∂
¸
1
]
1
· −
2
2
2
Q
V
b
km
k
kk
( ) ( )
∂
∂ ∂
¸
1
]
1
· − − −
2
Q
V V
g b
km
k m
km k m km k m
sin cos δ δ δ δ
∂
∂
¸
1
]
1
·
2
2
0
Q
V
km
m
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − − − −
2
2
Q
V V g b
km
k
k m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − −
2
Q
V V g b
km
k m
k m km k m km k m
δ δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂
¸
1
]
1
1
· − − − −
2
2
Q
V V g b
km
m
k m km k m km k m
δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − + −
2
Q
V
V g b
km
k k
m km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − − −
2
Q
V
V g b
km
k m
m km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· + − + −
2
Q
V
V g b
km
m k
k km k m km k m
δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
1
· − − − −
2
Q
V
V g b
km
m m
k km k m km k m
δ
δ δ δ δ cos sin
Calculation of the partial derivatives of the square of the MVA flow on a line, S
km

2
72
S P Q
km km km
2
2 2
· + (D.3)
From Equation (D.3), the following partial derivative can be found
∂
∂
¸
1
]
1
1
·
∂
∂
¸
1
]
1
+
∂
∂
¸
1
]
1
S
x
P
P
x
Q
Q
x
x V V
km
km
km
km
km
k m k m
2
2 2 where = or , , , δ δ
Equation (D.3) can also be simplified to the following
( ) S V y V V y V V y y
g jb y e g jb y e
km k kk k m km k m km kk k m kk km
kk kk kk
j
km km km
j
kk km
2
4 2 2 2 2 3
2 · + + − + −
+ · + ·
cos δ δ δ δ
δ δ
where and
(D.4)
From Equation (D.4), the following partial derivatives can be found
( )
∂
∂
¸
1
]
1
1
· − − + −
2
2
2
3
2
S
V V y y
km
k
k m km kk k m kk km
δ
δ δ δ δ cos
( )
∂
∂ ∂
¸
1
]
1
1
· − + −
2
2
3
2
S
V V y y
km
k m
k m km kk k m kk km
δ δ
δ δ δ δ cos
( )
∂
∂
¸
1
]
1
1
· − − + −
2
2
2
3
2
S
V V y y
km
m
k m km kk k m kk km
δ
δ δ δ δ cos
( )
∂
∂
¸
1
]
1
1
· + + − + −
2
2
2
2 2 2 2
12 2 12
S
V
V y V y V V y y
km
k
k kk m km k m km kk k m kk km
cos δ δ δ δ
( )
∂
∂ ∂
¸
1
]
1
1
· + − + −
2
2
2 2
4 6
S
V V
V V y V y y
km
k m
k m km k km kk k m kk km
cos δ δ δ δ
∂
∂
¸
1
]
1
1
·
2
2
2
2 2
2
S
V
V y
km
m
k km
( )
∂
∂ ∂
¸
1
]
1
1
· − − + −
2
2
2
6
S
V
V V y y
km
k k
k m km kk k m kk km
δ
δ δ δ δ sin
73
( )
∂
∂ ∂
¸
1
]
1
1
· − + −
2
2
2
6
S
V
V V y y
km
k m
k m km kk k m kk km
δ
δ δ δ δ sin
( )
∂
∂ ∂
¸
1
]
1
1
· − − + −
2
2
3
2
S
V
V y y
km
m k
k km kk k m kk km
δ
δ δ δ δ sin
( )
∂
∂ ∂
¸
1
]
1
1
· − + −
2
2
3
2
S
V
V y y
km
m m
k km kk k m kk km
δ
δ δ δ δ sin
Calculation of partial derivatives with respect to tap ratio and phase shift variables
Before beginning calculation of the partial derivative with respect to tap ratio and phase shift
variables, first consider what equations they affect. In the modeling of power systems, tap ratios
and phase shifts are typically represented as alterations to the network admittance matrix. Figure
D.1 shows the model for a transformer with tap changing and/or phase shifting capabilities.
Using this model, the entries in the network admittance matrix due to the transformer can be
found as in the book by Grainger [20, pp. 361362].
+
V
k
_
+
V
m
_
I
m
I
k
r + jx
Y
cap
2
Y
cap
2
1 : t
km
e
+jαkm
Figure D.1 Transformer Model
74
I
I
y
Y
e
t
y
e
t
y
t
y
Y
V
V
k
m
cap
j
km
j
km km
cap
k
m
km
km
Y
¸
1
]
1
1
·
+ −
− +
¸
1
]
1
1
1
1
¸
1
]
1
1
−
+
col k col m
row k
row m
2
1
2
2
α
α
1 2 4444 3 4444
where y
r jx
g jb ·
+
· +
1
From this model, the following derivatives are defined.
[ ] ∂
∂
·
∂
∂
∂
∂
∂
∂
∂
∂
¸
1
]
1
1
1
1
·
−
¸
1
]
1
1
1
1
·
−
− −
¸
1
]
1
1
1
1
−
+
Y
t
Y
t
Y
t
Y
t
Y
t
e
t
y
e
t
y
t
y
t
Y
t
Y
t
Y
km
kk
km
km
km
mk
km
mm
km
j
km
j
km km
km
km
km
mk
km
mm
km
km
0
2
0
1
1 2
2
2 3
α
α
[ ] ∂
∂
·
∂
∂
∂
∂
∂
∂
∂
∂
¸
1
]
1
1
1
1
·
−
−
¸
1
]
1
1
1
1
·
¸
1
]
1
1
1
1
−
+
2
2
2
2
2
2
2
2
2
2
3
3 4
2
2 2
0
2
2 6
0
2
2 6
Y
t
Y
t
Y
t
Y
t
Y
t
e
t
y
e
t
y
t
y
t
Y
t
Y
t
Y
km
kk
km
km
km
mk
km
mm
km
j
km
j
km km
km
km
km
mk
km
mm
km
km
α
α
[ ] ∂
∂
·
∂
∂
∂
∂
∂
∂
∂
∂
¸
1
]
1
1
1
1
·
−
¸
1
]
1
1
1
1
·
−
¸
1
]
1
1
·
−
+
−
¸
1
]
1
1
1
1
−
+
Y
Y Y
Y Y
j
e
t
y
j
e
t
y
jY
jY
jg
b
jg
b
km
kk
km
km
km
mk
km
mm
km
j
km
j
km
km
mk
km
km
mk
mk
km
km
α
α α
α α
α
α
0
0
0
0
0
0
[ ] ∂
∂
·
∂
∂
∂
∂
∂
∂
∂
∂
¸
1
]
1
1
1
1
·
¸
1
]
1
1
1
1
·
−
−
¸
1
]
1
1
·
−
−
−
−
¸
1
]
1
1
1
1
−
+
2
2
2
2
2
2
2
2
2
2
0
0
0
0
0
0
Y
Y Y
Y Y
e
t
y
e
t
y
Y
Y
g
jb
g
jb
km
kk
km
km
km
mk
km
mm
km
j
km
j
km
km
mk
km
km
mk
mk
km
km
α
α α
α α
α
α
From these equations, the partial derivatives of elements of the network admittance matrix with
respect to tap ratio and phase shift angle can be summarized as below.
75
∂
∂
·
∂
∂
· −
∂
∂
·
∂
∂
· −
∂
∂
· −
∂
∂
· −
∂
∂
· −
∂
∂
· −
∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
·
g
t
g
t
g
t
b
t
b
t
b
t
g
t
g
t
g
t
g
t
b
t
b
t
b
t
b
t
g
t
g
t
g
t
b
t
b
t
b
t
g
t
kk
km
km
km
km
km
kk
km
km
km
km
km
mk
km
mk
km
mm
km
mm
km
mk
km
mk
km
mm
km
mm
km
kk
km
km
km
km
km
kk
km
km
km
km
km
mk
km
0 0
2 2
0
2
0
2
2
2
2
2
2 2
2
2
2
2 2
2
2
g
t
g
t
g
t
b
t
b
t
b
t
b
t
mk
km
mm
km
mm
km
mk
km
mk
km
mm
km
mm
km
2
2
2 2
2
2 2
2
2 2
6 2 6 ∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
· −
∂
∂
· −
∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
·
∂
∂
· −
∂
∂
·
∂
∂
· −
∂
∂
· −
∂
∂
·
∂
∂
· −
∂
g g
b
b b
g
g
b
g b
g
b
g g
g
b b
b
g
g
g b
b
kk
km
km
km
km
kk
km
km
km
km
mk
km
mk
mm
km
mk
km
mk
mm
km
kk
km
km
km
km
kk
km
km
km
km
mk
km
mk
mm
km
mk
km
mk
α α α α
α α α α
α α α α
α α α
0 0
0 0
0 0
0
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
b
mm
km
∂
·
α
2
0
The equations that have to be differentiated are of the form
( ) ( ) ( ) ( ) ( ) w t u b t v g t
km km km km km
• · • + • , .
Therefore, the derivatives that need to be calculated may be found as follows.
∂
∂
·
∂
∂
∂
∂
w
t
u
b
t
v
g
t
km
km
km
km
km
+
∂
∂
·
∂
∂
∂
∂
2
2
2
2
2
2
w
t
u
b
t
v
g
t
km
km
km
km
km
+
∂
∂ ∂
·
∂
∂
∂
∂
∂
∂
¸
¸
_
,
·
2
w
t x x
u
b
t
v
g
t
x V V
km
km
km
km
km
k m k k
+ where or , , , δ δ
Using this process, the following partial derivatives with respect to tap ratio can be found.
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − + −
P
t
P
t
V V
t
g b
k
km
km
km
k m
km
km k m km k m
cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − − −
Q
t
Q
t
V V
t
g b
k
km
km
km
k m
km
km k m km k m
sin cos δ δ δ δ
76
( ) ( )
[ ] { }
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − + − + −
P
t
P
t
V
t
V g V g b
m
km
mk
km
m
km
m mm k mk m k mk m k
2 cos sin δ δ δ δ
( ) ( )
[ ] { }
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − + − − −
Q
t
Q
t
V
t
V b V g b
m
km
mk
km
m
km
m mm k mk m k mk m k
2 sin cos δ δ δ δ
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − + −
2
2
2
2 2
2 P
t
P
t
V V
t
g b
k
km
km
km
k m
km
km k m km k m
cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − −
2
2
2
2 2
2 Q
t
Q
t
V V
t
g b
k
km
km
km
k m
km
km k m km k m
sin cos δ δ δ δ
( ) ( )
[ ] { }
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· + − + −
2
2
2
2 2
2
3
P
t
P
t
V
t
V g V g b
m
km
mk
km
m
km
m mm k mk m k mk m k
cos sin δ δ δ δ
( ) ( )
[ ] { }
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − + − − −
2
2
2
2 2
2
3
Q
t
Q
t
V
t
V b V g b
m
km
mk
km
m
km
m mm k mk m k mk m k
sin cos δ δ δ δ
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
P
t V
P
t V
V
t
g b
k
km k
km
km k
m
km
km k m km k m
cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
P
t V
P
t V
V
t
g b
k
km m
km
km m
k
km
km k m km k m
cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − + −
2 2
P
t
P
t
V V
t
g b
k
km k
km
km k
k m
km
km k m km k m
δ δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
P
t
P
t
V V
t
g b
k
km m
km
km m
k m
km
km k m km k m
δ δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
Q
t V
Q
t V
V
t
g b
k
km k
km
km k
m
km
km k m km k m
sin cos δ δ δ δ
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
Q
t V
Q
t V
V
t
g b
k
km m
km
km m
k
km
km k m km k m
sin cos δ δ δ δ
77
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
Q
t
Q
t
V V
t
g b
k
km k
km
km k
k m
km
km k m km k m
δ δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − − −
2 2
Q
t
Q
t
V V
t
g b
k
km m
km
km m
k m
km
km k m km k m
δ δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
P
t V
P
t V
V
t
g b
m
km k
mk
km k
m
km
mk m k mk m k
cos sin δ δ δ δ
( ) ( )
[ ] { }
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − + − + −
2 2
1
4
P
t V
P
t V t
V g V g b
m
km m
mk
km m km
m mm k mk m k mk m k
cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
P
t
P
t
V V
t
g b
m
km k
mk
km k
k m
km
mk m k mk m k
δ δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − + −
2 2
P
t
P
t
V V
t
g b
m
km m
mk
km m
k m
km
mk m k mk m k
δ δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − + −
2 2
Q
t V
Q
t V
V
t
b g
m
km k
mk
km k
m
km
mk m k mk m k
cos sin δ δ δ δ
( ) ( )
[ ] { }
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + − − + −
2 2
1
4
Q
t V
Q
t V t
V b V b g
m
km m
mk
km m km
m mm k mk m k mk m k
cos sin δ δ δ δ
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − − −
2 2
Q
t
Q
t
V V
t
g b
m
km k
mk
km k
k m
km
mk m k mk m k
δ δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
Q
t
Q
t
V V
t
g b
m
km m
mk
km m
k m
km
mk m k mk m k
δ δ
δ δ δ δ cos sin
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
S
t
P
P
t t
x V V
km
km
km
km
km
km
km
km
k m k m
2
= 2 +2Q
Q
where = , , , or δ δ
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
S
t
P
P
t t
x V V
mk
km
mk
mk
km
mk
mk
km
k m k m
2
= 2 +2Q
Q
where = , , , or δ δ
78
∂
∂ ∂
¸
1
]
1
1
∂
∂ ∂
¸
1
]
1
+
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂ ∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
2
2
2 2
2 2
S
t x
P
P
t x
P
x
P
t t x
Q
x
Q
t
x V V
km
km
km
km
km
km km
km
km
km
km
km km
km
k m k m
= 2 +2Q
Q
+
where = , , , or δ δ
∂
∂ ∂
¸
1
]
1
1
∂
∂ ∂
¸
1
]
1
+
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂ ∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
2
2
2 2
2 2
S
t x
P
P
t x
P
x
P
t t x
Q
x
Q
t
x V V
mk
km
mk
mk
km
mk mk
km
mk
mk
km
mk mk
km
k m k m
= 2 +2Q
Q
+
where = , , , or δ δ
Using the same process for phase shift angles, the following partial derivatives can be found.
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − −
P P
V V b g
k
km
km
km
k m km k m km k m
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − + −
Q Q
V V b g
k
km
km
km
k m km k m km k m
α α
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − + −
P P
V V b g
m
km
mk
km
k m mk m k mk m k
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − − −
Q Q
V V b g
m
km
mk
km
k m mk m k mk m k
α α
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − + −
2
2
2
2
P P
V V g b
k
km
km
km
k m km k m km k m
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − − −
2
2
2
2
Q Q
V V g b
k
km
km
km
k m km k m km k m
α α
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − + −
2
2
2
2
P P
V V g b
m
km
mk
km
k m mk m k mk m k
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂
¸
1
]
1
·
∂
∂
¸
1
]
1
· − − − −
2
2
2
2
Q Q
V V g b
m
km
mk
km
k m mk m k mk m k
α α
δ δ δ δ sin cos
79
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − −
2 2
P
V
P
V
V b g
k
km k
km
km k
m km k m km k m
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − −
2 2
P
V
P
V
V b g
k
km m
km
km m
k km k m km k m
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
P P
V V b g
k
km k
km
km k
k m km k m km k m
α δ α δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − + −
2 2
P P
V V b g
k
km m
km
km m
k m km k m km k m
α δ α δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − + −
2 2
Q
V
Q
V
V b g
k
km k
km
km k
m km k m km k m
α α
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − + −
2 2
Q
V
Q
V
V b g
k
km m
km
km m
k km k m km k m
α α
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − −
2 2
Q Q
V V b g
k
km k
km
km k
k m km k m km k m
α δ α δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
Q Q
V V b g
k
km m
km
km m
k m km k m km k m
α δ α δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
P
V
P
V
V b g
m
km k
mk
km k
m mk m k mk m k
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
P
V
P
V
V b g
m
km m
mk
km m
k mk m k mk m k
α α
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
P P
V V b g
m
km k
mk
km k
k m mk m k mk m k
α δ α δ
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − + −
2 2
P P
V V b g
m
km m
mk
km m
k m mk m k mk m k
α δ α δ
δ δ δ δ sin cos
80
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
Q
V
Q
V
V b g
m
km k
mk
km k
m mk m k mk m k
α α
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − − −
2 2
Q
V
Q
V
V b g
m
km k
mk
km k
k mk m k mk m k
α α
δ δ δ δ sin cos
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − −
2 2
Q Q
V V b g
m
km k
mk
km k
k m mk m k mk m k
α δ α δ
δ δ δ δ cos sin
( ) ( )
[ ]
∂
∂ ∂
¸
1
]
1
·
∂
∂ ∂
¸
1
]
1
· − − + −
2 2
Q Q
V V b g
m
km m
mk
km m
k m mk m k mk m k
α δ α δ
δ δ δ δ cos sin
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
S
P
P
x V V
km
km
km
km
km
km
km
km
k m k m
2
α α α
δ δ = 2 +2Q
Q
where = , , , or
∂
∂
¸
1
]
1
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
S
P
P
x V V
mk
km
mk
mk
km
mk
mk
km
k m k m
2
α α α
δ δ = 2 +2Q
Q
where = , , , or
∂
∂ ∂
¸
1
]
1
1
∂
∂ ∂
¸
1
]
1
+
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂ ∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
2
2
2 2
2 2
S
x
P
P
x
P
x
P
x
Q
x
Q
x V V
km
km
km
km
km
km km
km
km
km
km
km km
km
k m k m
α α α α α
δ δ
= 2 +2Q
Q
+
where = , , , or
∂
∂ ∂
¸
1
]
1
1
∂
∂ ∂
¸
1
]
1
+
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂ ∂
¸
1
]
1
∂
∂
¸
1
]
1
∂
∂
¸
1
]
1
2
2
2 2
2 2
S
x
P
P
x
P
x
P
x
Q
x
Q
x V V
mk
km
mk
mk
km
mk mk
km
mk
mk
km
mk mk
km
k m k m
α α α α α
δ δ
= 2 +2Q
Q
+
where = , , , or
81
APPENDIX E. SIXBUS SAMPLE POWER SYSTEM
Appendix E contains information on the sixbus sample power system discussed in the thesis.
A oneline diagram of the system is shown in Figure E.1.
91 MW
92 MVR
100 MW
100 MW
20 MVR
196 MW
27 MVR
1.00 PU
1.00 PU
1.00 PU
Bus 3
Bus 2
Bus 4
100 MW
20 MVR
20 MVR
20 MVR
100 MW
86 MW
237 MW
23 MVR
60 MVR
1.00 PU
Bus 5
Bus 6
0.97 PU
0.97 PU
100 MW
50 MVR
Bus 1
100 MW
10 MVR
14.85 $/MWH
14.19 $/MWH
13.37 $/MWH
13.85 $/MWH
15.25 $/MWH
15.18 $/MWH
0.00 $/MVRH
0.23 $/MVRH
0.00 $/MVRH
0.11 $/MVRH
0.00 $/MVRH
0.00 $/MVRH
Figure E.1 OneLine Diagram of SixBus System
The line characteristics of the system are shown in Table E.1. The bus characteristics of the
system are shown in Table E.2. The economic information of the system is shown in Table E.3.
Table E.1 Line characteristics for sixbus system
From Bus To Bus Circuit Resistance
[p.u.]
Reactance
[p.u.]
Line Charging
[p.u.]
Line Limit
[MVA]
1 2 1 0.04 0.08 0.02 100
1 5 1 0.04 0.08 0.02 100
2 4 1 0.04 0.08 0.02 100
3 5 1 0.04 0.08 0.02 100
3 6 1 0.04 0.08 0.02 100
4 5 1 0.04 0.08 0.02 50
4 6 1 0.04 0.08 0.02 100
82
Table E.2 Bus characteristics for sixbus system
Bus
Number
Load
[MW]
Load
[MVAR]
Min Generation
[MW]
Max Generation
[MVAR]
1 100 20 50 250
2 100 20 50 250
3 100 20 50 250
4 100 20 50 250
5 100 50 0 0
6 100 10 0 0
Table E.3 Economic information for sixbus system
Generator
Bus
a
$
hr
¸
1
]
1
b
$
MW hr
¸
1
]
1
c
$
MW hr
2
¸
1
]
1
1 105 12.0 0.0120
2 96 9.6 0.0096
3 105 13.0 0.0130
4 94 9.4 0.0094
83
APPENDIX F. TWENTYTHREE BUS SAMPLE POWER SYSTEM
Appendix F contains information on the twentythree bus sample power system discussed in
the thesis. A oneline diagram of the system is shown in Figure F.1.
Bus 1
Bus 2
Bus 3
Bus 4
Bus 5
Bus 6
Bus 7
Bus 8
Bus 9
Bus 23
Bus 25
Bus 28
Bus 101
Bus 102
Bus 103
Bus 104
Bus 105
Bus 122
Bus 201
Bus 202
Bus 203
Bus 204
Bus 224
1.03 PU
700 MW
464 MW
1.03 PU
200 MW
763 MW
450 MW
0.98 PU
1.02 PU
200 MW
0.99 PU
350 MW
543 MW
0.97 PU
300 MW 0.96 PU
200 MW
1.03 PU
300 MW 200 MW
657 MW
1.03 PU
1.01 PU
1.01 PU
359 MW
1.01 PU
250 MW
0.98 PU
75 MW
1.01 PU
350 MW
1.03 PU
430 MW
0.99 PU
100 MW
1.00 PU
150 MW
340 MW
1.01 PU
0.99 PU
225 MW
1.02 PU
175 MW
400 MW
300 MW
0.98 PU
1.02 PU
432 MW
125 MW
1.03 PU
97 MVR
106 MVR
106 MVR
98 MVR
106 MVR
97 MVR
16.55 $/MWH
13.90 $/MWH
13.57 $/MWH
14.52 $/MWH
17.25 $/MWH
16.64 $/MWH
16.51 $/MWH
15.58 $/MWH
16.50 $/MWH
15.22 $/MWH
13.39 $/MWH
13.49 $/MWH
13.16 $/MWH
13.68 $/MWH
13.53 $/MWH
13.44 $/MWH
13.51 $/MWH
13.49 $/MWH
13.61 $/MWH
12.86 $/MWH
13.27 $/MWH
12.48 $/MWH
12.86 $/MWH
1.000 tap
1.000 tap
1.000 tap
Figure F.1 OneLine Diagram of the TwentyThree Bus System
The line characteristics of the system are shown in Table F.1. The bus characteristics of the
system are shown in Table F.2. The economic information of the system is shown in Table F.3.
84
Table F.1 Line characteristics for twentythree bus system
From Bus To Bus Circuit Resistance
[p.u.]
Reactance
[p.u.]
Line Charging
[p.u.]
Line Limit
[MVA]
1 2 1 0.015 0.040 0.0005 1000
3 23 1 0.010 0.050 0.0000 1000
6 7 1 0.030 0.070 0.0000 1000
6 9 1 0.000 0.100 0.0000 1000
2 6 1 0.015 0.040 0.0006 1000
8 9 1 0.010 0.030 0.0050 1000
5 7 1 0.030 0.070 0.0000 1000
4 7 1 0.020 0.100 0.0005 1000
3 4 1 0.030 0.100 0.0000 1000
1 3 1 0.025 0.060 0.0005 1000
8 28 1 0.010 0.050 0.0000 1000
25 28 1 0.002 0.009 0.0008 1000
23 25 1 0.002 0.009 0.0008 1000
5 25 1 0.010 0.050 0.0000 1000
102 122 1 0.010 0.050 0.0000 1000
122 23 1 0.000 0.100 0.0000 1000
204 224 1 0.010 0.050 0.0000 1000
224 28 1 0.001 0.007 0.0005 500
105 201 1 0.010 0.030 0.0005 500
201 203 1 0.020 0.060 0.0005 1000
203 204 1 0.020 0.060 0.0000 1000
202 204 1 0.020 0.060 0.0000 1000
201 202 1 0.020 0.050 0.0005 1000
102 101 1 0.020 0.060 0.0000 1000
102 104 1 0.015 0.040 0.0008 1000
104 105 1 0.020 0.060 0.0005 1000
101 103 1 0.030 0.070 0.0000 1000
103 105 1 0.000 0.100 0.0000 1000
85
Table F.2 Bus characteristics for twentythree bus system
Bus
Number
Load
[MW]
Load
[MVAR]
Min Generation
[MW]
Max Generation
[MW]
1 125 50 100 1000
2 300 100 0 0
3 250 50 100 600
4 75 10 0 0
5 350 100 150 1000
6 150 35 100 400
7 100 10 0 0
8 225 75 0 0
9 175 50 100 800
23 0 0 0 0
25 0 0 0 0
28 0 0 0 0
101 700 200 100 600
102 200 20 200 1000
103 350 100 0 0
104 450 75 0 0
105 200 20 100 600
122 0 0 0 0
201 300 100 0 0
202 200 20 0 0
203 300 100 50 200
204 0 0 100 800
224 0 0 0 0
Table F.3 Economic information for sixbus system
Generator
Bus
a
$
hr
¸
1
]
1
b
$
MW hr
¸
1
]
1
c
$
MW hr
2
¸
1
]
1
1 55.43 7.9181 0.005279
3 58.90 7.6570 0.007657
5 64.28 7.7141 0.006428
6 58.90 7.6570 0.007657
9 80.47 7.5102 0.006897
101 72.94 7.2937 0.009981
102 74.24 7.8609 0.004367
105 65.59 7.5255 0.008285
203 65.91 7.5330 0.008161
204 70.81 7.9140 0.004165
86
APPENDIX G. SEVENBUS SAMPLE POWER SYSTEM
Appendix G contains information on the twentythree bus sample power system discussed
in the thesis. A oneline diagram of the system is shown in Figure G.1.
Bus 2
Bus 3
Bus 4 Bus 5
Bus 7
Bus 1
Bus 6
0.99 PU
1.00 PU
1.00 PU 1.04 PU
1.01 PU
1.01 PU
1.04 PU
50 MW
10 MVR
50 MW
10 MVR
100 MW
30 MVR
100 MW
30 MVR
50 MVR
50 MW
10 MVR
50 MW
10 MVR
150 MW
60 MVR
1.0000 tap
1.0000 tap
136 MW
81 MVR
431 MW
59 MVR
20.11 $/MWH
0.02 $/MVRH
18.51 $/MWH
0.36 $/MVRH
18.50 $/MWH
0.28 $/MVRH
18.18 $/MWH
0.21 $/MVRH
19.07 $/MWH
0.00 $/MVRH
16.75 $/MWH
0.00 $/MVRH
18.19 $/MWH
0.32 $/MVRH
Figure G.1 OneLine Diagram of the SevenBus System
The line characteristics of the system are shown in Table G.1. The bus characteristics of the
system are shown in Table G.2. The economic information of the system is shown in Table G.3.
Table G.1 Line characteristics for sevenbus system
From
Bus
To
Bus
R
[p.u.]
X
[p.u.]
C
[p.u.]
Limit
[MVA]
1 3 0.03 0.05 0.00 1000
2 1 0.03 0.05 0.00 1000
2* 4* 0.00 0.02 0.00 1000
3* 5* 0.11 0.02 0.00 1000
4 6 0.07 0.01 0.00 1000
6 7 0.06 0.10 0.00 1000
7 5 0.07 0.11 0.00 1000
87
* Note: Elements from buses 24 and 35 are transformers with tap ranges of 0.9  1.1 p.u.
Table G.2 Bus characteristics for sevenbus system
Bus
Number
Load
[MW]
Load
[MVAR]
Min Gen.
[MW]
Max Gen.
[MW]
1 150 60 0 1000
2 50 10 0 0
3 50 10 0 0
4 50 10 0 0
5 50 10 0 0
6* 100 30 0 0
7 100 30 0 1000
* A 50 MVAR capacitor bank is at bus 6
Table G.3 Economic information for sevenbus system
Generator
Bus
a
$
hr
¸
1
]
1
b
$
MW hr
¸
1
]
1
c
$
MW hr
2
¸
1
]
1
1 90 9.0 0.0090
7 115 15.0 0.0150
88
APPENDIX H. THREEBUS SAMPLE POWER SYSTEM
Appendix H contains information on the twentythree bus sample power system discussed
in the thesis. A oneline diagram of the system is shown in Figure H.1.
1.04 PU
0.9710 PU
504 MW
193 MVR
411 MW
70 MVR
300 MW
75 MW
Bus 1
Bus 2
100 MW
1.04 PU
Bus 3
100 MVR
Area Two
Area One
14 MVR 500 MW
100 MVR
30 MVR
50.18 $/MWH
0.00 $/MVRH
20.22 $/MWH
0.00 $/MVRH
22.85 $/MWH
0.94 $/MVRH
72 MW
21 MVR
72 MVR
73 MW
15 MVR
238 MW
55 MVR
73 MW
64 MVR
227 MW
22 MVR
Figure H.1 OneLine Diagram of the ThreeBus System
The line characteristics of the system are shown in Table H.1. The bus characteristics of the
system are shown in Table H.2. The economic information of the system is shown in Table H.3.
Table H.1 Line characteristics for threebus system
From
Bus
To
Bus
R
[p.u.]
X
[p.u.]
C
[p.u.]
Limit
[MVA]
1 2 0.02 0.08 0.00 500
1 3 0.03 0.12 0.00 500
2 3 0.02 0.06 0.00 600
89
Table H.2 Bus characteristics for threebus system
Bus
Number
Load
[MW]
Load
[MVAR]
Min Gen.
[MW]
Max Gen.
[MW]
1 500 100 100 800
2 300 100 0 0
3 100 30 100 800
Table H.3 Economic information for threebus system
Generator
Bus
a
$
hr
¸
1
]
1
b
$
MW hr
¸
1
]
1
c
$
MW hr
2
¸
1
]
1
1 100 25.0 0.0250
3 150 13.0 0.0100
90
REFERENCES
[1] S. Hunt and G. Shuttleworth, “Unlocking the GRID,” IEEE Spectrum, Vol. 33,
No. 7, July 1996, pp. 2025.
[2] B. R. Barkovich and D. Hawk, “Charting a new course in California,” IEEE
Spectrum, Vol. 33, No. 7, July 1996, pp. 2631.
[3] M. G. Morgan and S. Talukdar, “Nurturing R&D,” IEEE Spectrum, Vol. 33, No.
7, July 1996, pp. 32,33.
[4] H. Rudnick, “Pioneering electricity reform in South America,” IEEE Spectrum,
Vol. 33, No. 8, August 1996, p. 3844.
[5] R. D. Tabors, “Lessons from the UK and Norway,” IEEE Spectrum, Vol. 33, No.
8, August 1996, pp. 4549.
[6] R. D. Masiello, “It’s put up or shut up for grid controls,” IEEE Spectrum, Vol. 33,
No. 8, August 1996, pp. 50,51.
[7] Federal Energy Regulatory Commission of the United States of America,
“Promoting wholesale competition through open access nondiscriminatory
transmission services by public utilities,” Docket No. RM95800, March 1995.
[8] “WSCC Unfolds Causes of the July 2 Disturbance,” IEEE Power Engineering
Review, Vol. 16, No. 9, September 1996, pp. 5,6.
[9] The joint PSERC/EPRI workshop on underlying technical issues in electricity
deregulation, Tucson, Arizona, April 2527, 1996.
[10] J. Carpienter, “Contribution e l’étude do Dispatching Economique,” Bulletin
Society Française Electriciens, Vol. 3, August 1962.
[11] A. J. Wood and B. F. Wollenberg, Power Generation Operation and Control,
New York, NY: John Wiley & Sons, Inc., 1996, pp. 39,517.
[12] H. W. Dommel and W. F. Tinney, “Optimal Power Flow Solutions,” IEEE
Transactions on Power Apparatus and Systems, Vol. PAS87, October 1968, pp.
18661876.
[13] D. I. Sun, B. Ashley, B. Brewer, A. Hughes and W. F. Tinney, “Optimal Power
Flow by Newton Approach,” IEEE Transactions on Power Apparatus and
Systems, Vol. PAS103, October 1984, pp. 28642880.
91
[14] O. Alsac, J. Bright, M. Prais and B. Stott, “Further Developments in LPBased
Optimal Power Flow,” IEEE Transactions on Power Systems, Vol. 5, No. 3,
August 1990, pp. 697711.
[15] Y. Wu, A. S. Debs and R. E. Marsten, “Direct Nonlinear PredictorCorrector
PrimalDual Interior Point Algorithm for Optimal Power Flows,” 1993 IEEE
Power Industry Computer Applications Conference, pp. 138145.
[16] M. Huneault and F. D. Galiana, “A Survey of the Optimal Power Flow
Literature,” IEEE Transactions on Power Systems, Vol. 6, No. 2, May 1991, pp.
762770.
[17] T. J. Overbye, P. W. Sauer, C. M. Marzinzik and G. Gross, “A UserFriendly
Simulation Program for Teaching Power System Operations,” IEEE Transactions
on Power Systems, Vol. 10, No. 4, November 1995, pp. 17251733.
[18] W. F. Tinney and C. E. Hart, “Power Flow Solution by Newton’s Method,” IEEE
Transactions on Power Apparatus and Systems, Vol. PAS86, No. 11, November
1967, pp. 18661876.
[19] D. G. Luenberger, Linear and NonLinear Programming, Reading, MA:
AddisonWesley Publishing Company, 1984, pp. 295392,423450.
[20] J. J. Grainger and W. D. Stevenson, Power System Analysis, New York, NY:
McGraw Hill, Inc., 1994, pp. 33,330.
[21] E. Liu, A. D. Papalexopoulos and W. F. Tinney, “Discrete Shunt Controls in A
Newton Optimal Power Flow,” IEEE Transactions on Power Systems, Vol. 7, No.
4, November 1992, pp. 15191528.
[22] J. A. Momah, R. J. Koessler, M. S. Bond, B. Stott, D. Sun, A. D. Papalexopoulos
and P. Ristanovic, “Challenges to Optimal Power Flow,” 1996 IEEE/PES Winter
Meeting, 96 WM 3129 PWRS, January 2125, 1996, Baltimore, MD.
[23] C.A. Gross, Power System Analysis, New York, NY: John Wiley & Sons, Inc.,
1979.
[24] G. E. Forsythe and C. B. Moler, Computer Solution of Linear Algebraic Systems,
Englewood Cliffs, N.J.: PrenticeHall, 1967, Chapters 9 and 16.
[25] S. M. Chan and V. Brandwajn, “Partial Matrix Refactorization,” IEEE
Transactions on Power Systems, Vol. PWRS1, No. 1, February 1986, pp. 193
200.
[26] J. D. Weber, M. J. Laufenberg, T. J. Overbye and P. W. Sauer, “Assessing the
Value of Reactive Power Services in Electric Power Systems,” Conference on
Unbundled Power Quality Services, Key West, Florida, November 1719, 1996.
92
[27] R. D. Tabors, “Transmission System Management and Pricing: New Paradigms
and International Comparisons,” IEEE Transactions on Power Systems, Vol. 9,
No. 1, February 1994, pp. 206215.
[28] D. Shirmohammadi, X.V. Filho, B. Gorenstin and M. Pereira, “Some fundamental
technical concepts about cost based transmission pricing,” IEEE Transactions on
Power Systems, Vol. 11, No. 2, May 1996, pp. 10021008.
[29] D. A. Wismer and R. Chattergy, Introduction to NonLinear Optimization, A
Problem Solving Approach, New York, NY: North Holland Publishing, 1978.
REDBORDERED FORM
ii
ABSTRACT
In this thesis, a Newtonbased optimal power flow (OPF) is developed for implementation into a power system simulation environment. The OPF performs all system control while
maintaining system security. System controls include generator megawatt outputs, transformer taps, and transformer phase shifts, while maintenance of system security ensures that no power system component’s limits are violated. Special attention is paid to the heuristics important to creating an OPF which achieves solution in a rapid manner. Finally, sample applications of the OPF are discussed. These include transmission line overload removal, transmission system control, available transfer capability calculation (ATC), real and reactive power pricing, transmission component valuation, and transmission system marginal pricing.
iii
ACKNOWLEDGEMENTS
I would like to thank Professor Thomas Overbye for his knowledge, guidance and support throughout my graduate studies. I would also like to thank the University of Illinois Power Affiliates Program and the University of Illinois Graduate College for their generous financial support. I would also like to thank my professors and friends from my undergraduate studies while at the University of Wisconsin  Platteville. The preparation and experience I gained there were invaluable. Special thanks go to Dr. Mesut Muslu and Dr. Richard Shultz. Finally, I would like to thank my family. My brothers Brian and Scott continue to be there for support and friendship. Most of all I thank my mother and father, Jan and Gene. They continue to supply the unconditional love and support which allow me and my brothers to achieve what we have and will.
iv
TABLE OF CONTENTS
Page 1. INTRODUCTION................................................................................................................... 1 1.1 1.2 1.3 1.4 2. Motivation ........................................................................................................................... 1 Literature Survey ................................................................................................................. 3 Goals of the OPF ................................................................................................................. 4 Overview ............................................................................................................................. 5
DEVELOPMENT OF NEWTONBASED OPTIMAL POWER FLOW............................... 7 2.1 Background on Newton’s Method....................................................................................... 7 2.1.1 Problem statement ....................................................................................................... 7 2.1.2 Development of Lagrangian, gradient and Hessian..................................................... 8 2.1.3 Application of inequality constraints........................................................................... 9 2.1.4 Solution method .......................................................................................................... 9 2.2 Application of Newton’s Method to OPF ......................................................................... 11 2.2.1 The objective function............................................................................................... 11 2.2.2 Equality constraints ................................................................................................... 12 2.2.3 Inequality constraints................................................................................................. 13 2.2.4 Soft constraints by using penalty functions............................................................... 14 2.2.5 Treatment of discrete variables ................................................................................. 17 2.2.6 Summary of optimal power flow problem ................................................................ 17 2.2.7 Summary of Lagrangian terms .................................................................................. 18 2.2.8 Calculation of gradient and Hessian.......................................................................... 19 2.2.9 Solution of the optimal power flow........................................................................... 20 2.3 A Sample Case Illustrating OPF Algorithm Process......................................................... 21 2.4 Information Gained from the OPF Solution...................................................................... 22
3.
HEURISTICS OF THE OPTIMAL POWER FLOW SOLUTION ...................................... 24 3.1 3.2 3.3 3.4 3.5 Classification of OPF Variables ........................................................................................ 24 Implementation of Sparse Matrix Techniques .................................................................. 25 Determination of the Set of Binding Inequality Constraints ............................................. 28 Searching Algorithms........................................................................................................ 30 Solution of an OPF Repeatedly Over Time....................................................................... 31
4. USES OF AN OPTIMAL POWER FLOW IN A POWER SYSTEM SIMULATION ENVIRONMENT.......................................................................................................................... 33 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 Example Line Overload Removal ..................................................................................... 33 Use for Bus Real and Reactive Power Pricing .................................................................. 34 Use for Area Real Power Pricing ...................................................................................... 36 Example Transformer Tap Control ................................................................................... 38 System MVAR Control Using Transformer Taps [26] ................................................... 41 Transmission Line Valuation by TimeDomain Simulation ............................................. 45 Capacitor Bank Valuation by TimeDomain Simulation .................................................. 47 Limit on Available Transfer Capability (ATC) Due to a Voltage Constraint [26] .......... 48 v
4.9 Transmission System Pricing Through ShortRun Marginal Costing (SRMC)................ 52 5. CONCLUSION ..................................................................................................................... 56
APPENDIX A. ECONOMIC INTERPRETATION OF THE LAGRANGE MULTIPLIERS .... 58 APPENDIX B. CALCULATION OF THE GRADIENT OF THE LAGRANGIAN ................... 61 APPENDIX C. CALCULATION OF THE HESSIAN OF THE LAGRANGIAN ...................... 64 APPENDIX D. SUMMARY OF DERIVATIVE CALCULATIONS .......................................... 68 APPENDIX E. SIXBUS SAMPLE POWER SYSTEM.............................................................. 82 APPENDIX F. TWENTYTHREE BUS SAMPLE POWER SYSTEM...................................... 84 APPENDIX G. SEVENBUS SAMPLE POWER SYSTEM....................................................... 87 APPENDIX H. THREEBUS SAMPLE POWER SYSTEM....................................................... 89 REFERENCES.............................................................................................................................. 90
vi
....................................................................................... 46 Table 4..... twoarea system ...2 Results for various interchanges with transformer taps left inactive ... 24 Table 4.................... 37 Table 4......................................................................2 Bus characteristics for sixbus system ....................3 Economic information for threebus system ..... 83 Table F.......... twoarea system where no limit is encountered...... 83 Table E............. 90 vii .............................2 Bus characteristics for threebus system ................................................ 90 Table H............................... 87 Table G....................... 89 Table H.......................................................1 Simulation data for sixbus system with transmission lines removed ................................................... 42 Table 4......................................................... 48 Table 4..............................1 Line characteristics for sixbus system ...2 Bus characteristics for twentythree bus system ......................1 Summary of tap ratio control experiment............................1 Simulation data for sixbus system with transmission lines removed .......................................................................................................... 86 Table F.............. 44 Table 4.... 88 Table G..........................3 Economic information for sixbus system ...1 Line characteristics for sevenbus system ................................... 82 Table E.................................. 54 Table E..............................................1 OPF problem variables.................................1 Comparison of two systems ...............................3 Results for various interchanges with transformer taps active....................................................2 Bus characteristics for sevenbus system ............. 41 Table 4.... 44 Table 4...........LIST OF TABLES Table Page Table 3.............3 Economic information for sevenbus system .............................................. 53 Table 4...... 85 Table F.. 88 Table H.................... 86 Table G.1 Line characteristics for threebus system .......1 Line characteristics for twentythree bus system ....3 Economic information for sixbus system .........................................................................1 Sample transaction scenarios ...........................................................................2 SRMC calculation for sixbus.............................................1 SRMC calculation for sixbus.........
............................. SingleArea System on OPF Control......1 Sixbus............... 35 Figure 4............................................ 33 Figure 4................1 OneLine Diagram of the SevenBus System ...........5 MW Undertaken ......... 37 Figure 4......................................2...................1 OneLine Diagram of SixBus System ..... 36 Figure 4................................ 47 Figure 4.......... 38 Figure 4.................................1 Transformer Model ............... 50 Figure 4.................... OneArea.....................LIST OF FIGURES Figure Page Figure 2........... SevenBus..1 Multiarea System with Scheduled Interchanges ..........2 Transaction of 65................................................................. 46 Figure 4..............................3........................... 10 Figure 2.......1 TwentyThree Bus System on OPF Control with Tap Control Off .....2 Penalty Function for Line MVA Flow Limit..................................... 84 Figure G... SevenBus.............. 41 Figure 4. OneArea System with Taps Inactive .. SingleArea System Not on OPF Control......1 Line Limit from Bus 5 to 4 Raised to 100 MVA ....................................1 Sixbus................ System with Taps Active ......................................... 51 Figure 4............ 74 Figure E. 34 Figure 4.. TwoArea System on OPF Control.....2 TwentyThree Bus System with Tap Control On......................... 16 Figure 2.....4 ThreeBus Example at Maximum Power Transfer (45 MVAR Capacitor Support). 30 Figure 4.....1 Sixbus.......1 SevenBus..........................................................................2 Sixbus...1 OneLine Diagram of the TwentyThree Bus System .................1 Load Factor Curve for sixbus System............ 54 Figure D................................................................................................ 89 viii ........1 Penalty Function for Bus Voltage . 43 Figure 4...............................................2 ThreeBus Example at Maximum Power Transfer (15 MVAR Capacitor Support)..............................1 Sample Binary Tree...............1 Load Factor Curve for TwentyThree Bus System ...........................................................1 OneLine Diagram of the ThreeBus System....................1 Newton’s Method Flowchart .....................1 ThreeBus Base Case with No Area Power Transfer..3 ThreeBus Example at Maximum Power Transfer (30 MVAR Capacitor Support)....... 51 Figure 4................................................................... 16 Figure 3.. 87 Figure H..................................................... 82 Figure F................................... 49 Figure 4.............................................................. TwoArea System Undergoing Transaction with Line Limit Doubled .................................................. 41 Figure 4.... TwoArea System with TapChanging Transformers . 13 Figure 2........ 42 Figure 4......................................................................
Element of the real part of network admittance matrix. Inequality constraints. The real power interchange for an area. Coefficients for the quadratic cost curve of a generator Signifies a maximum bound on a variable. The scheduled real power interchange for an area. The real power flow from bus k to bus m. The gradient of the Lagrangian. Vk Magnitude of voltage at bus k. The objective function. Xmax . ix PGk bkm gkm ykm δ Pk Qk km Pkm Qkm Skm Pint Psched f(• ) g(• ) h(• ) L(•) H(•) ∇L(•) ai. Reactive power injection at bus k. The Lagrange function or Lagrangian. k δ tkm Angle of voltage at bus k.NOTATION Throughout this thesis. The real power generated at bus k. while all variables that refer to a scalar will have a subscript next to them. Phasor angle of an element of the network admittance matrix. The reactive power flow from bus k to bus m. bi. α km Transformer phase shift between buses k and m. The MVA flow from bus k to bus m. and ci. Element of the imaginary part of the network admittance matrix. The Hessian of the Lagrangian. Magnitude of an element of the network admittance matrix. Real power injection at bus k. Transformer tap ratio between buses k and m. variables that do not have a subscript next to them may be considered vectors or matrices. Equality constraints.
For Qk. For minimum transformer phase shift constraint. For MVA constraint on line from bus k to m. µ λ List of Lagrange multipliers µ µ µ µ λ λ λ λ λ λ λ λ λ Pk Qk viset int Skm PGih PGil Vih Vil tkm max tkm min αkm αkm For Pk. For minimum bus voltage constraint at bus i. For generator voltage set point. In general. For maximum transformer phase shift constraint. For generator voltage set point. For generator maximum power output constraint at bus i. the reactive power injection at bus k. For generator minimum power output constraint at bus i. For minimum transformer tap ratio constraint. max min x . this is a Lagrange multiplier for an equality constraint. For maximum bus voltage constraint at bus i. For maximum transformer tap ratio constraint .Xmin Signifies a minimum bound on a variable. this is a Lagrange multiplier for an inequality constraint. In general. the real power injection at bus k.
“participants in wholesale power markets will have nondiscriminatory open access to the transmission systems of public utilities. In the United States. and industries were required to purchase their power from the local monopolistic power company. businesses. As stated in [7]. countries have begun to split up these monopolies in favor of the free market. the electric power industry has undergone a considerable change in the past decade and will continue to do so for the next several decades. the change from the regulated monopoly to the free market system has become known as restructuring. the Federal Energy Regulatory Commission (FERC) oversees issues involving the transmission system. but a physical engineering requirement as well. however. It just didn’t appear feasible to duplicate the resources required to connect everyone to the power grid.1 Motivation Throughout the entire world. Over the past decade. FERC presently believes that the only manner in which everyone will be on an equal playing field is to create open access to all. Numerous papers and articles have been written on this topic with a good overview of the topic found in a series of articles written for IEEE Spectrum in July and August of 1996 [1 . One of the cornerstones of any restructuring plan is the ability to operate the transmission system in a manner which is fair to all participants in the industry. INTRODUCTION 1. it will be referred to as restructuring. In the past the electric power industry has been either a governmentcontrolled or a governmentregulated industry which existed as a monopoly in its service region. In the United States.” 1 . For the remainder of this thesis. This was not only a legal requirement. All people.1.6].
justifiable and economic manner • Tools for congestion management (including congestion pricing) • Tools for determining the ISO action before the contingency occurs Economic Problems • Realtime pricing and price risk management services • Tools for operating the power system in the most economical manner • Transaction evaluation tools that enable players to evaluate their own costs • Methodologies for determining the value (cost) of ancillary services in improving efficiency and flexibility UserInterface and Simulation Problems • Market simulation models • Better tools for communication and display of information will permit the better operation of the min ISO scenario 2 . In the past year. the western United States has seen the consequences of pushing the transmission system too hard on two separate occasions. If these engineering problems become overshadowed by short term economic concerns. In a presently unpublished report from a joint PSERC/EPRI workshop. It is very important that these problems be addressed early in the restructuring process.In order to achieve the ideal of open access. many outstanding engineering problems will need to be investigated and tools created for their solution. then the result could be decreased electricity reliability. The two multistate blackouts in the Western States Coordinating Council (WSCC) system in the last several months may be destined to repeat themselves [8]. A partial list of problems that can be addressed from the work done in this thesis follows: Control Problems • Computation of realtime available transfer capability (ATC) • Realtime control of power flows • Tools to relieve congestion in fair. many of these tools were identified [9].
the OPF very adeptly addresses both the control and economic problems.See paper by Alsac et al. it has taken decades to develop efficient algorithms for its solution.The work presented in this thesis utilizes an optimal power flow program. [14] Interior point method . 1. OPF. and Marsten [15] 3 . p. This method has its roots in the common method of economic dispatch used since the 1930s. See [11. In this way. Because the OPF is a very large. the userinterface and simulation problems may also be addressed by implementing the OPF into a power system simulator. The OPF is a natural choice for addressing these concerns because it is basically an optimal control problem. Debs.See paper written by Dommel and Tinney [12] Newton’s method .Also called the equal incremental cost criterion (EICC) method. After creating the OPF program. It also yields valuable economic information and insight into the power system.See paper by Sun et al.2 Literature Survey The optimal power flow problem has been discussed since its introduction by Carpentier in 1962 [10].517]: • Lambda iteration method . p. as the tool for solving these problems. In these ways. the results of the economic and control operations of the OPF can easily be utilized by the user of the program. The majority of the techniques discussed in the literature use one of the following five methods [11. [13] Linear programming method . 39 ] • • • • Gradient method . Many different mathematical techniques have been employed for its solution. The OPF utilizes all control variables to help minimize the costs of the power system operation.See paper by Wu. nonlinear mathematical programming problem.
4 . maximum MVA flows on transmission lines and transformers. The primary goal of a generic OPF is to minimize the costs of meeting the load demand for a power system while maintaining the security of the system. Though it does not discuss the interior point method.An excellent literature survey of the different techniques can be found in a paper by Huneault and Galieana published in 1991 [16]. and steadystate contingency analysis are not addressed. This thesis will explore the application of Newton’s method to the OPF problem. dynamic stability. This will include maximum and minimum outputs for generators. Topics such as transient stability. but in general they can be attributed to the cost of generating power (megawatts) at each generator. it will explore the implementation of a Newton’s method based OPF in the power system simulator POWERWORLD [17]. From the viewpoint of an OPF. it does make reference to over 150 papers on the optimal power flow problem covering all the other methods for solving the OPF. as well as keeping system bus voltages within specified ranges.3 Goals of the OPF Before beginning the creation of an OPF. It should be noted that the OPF only addresses steadystate operation of the power system. The costs associated with the power system may depend on the situation. the maintenance of system security requires keeping each device in the power system within its desired operation range at steadystate. Specifically. 1. it is useful to consider the goals that the OPF will need to accomplish.
4 Overview The OPF program written in conjunction with this thesis uses Newton’s method as its solution algorithm. These functions may include generator control and transmission system control. 1. Chapter 4 will show several sample applications of the OPF. A secondary goal of an OPF is the determination of system marginal cost data. and all other flexible ac transmission system (FACTS) devices. This marginal cost data can aid in the pricing of MW transactions as well as the pricing ancillary services such as voltage support through MVAR support. For the transmission system. It will tackle all of the goals set forth for an OPF except the control of switched shunts and other FACTS devices. the marginal cost data are determined as a byproduct of the solution technique. The sample applications 5 . In solving the OPF using Newton’s method. The control of these may be added at a later time as desired.To achieve these goals. the OPF will control generator MW outputs as well as generator voltage.4 on page 22. This will be discussed later in Section 2. It will lay a framework for the mathematics and engineering behind the OPF computer source code. the OPF may control the tap ratio or phase shift angle for variable transformers. The remainder of this thesis will discuss the development of the OPF. the OPF will perform all the steadystate control functions of the power system. switched shunt control. Chapter 3 will discuss some special heuristics important to creating an OPF which achieves solution in a rapid manner. Chapter 2 of this thesis will discuss the theory of the Newtonbased optimal power flow. For generators.
Finally. and transmission system marginal pricing. available transfer capability (ATC) calculations. transmission component valuation. 6 . Chapter 5 will conclude with a summary and several improvements which would aid in creating a truly useful OPF.discussed will include transmission line overload removal. real and reactive power pricing. transmission system control.
. which describes Newton’s method as well as its quadratic convergence properties.1. 2. 2. 2.1 Problem statement A general minimization problem can be written in the following form. Newton’s method is a very powerful solution algorithm because of its rapid convergence near the solution. It has been the standard solution algorithm for the power flow problem for decades [18]. . and transformer tap ratios will be near 1. .2. System voltages will be near rated system values. generator outputs can be estimated from historical data. Minimize 7 . This property is especially useful for power system applications because an initial guess near the solution is easily attained.u. n (inequality constraints) There are m equality constraints and n inequality constraints and the number of variables is equal to the dimension of the vector x.1 Background on Newton’s Method Newton’s method is wellknown in the area of power systems. . . (the objective function) f(x) subject to: h i (x) = 0. The detailed explanation in [19] is left to the interested reader.0 p. m (equality constraints) g j(x) ≤ 0. A good reference for the theory of Newton’s method is a book by Luenberger [19]. DEVELOPMENT OF NEWTONBASED OPTIMAL POWER FLOW 2. . . This thesis will only cover the process of applying Newton’s method to a minimization problem such as the OPF. i = 1. . j = 1.
314]: ∇x L(z*) = ∇x L([x*. the inequality constraint is active) λ i * = 0 if g(x*) ≤ 0 (i. µ*]) = 0.2. λ*.2 Development of Lagrangian. p. where z = x µ λ . ∇λ L(z*) = ∇λ L([x*. ∇µ L(z*) = ∇µ L([x*. 8 . A gradient and Hessian of the Lagrangian may then be defined.1. µ*] is the optimal solution... µ*]) = 0. µ and λ are vectors of the Lagrange multipliers. according to optimization theory.e. Thus solving the equation ∇z L(z*) = 0 will yield the optimal problem solution. λ*. This sparsity will be exploited in the solution algorithm. λ*. the inequality constraint is not active) µ i * = Real where z* = [x*. and g(x) only includes the active (or binding) inequality constraints. λ*. λ i* ≥ 0 if g(x*) = 0 (i.e. µ*]) = 0. [ L(z ) = f ( x ) + µ T h( x ) + λ T g( x ) = the Lagrangian ] T ∂L( z) Gradient = ∇L(z) = = a vector of the first partial derivatives of the Lagrangian ∂zi ∂ 2 L( z ) ∂xi ∂x j ∂ 2 L( z ) = ∂µ i ∂x j ∂ 2 L( z ) ∂λ i ∂x j ∂ 2 L( z) ∂xi ∂µ j 0 0 ∂ 2 L( z ) ∂xi ∂λ j 0 = 0 ∂ 2 L( z ) 2 Hessian = ∇ L(z) = H = ∂zi ∂z j a matrix of the second partial derivatives of the Lagrangian Note the extremely sparse structure of the Hessian matrix shown. From this. gradient and Hessian The solution of this problem by Newton’s method requires the creation of the Lagrangian as shown below. the KuhnTucker necessary conditions of optimality are [19.
continued enforcement will result in a decrease of the function. 326]. p. While an inequality constraint is being enforced. the determination of those inequality constraints that are active is of utmost importance. A more detailed 9 . the inequality constraints will be handled by separating them into two sets: active and inactive [19. the sign of its associated Lagrange multiplier at solution determines whether continued enforcement of the constraint is necessary. if the multiplier is positive. Therefore. then enforcement will result in an increase of the function. 2. This flowchart will be useful for any generic minimization problem. the Lagrangian only includes those inequalities that are being enforced.1 performs this search for the binding or active constraints. For example. Essentially the Lagrange multiplier is the negative of the derivative of the function that is being minimized with respect to the enforced constraint (see Appendix A for a derivation of this fact). then there is no need to activate the inequality constraint associated with that bus voltage.2.3 Application of inequality constraints It should be noted that special attention must be paid to the inequality constraints of this problem. if a bus voltage is within the desired operating range. For this Newton’s method formulation.1. If it is negative.1. The outer loop of the flow chart in Figure 2.1. For efficient algorithms. A flowchart of this process is shown in Figure 2. the solution of the minimization problem can be found by applying Newton’s method to ∇z L(z) = 0. As noted. and enforcement is thus maintained.4 Solution method Considering the issues discussed above. and enforcement is thus stopped.
Create the Lagrangian given the active inequality constraints.discussion of this flowchart will be reserved for the following section.∆z NO Check tolerance Determine new set of inequalities to enforce using Lagrange multipliers ∆z< ε YES NO Are correct inequalities enforced? YES Problem Completed Figure 2. Make initial guess of vector z = [ x µ λ ]T and which inequality constraints to enforce. Calculate the Hessian and gradient of the Lagrangian Solve the equation [H] ∆z = ∇L(z) for ∆z Calculate new z znew = zold . There the application of Newton’s method to the optimal power flow problem will be discussed.1 Newton's Method Flowchart 10 .
will be introduced. The quadratic cost model for generation of power will be utilized: CPGi = ai + bi PGi + ci PGi 2 where PGi is the amount of generation in megawatts at generator i. It will reflect the desire to minimize the costs of the system. f ( x ) = ∑ (ai + bi PGi + ci PGi 2 ) i This objective function will minimize the total system costs. the concept of soft constraints using penalty functions will be introduced.3. This section of the thesis will discuss the application of Newton’s method in a manner that will achieve this desired goal. At this point. and Hessian will be summarized. the objective function. First.2. Penalty functions also model the need to maintain system security. Then the equality and inequality constraints will be discussed. Finally. 11 . The objective function for the entire power system can then be written as the sum of the quadratic cost model at each generator. the goal of the OPF is to minimize the costs of meeting the load demand for a power system while maintaining the security of the system. f(x). all terms in the Lagrangian.2. and does not necessarily minimize the costs for a particular area within the power system.2 Application of Newton’s Method to OPF As discussed in Section 1. These constraints model the physical laws of the power system as well as the need to maintain system security.1 The objective function The objective function for the OPF reflects the costs associated with generating power in the system. gradient. 2.
2 Equality constraints The equality constraints of the OPF reflect the physics of the power system as well as the desired voltage set points throughout the system. In this case. for multiarea power systems. 330]. p. it will 12 . The physics of the power system are enforced through the power flow equations which require that the net injection of real and reactive power at each bus sum to zero. VGi − VGi setpo int = 0 Finally.2. With the loads given and the interchanges constrained for two areas. an equality constraint for each generator is added. It is also common for the power system operators to have voltage set points for each generator. This adds an equality constraint for all but one area. Pinterchange − Psceduled interchange = tie lines ∑ [P ] − P km sceduled interchange =0 This last area must not have the equality constraint and essentially becomes a “slack area. p. Thus. a contractual constraint requires that the net power interchange be equal to the scheduled power interchange.” Consider a simple threearea system with scheduled interchanges as shown in Figure 2. 33]. the last area will be forced to have its interchanges set as planned. adding another constraint is not needed. For an explanation of forming the network admittance matrix see Grainger and Stevenson [20. Indeed.1.2. Pk = 0 = Vk ∑ Vm g km cos(δk − δm ) + bkm sin(δk − δ m ) − PGk + PLk m=1 N Qk [ [ = 0 = V ∑ [V [g N k m=1 m ]] km sin (δk − δm ) − bkm cos(δk − δ m ) − QGk + QLk ]] For a derivation of the power flow equations see Grainger and Stevenson [20.
cause numerical problems if added because the equations associated with this constraint are dependent on the other interchange constraints and would thus lead to a singular Hessian matrix. tap changing transformers. Generators have maximum and minimum output powers and reactive powers which add inequality constraints. Load3 AREA THREE Load2 P23int AREA TWO P13int AREA ONE P12int Load1 Figure 2. and phase shifting transformers. PGi min ≤ PGi ≤ PGi max QGi min ≤ QGi ≤ QGi max 13 .3 Inequality constraints The inequality constraints of the OPF reflect the limits on physical devices in the power system as well as the limits created to ensure system security.2.1 Multiarea System with Scheduled Interchanges 2. This section will lay out all the necessary inequality constraints needed for the OPF implemented in this thesis. Physical devices that require enforcement of limits include generators.
bus voltages usually have maximum and minimum magnitudes. The determination of these MVA ratings will not be of concern in this thesis.4 Soft constraints by using penalty functions One issue sometimes encountered when trying to solve a minimization problem is the nonexistence of a feasible solution. It is assumed that they are given. These limits again require the addition of inequality constraints. the constraint used in the OPF will limit the square of the MVA flow on a transformer or transmission line. One way to avoid 14 . these MVA ratings will result in another inequality constraint.Load tap changing transformers have a maximum and a minimum tap ratio which can be achieved and phase shifting transformers have a maximum and a minimum phase shift. Vi min ≤ Vi ≤ Vi max 2 2 2. Regardless. These ratings may come from thermal ratings (current ratings) of conductors. or they may be set to a level due to system stability concerns.2. Both of these create inequality constraints. which can be achieved. S km − S km max ≤ 0 To maintain the quality of electrical service and system security. t km min ≤ t km ≤ t km max α km min ≤ α km ≤ α km max For the maintenance of system security. To make the mathematics less complex. power systems have transmission line as well as transformer MVA ratings. Essentially this means that too many constraints have been added to the problem and no solution exists which obeys all of the constraints.
k Vi − Vi max ( ) ( ) Figure 2. Ideally. In the OPF presented in this thesis. soft inequality constraints are available for transmission line MVA limits as well as bus voltage limits. soft equality constraints are not used. 15 . and the generator set points of a power system are normally not moved around frequently. because of the nature of the equality constraints in the OPF problem. a penalty function will be very small near a limit and increase rapidly as the limit is violated more.1 and Figure 2.2 show the graphs of these penalty functions used in the OPF. The word “soft” signifies that the constraint is not absolutely enforced. . Penalty functions are added to the objective function of the minimization problem. For the inequality constraints. the penalty functions offer a viable option. 2 2 2 Wkm = k S km − S km max 2 Penalty Functions k Vi min − Vi Vi < Vi min . The power flow equations can not be violated as they are imposed by physics. 443] which meets the requirements of a penalty function and is also easily differentiated for use by Newton’s method. Vi min ≤ Vi ≤ Vi max 0 Wi = 2 Vi > Vi max .this issue is to implement soft inequality constraints in the form of penalty functions. The soft constraint only encourages the solution to meet the constraint by enforcing a penalty if the constraint is not met. A wellsuited penalty function for use in Newton’s method is the quadratic penalty function [19. p. In the OPF problem.
After doing trial and error experimentation. one can control the importance given to the limit. 16 . For soft transmission line constraints. By adjusting it to smaller values. For soft bus voltage constraints. the penalty function has a value of zero. Another advantage of the quadratic penalty function is the ability to control how hard or soft to make the constraint. While the inequality constraint is not violated. As the constraint begins to be violated. the penalty function quickly increases.2 Penalty Function for Line MVA Flow Limit Note that these penalty functions fit the requirements perfectly. k was chosen to be $200/V2.1 Penalty Function for Bus Voltage Transmission Line MVA Flow Penalty Function Value Skm max2 Skm2 Figure 2. For very large values of k. the quadratic penalty function behaves much like a hard constraint. values for k were chosen for use in the OPF of this thesis.Voltage Penalty Function Value Vi min Vi max Vi Figure 2. k was chosen to be $100/MVA4.
16 above and 16 below). the optimal power flow problem can be written in the following form. although this is not true for them. three problems arise from this methodology.6 Summary of optimal power flow problem In summary. This problem should be addressed in future work. 17 . One possible solution for this problem is to round the optimal setting found assuming a continuous tap to the nearest discrete tap.. First. For example.e.2. This could be done for all transformers. This final problem is addressed in references [21. The OPF algorithm as presented in this thesis also assumes this for the tap ratios and phase shift angles of variable transformers. there is no guarantee that the rounded solution is the optimal solution. 2. the constraint Vi min ≤ Vi ≤ Vi max allows Vi to take on all values within the specified range.2.2. Finally. some constraints may be violated. this methodology will not work well for discrete variables that have very large step sizes such as switched capacitor banks. Second the solution may become infeasible after rounding. a typical tap changing transformer has 33 discrete positions (nominal. i.5 Treatment of discrete variables In all the discussions thus far. all variables have been assumed to be continuous. However. Variable transformers have a fixed number of discrete tap positions at which they may operate. 22]. For example.
These terms are summarized as follows: 18 .minimize: generators ∑ (a i + bi PGi + ci PGi 2 ) + Pk = 0 Qk = 0 Vi − Vi set = 0 Pint − Pscheduled = 0 2 2 penalties ∑α i h( x ) = 0 subject to: S km − S km max ≤ 0 PGi − PGi max ≤ 0 PGi min − PGi ≤ 0 Vi − Vi max ≤ 0 Vi min − Vi ≤ 0 g ( x ) ≤ 0 t km − t km max ≤ 0 t km min − t km ≤ 0 α km − α km max ≤ 0 α km min − α km ≤ 0 It should be noted that the constraints on the reactive power at each generator are not included in the problem as stated above. This is commonly done in a power system when modeling generator reactive power limits [23.2. 228].2. p. the Lagrangian can be written as a summation of several terms. 2.7 Summary of Lagrangian terms Given the problem statement for the OPF as shown in Section 2.5. These constraints will be taken care of by treating a generator bus at a Q limit as a load bus.
2 2 λ Skm S km − S km max * λ Pgih PGi − PGi max * λ PGil PGi min − PGi * λ Vih Vi − Vi max * λ Vil Vi min − Vi * inequality constraints λ tkmh t km − t km max * λ tkml t km min − t km * λ αkm max α km − α km max * λ αkm min α km min − α km * µ Pk (Pk ) µ Qk (Q k ) equality constraints µ Viset Vi − Vi set µ int (Pint − Pscheduled ) ∑ ( ai + bi PGi + ci PGi 2 ) generator cost functions generators 2 2 2 * k S km − S km max 2 * k Vi min − Vi penalty functions 2 * k Vi − Vi max ( ( ( ( ( ( ( ( ) ) ) ) ) ) ) ) ( ) ( ( ) ) Note: The terms with an asterisk (*) next to them are only included when the corresponding constraints are being enforced 2. and D. calculation of the gradient and Hessian is a very straightforward process. For completeness.8 Calculation of gradient and Hessian Given the terms of the Lagrangian. C. details on calculating the gradient and Hessian are found in Appendices B. albeit an extremely tedious one.2. 19 .
2. If not. Step 2. Update solution znew = zold . a. b. Check whether correct inequalities have been enforced. Initial guess z vector (bus voltages and angles. Calculate the gradient and Hessian of the Lagrangian.1 on page 10. problem is solved. the solution of the OPF can be achieved by using the Newton’s method algorithm. 20 . Step 8. go to Step 4. The Newton’s method algorithm is summarized in the flowchart in Figure 2. Initialize the OPF solution. Step 4. Initial guess at which inequalities are violated. all Lagrange multipliers). Step 5. Check whether ∆z < ε. Evaluate those inequalities that have to be added or removed using the information from Lagrange multipliers for hard constraints and direct evaluation for soft constraints.9 Solution of the optimal power flow Once an understanding of the calculation of the Hessian and gradient is attained. Solve the equation [H]∆z = ∇L(z). Step 7. If not go to Step 2. If so. Step 3.∆z. The application of Newton’s method to the OPF algorithm used in this thesis is summarized as follows: Step 1. Presently this ensures that at least one generator is not at a limit. generator output power. Determine viability of the OPF solution. otherwise continue. transformer tap ratios and phase shifts.2. Step 6.
The details of this system are not included in this thesis because this case is only shown to demonstrate the solution process described in the flowchart of Figure 2.0000 Constraints: 0.0004 Angles: 0.2151 OPFitr 2 MaxMis: Controls: 0. 15. 7.0067 Angles: 0.0393 Constraints: 0. several inequality constraints are found to be violated: a transmission line has exceeded its MVA capacity. 8.0005 OPFitr 10 MaxMis: Controls: 0.0000 Voltages: 0.0001 Adding Hard MVA Constraint from bus 64 to 65 Adding Soft Voltage (high) Constraint at bus 9 Adding BusQ Constraint for bus 12 Adding Soft Voltage (low) Constraint at bus 53 Bus 40 Gen at Max MW Limit OPFitr 5 MaxMis: Controls: 10.70 at 15:15:20 Angles: 3216.0000 Voltages: 0.2391 Angles: 183.1888 Angles: 0. 6.7240 Voltages: 3113.0233 OPFitr 9 MaxMis: Controls: 0.2. 9. As can be seen from lines 8 to 12.3324 Angles: 0.0003 Angles: 0.0003 Constraints: 1.3480 OPFitr 3 MaxMis: Controls: 0.5599 OPFitr 7 MaxMis: Controls: 0. 2.3324 Constraints: 0.8894 Angles: 29. 13.3 A Sample Case Illustrating OPF Algorithm Process In order to gain insight into the process which the OPF algorithm undergoes while determining the optimum solution.0000 Voltages: 571. a 21 . 3. 16. 10. 20.1469 Angles: 0.0000 Constraints: 0.0000 Voltages: 0. 18. Lines 3 to 7 then perform the inner loop of Newton’s method. 12.4475 OPFitr 1 MaxMis: Controls: 0.1868 Constraints: 0.0001 OPFitr 11 MaxMis: Controls: 0.Final Cost 51324.4009 OPFitr 6 MaxMis: Controls: 0.0000 Angles: 896. 11.0000 Voltages: 73.0000 Voltages: 0.7967 Angles: 19.1 on page 10. The output to a message file during the solution of the 118bus system is shown in the following.0626 Constraints: 0. the inequality constraints are checked to ensure that none are violated.0025 OPFitr 4 MaxMis: Controls: 0. Case OPF Case initialized Starting OPF solution at 15:15:19 OPFitr 0 MaxMis: Controls: 25. After having reached convergence here. 5.0000 Voltages: 0.0000 Angles: 0. 19.0007 Constraints: 0.0000 Voltages: 21. 4. 17.0000 Voltages: 0. Lines 1 and 2 initialize the OPF variables and assume a Lagrangian that contains no constraints.0000 This output file can help in the understanding of the Newton’s method solution process.0000 Voltages: 0.6602 Constraints: 0.0009 Constraints: 0. which continuously solves [H]∆z = ∇L(z) until convergence.0002 Constraints: 0. 14.6133 Constraints: 5. a test was run on a 118bus system.5178 OPFitr 8 MaxMis: Controls: 0.0000 Voltages: 1. 1.0001 Successful OPF .
probably the greatest advantage of the OPF is the great wealth of knowledge it yields concerning the economics of the power system. the OPF controls transformer tap ratios and phase shift angles as well. A derivation of this fact can be found in Appendix A. the Lagrange multipliers. respectively. while difficult. the inequality constraints are again checked. While the economic dispatch of a power system does control generator MW output. has many great advantages over the classical economic dispatch [11] of a power system. The OPF is capable of performing all of the control functions necessary for the power system. With these violations noted. If any security problems occur. 2. the OPF will modify its controls to fix them. µ PK and µ QK .e. The OPF also is able to monitor system security issues including line overloads and low or high voltage problems. Line 20 shows that new violations were found and the successful final cost is shown. Therefore. the Lagrangian is recalculated and the inner loop is begun again.4 Information Gained from the OPF Solution The solution of the OPF. one can show that they can be interpreted as the marginal costs associated with meeting the constraint. Besides performing these enhanced engineering functions.generator has exceeded its MVAR limit. After convergence is achieved again. Lines 13 to 19 show the inner loop process proceeding successfully again. $ can be seen as the marginal cost of real and reactive power generation at bus k in and MW hr $ MVAR hr . In studying the Lagrange multipliers associated with each constraint. These prices could then be used to determine electricity prices at bus 22 . and two buses are outside their desired ranges. remove a transmission line overload.. a generator has exceeded its MW limit. i.
the Lagrange multiplier. Consider the Lagrange multiplier associated with a hard transmission line MVA constraint. This can be interpreted as the cost savings per hour for each additional 1 MVA increase in a line’s rating. This will be discussed in Chapter 4. λ Skm . On a larger level. If this cost is positive. 23 . µ int . This information may be of use in planning where new transmission lines will have the greatest economic impact on the power system. associated with the area interchange constraint. Much more information can be gained from a single OPF solution. While these economic data are very helpful for realtime pricing algorithms. other data could be used to help with transmission system planning. These costs may be of use in determining the price which an area would charge for a megawatt transaction with another area. the area would benefit from selling electricity.k. but a great potential also lies in the simulation of a power system over time using a power system simulator with an OPF as its solution engine. can be seen as the marginal cost of allowing an area to break its interchange. while if it is negative. then the area would benefit from buying electricity.
Because of this. within their limits. λ λ α km min Pk. These include generator MW outputs. the constraint variables are variables associated with the constraints. The variables in the OPF problem are summarized in Table 3. Table 3. They are also of interest at the solution. The control variables correspond to quantities that can be arbitrarily manipulated. and α Vk and δ k km µ viset. and In addition to the OPF variables. but must be monitored. λ tkm max. and constraints. The states correspond to quantities that are set as a result of the controls. λ Vih. µ µ Skm. in order to minimize the costs. α km max. tkm. but for soft constraints. λ tkm min. HEURISTICS OF THE OPTIMAL POWER FLOW SOLUTION 3. a fourth 24 . transformer tap ratios. in order to handle the variables in the OPF problem efficiently. λ λ PGil. a primary concern is identification of variables during the process.1 Classification of OPF Variables While writing software to perform an OPF solution. µ int.1. and transformer phase shift angles.3. Qk.1 OPF problem variables Variable Classification Control State Constraint Variables in Classification PGk. states. These include all the Lagrange multipliers. it is convenient to separate them into three categories: controls. The states include all system voltages and angles. Finally. λ PGih. λ Vil. the Lagrange multiplier is monitored. For hard constraints. it is also important to keep track of any added soft constraints.
By a closer examination of the Hessian structure for the OPF problem. Much of the structure of the Hessian matrix for the OPF problem is known prior to any calculation. and then constraints. and because one of the primary objectives of an OPF is to find its solution in a short amount of time. In general. states. the Hessian matrix is extremely sparse. Fortunately. the Hessian takes the following form as discussed in Section 2. During the Newton’s method solution process.variable classification is added: penalty. No variable data are stored for this classification. 25 . The first step was to reorder the variables to include the generator megawatt controls first. This resulted in the following structure. it is seen that the lower right quadrant of the matrix has entries of zero. Because of this. the speed of the solution of this linear system of equations is essential to a successful OPF solution. It simply serves as a place holder for the penalty function.2: W H= J JT 0 Already.2 Implementation of Sparse Matrix Techniques At the heart of the solution of the OPF using Newton’s method is the solution of the linear system of equations.1. the equations can be quickly solved. for any Newton’s method routine. followed by the remaining controls. this system of equations is solved repeatedly. By implementing sparse matrix routines. as is the case with many power system matrices. 3. advantage can be taken of this sparsity. [H]∆z = ∇L(z).
and column λ PGih) . therefore. then the matrix n PG n PG T will have some offdiagonal entries. 26 . will result in the following matrix. so this is a desired event. and Wremaining and Jremaining are the remaining parts of the Hessian matrix. it will not affect the zero/nonzero structure of the resulting matrix. The diagonal entries are needed for factorization anyway. PGih .WPG H= 0 n PG 0 Wremaining J remaining J remaining T 0 n PG T where WPG is a purely diagonal matrix with entries of 2ci. then each column of nPG will have only a single entry of (1). In order to determine the fills created by this matrix. WPG Hmodified = 0 n PG 0 Wremaining J remaining −1 WPG n PG T J remaining T −1 − n PGWPG n PG T Wremaining and Jremaining are unaffected by these steps and fills are only created by the term −1 n PG WPG n PG T . and possibly an entry of (±1) in a column corresponding to a generator with a maximum or a minimum MW constraint being enforced. If no generators are at a megawatt limit. nPG is a matrix with a (1) entry in each column corresponding to the net power injection constraint. One will find that most fills created by this term are actually desired. Because WPG is a diagonal matrix. If. using Gauss’s method [24].column µ Pk). some generators are at a megawatt limit. These entries will occur in offdiagonal pairs such as (row λ (row µ Pk. however. the sparse LU factorization up to the rows of Wremaining. the matrix will result in only diagonal entries in the rows corresponding to the Lagrange multipliers µ Pk. Due to this structure. In this situation. only consider the matrix n PG n PG T . consider only its zero/non zero structure.
This results in the following structure for the Hessian matrix WPG 0 H= 0 0 n PG where I is the identity matrix. 0 Wvset I Wrvset J rvset 0 I 0 0 0 0 Wrvset T 0 Wr J rr n PG T J rvset T 0 J rr T 0 Given this ordering.Because generators will not be normally at their limits. Lagrange multipliers corresponding to generator voltages (in same order as the states) third. WPG 0 H= 0 0 n PG 0 I Wvset Wrvset J rvset 0 0 0 0 I Wrvset T 0 Wr 0 J rr n PG T 0 J rvset T J rr T 0 Given this ordering. switching the second. will result in the following matrix.and thirdrow partitions around yields the following. In order to exploit this structure. one can see that the Hessian rows corresponding to the Lagrange multipliers corresponding to the voltage set points contain only one entry. and because only two nondiagonal fills are created for each generator at a limit. using Gauss’s method [24]. 27 . a (+1). and finally the remaining states and constraints. the sparse LU factorization up to the rows of Wr. voltage states corresponding to generators second. In studying the Hessian matrix structure further. this ordering can help speed the solution. consider ordering the variables in the following manner: generator MW outputs first. This ordering is what is presently done in the OPF source code written for this thesis.
in addition.WPG 0 Hmodified = 0 0 n PG 0 I Wvset Wrvset J rvset 0 0 I 0 0 0 0 Wrvset T Wr J rr J rvset T J rr T T −1 − n PGWPG n PG −1 WPG n PG T 0 The fills created by this method would be identical to those created by the previous ordering.3 Determination of the Set of Binding Inequality Constraints When reading any paper written about the application of Newton’s method to the OPF problem.1 showing the Newton’s method solution algorithm. however. the implementation of this ability into the OPF source code has not been done.33% of the execution time assuming that the inner loop time remains relatively constant. it does require row pivoting to be done by the sparse matrix solution routines. and thus ultimately its usefulness. Therefore. However. In referring to Figure 2. the sparse matrix routines could also take advantage of simply “skipping” over the processing of the second. even removing one of these iterations could save 25 . Presently. one sees that this determination of the binding set makes up the outer loop of the flowchart. the OPF source code does ordering with generator megawatt output controls first. At present. In a typical OPF problem.and thirdrow partitions because the only operations required by Gauss’s method would be divisions by 1. 3. this outer loop will be executed about three or four times. This ordering could be implemented and would result in increased solution speed. one runs into discussions of finding the binding set of inequalities [13]. but it can be added at a later time. 28 . The reason for this emphasis is the role this process has in determining the speed of an OPF solution.
add a hard constraint or soft penalty function depending on user’s preference. Step 2. a. 29 . Maximum and minimum transformer phase shifts. Maximum and minimum transformer tap ratios e. add a hard constraint or soft penalty function depending on user’s preference. Step 4. Maximum and minimum transformer phase shifts. Using this process. a. Step 5. Step 1. Determine if any line MVA limits are violated. Hard bus voltage constraints c. Check if any inequality constraints can be removed. the following process was used to determine if the correct inequalities had been enforced. Check if any bus voltage limits are being violated. Step 3. If they are not. If they are.For the OPF source code written for this thesis. 228]. An inequality constraint can be removed if the Lagrange multiplier associated with it is negative. Maximum and minimum generator megawatt outputs d. Check all control variables to see if they are operating within their limits. Maximum and minimum generator MW outputs b. Determine if any generator reactive power limits are being violated. change the generator bus to a load bus [23. For those that are violated. add a constraint. Hard line MVA constraints b. Maximum and minimum transformer tap ratios c. the OPF will determine which inequality constraints to enforce. If they are. p.
it is often necessary to search through the list of controls. all values to the left of the node are smaller. At each node. when the system becomes large. It is therefore very useful to take advantage of a data structure suited for repeated searches.1 Sample Binary Tree 12 7 10 13 A binary tree obeys very simple rules which make searching for an individual element of the tree a very quick process. because with a linked list. In this way. and all values to the right of the node are larger. it can be shown that the maximum search time for an element of the tree is proportional to the base two logarithm of the number of elements (log2N). states. The saving in time is substantial when the number of nodes becomes very 30 . the average search time is proportional to half of the number of elements in the list (N/2). While for small systems the use of a simple linked list is adequate.3. 8 5 3 1 4 6 Figure 3. this searching can begin to dominate the CPU time spent. One such data structure is the binary tree.1. A sample binary tree is shown in Figure 3. and constraints for a particular variable.4 Searching Algorithms While performing the OPF solution.
into controls.large. Type of state with ranking: Vk and δ k. (To bus number for transformer tap or phase shift) 3. λ tkm max. Within each variable type precedence is defined as follows: Controls 1. The binary tree structure can then be applied to these structures with this ranking. as discussed in Section 3. µ λ Vih. Type of control with ranking: PGk. Type of constraint with ranking: µ Pk. precedence order is controls. The binary tree results in an average search which is 50 times faster.1. Using the binary tree data structure greatly aids in reducing the amount of time spent searching. (To bus number for line constraints) 3. States 1. This will be discussed in the next section. and then constraints. and λ α km min. 31 . and α km. λ α km max.5 Solution of an OPF Repeatedly Over Time One of the great opportunities of an OPF solution is the ability to simulate a power system over time while keeping it at its optimal solution. Consider a system which has 1024 elements: 1024/2 = 512 and log21024 = 10. λ Skm. and constraints. Bus number 2. Bus number (or area number for interchange constraints) 2. Bus number (from bus for transformer tap or phase shift) 2. int. µ Qk. λ Vil. λ Pgih. For the ranking in this thesis. Applying this thinking to OPF variables requires that they be given an ordering so that “smaller than” and “larger than” comparisons can be made. µ viset. states. λ tkm min. the variables are separated. λ PGil. 3. but heuristics for increasing the computational speed of this process will be discussed here. states. In the OPF source code. Constraints 1. tkm.
the speed of solution in the simulation environment will be greatly increased. as well as which inequalities are binding) can be fed as the initial guess for the next OPF solution. These schemes only update those elements of the L and U factors of the Hessian that are affected by changes to the Hessian. angles. Because of this. Lagrange multipliers. 32 .1. voltages. By keeping the solved OPF solution. however. the initial guess for the variables is very important. In calculating the Hessian a time step. the convergence of Newton’s method algorithm is very rapid near the solution. These modifications will greatly enhance the power system simulation over time. While this holds great promise for future work. Both of these important parameters for solving the OPF quickly can be met when simulating a power system over time. gradient. At present. It was also mentioned in the previous section that the speed of the OPF is greatly influenced by how quickly the binding set of inequalities is found.As mentioned in Section 2. these schemes have not yet been implemented into the OPF written for this thesis. the Hessian and gradient of the Lagrangian are recalculated from scratch at each time step of the simulation. the output of one OPF solution (the Hessian. By assuming that the power system will not undergo drastic change over the next time step. further advantage could be gained by implementing partial refactorization schemes for the Hessian as seen in [25]. The search for the binding inequalities is also done from scratch for each time step.
USES OF AN OPTIMAL POWER FLOW IN A POWER SYSTEM SIMULATION ENVIRONMENT While a single OPF solution yields valuable information regarding a power system. 196 MW 27 MVR 0. For details concerning this power system see Appendix E. a vast amount of economic data can be gleaned from the simulation. SingleArea System Not on OPF Control 33 .97 PU 100 MW 50 MVR 1.00 $/MWH 0.00 PU 0.00 PU 91 MW 60 MVR Bus 5 0.00 PU 0.00 $/MWH 0.00 $/MWH 0. 4.00 $/MVRH 20 MVR Bus 2 1.00 PU 100 MW 20 MVR Bus 4 0. In this environment.00 $/MVRH 237 MW 23 MVR 0.1 Sixbus.1.00 $/MVRH 100 MW 10 MVR 100 MW 20 MVR 100 MW 20 MVR Bus 1 1.00 $/MWH 0. POWERWORLD [17].00 $/MWH 100 MW 0.00 $/MVRH 147% 0.00 $/MVRH 86 MW 92 MVR Figure 4. the implementation of the OPF into a power system simulation environment holds even greater promise. In this way. This chapter will give several examples of the use of the OPF code as implemented into the power system simulator. simulation of a system over time can be done while maintaining it at its optimal condition.00 $/MVRH Bus 3 1.00 $/MWH 0.1 Example Line Overload Removal A simple power system not operating under the OPF control is shown in Figure 4.4.97 PU Bus 6 0.
02 $/MWH 0. 100 MW 20 MVR 100 MW 20 MVR Bus 1 1.1 is placed on OPF control with its line limit raised.93 $/MWH 0. 4.00 PU 100 MW 20 MVR 116 MW 76 MVR Figure 4. the generators have redispatched themselves in order to remove the line overload. 34 .00 $/MVRH 14. Note that the OPF results in the same dispatch seen in Figure 4.97 PU Bus 6 15.1 now that the line limit has been removed. the same system shown in Figure 4.00 PU 190 MW 25 MVR 13.2 Use for Bus Real and Reactive Power Pricing To illustrate the real and reactive power pricing potentials of the OPF solution.1.03 $/MWH 0.00 $/MVRH 20 MVR Bus 2 1.2. the OPF control is turned on.65 $/MVRH Bus 3 1.00 $/MVRH 100% 17.25 $/MWH 100 MW 0.23 $/MWH 0. Further analysis of this process from an economic viewpoint will be discussed in the next section. SingleArea System on OPF Control As can be seen.12 $/MVRH 100 MW 10 MVR 16.77 $/MWH 0.2 Sixbus.97 PU 100 MW 50 MVR 1.00 $/MVRH 179 MW 48 MVR 0.In order to remove the line constraint. See Figure 4.00 PU Bus 4 12. and the line overload is removed as shown in Figure 4.00 PU 122 MW 43 MVR Bus 5 0.
the bus MW marginal costs at buses 1.97 PU Bus 6 15.00 PU 91 MW 60 MVR Bus 5 0. and 5 increased.19 $/MWH 0.97 PU 100 MW 50 MVR 1. 3.11 $/MVRH 100 MW 10 MVR 15.2 and Figure 4.2 and Figure 4.1.100 MW 20 MVR 100 MW 20 MVR Bus 1 1. their bus MW marginal costs also decreased.18 $/MWH 0.00 PU Bus 4 13.00 PU 196 MW 27 MVR 13.1 Line Limit from Bus 5 to 4 Raised to 100 MVA Comparison of the OPF solutions in Figure 4.00 $/MVRH 20 MVR Bus 2 1. With the line limited as in Figure 4. to $7895/hr. the largest changes occurred at the ends of the limited line. as would be expected. As might be expected.37 $/MWH 100 MW 0.00 $/MVRH 237 MW 23 MVR 0. buses 4 and 5.85 $/MWH 0.1 is $7824/hr. Also note the differences between the bus MW marginal costs in Figure 4.85 $/MWH 0.23 $/MVRH Bus 3 1. 35 .00 $/MVRH 14.1 yield valuable insight.00 $/MVRH 14.25 $/MWH 0.00 PU 100 MW 20 MVR 86 MW 92 MVR Figure 4.2 this costs increases. Conversely. Because the generators at buses 2 and 4 were forced to decrease their output in order to remove the overload. The total system cost without the line limited as in Figure 4.
00 $/MVRH 169 MW 50 MVR 0. Using the capabilities of the POWERWORLD Area Transactions/Information display [17].00 $/MVRH 20 MVR Bus 2 1.97 PU 100 MW 50 MVR 1.75 $/MWH 0.4. In Figure 4. therefore.86 $/MWH 0.15 $/MWH 100 MW 0.71 $/MWH 0.00 PU 161 MW 23 MVR Bus 5 0. Table 4.00 PU 100 MW 20 MVR Area One 143 MW 62 MVR Figure 4.00 PU 133 MW 2 MVR Area Two 12. the simple sixbus system from before is split into two areas as shown.00 $/MVRH 16.1. 36 .39 $/MWH 0.98 PU Bus 6 13.58 $/MWH 0. TwoArea System on OPF Control As can be seen. for this case the generation in Area Two is less expensive than in Area One. 100 MW 20 MVR 100 MW 20 MVR Bus 1 1.3 Use for Area Real Power Pricing The OPF solution method may also be used with multiarea power systems.1 summarizes several possibilities. scheduled transactions can be set up between the two areas to optimize their costs. The OPF will enforce the scheduled area interchange in these systems.09 $/MVRH 100 MW 10 MVR 16.24 $/MVRH Bus 3 1. it would be advantageous for both areas if Area One were to purchase some power from Area Two.1 Sixbus.00 $/MVRH 15.00 PU Bus 4 12.
2 one can see that they show the same solution. This must happen so that all areas are at their optimal solution.Table 4.00 PU 100 MW 20 MVR Area One 116 MW 76 MVR Figure 4.2.2 and Figure 4.97 PU Bus 6 15.00 $/MVRH 100% 17.25 $/MWH 100 MW 0.5 70.2 Transaction of 65.00 $/MVRH 179 MW 48 MVR 0.97 PU 100 MW 50 MVR 1.00 $/MVRH 20 MVR Bus 2 1.0 AreaOne Cost [$/hr] 4564 4489 4482 4481 4481 AreaTwo Cost [$/hr] 3496 3423 3413 3415 3428 Sum of Both Areas [$/hr] 8060 7912 7895 7896 7909 As can be seen.77 $/MWH 0.5 MW Undertaken Comparing the OPF solution in Figure 4.00 PU 122 MW 43 MVR Bus 5 0.93 $/MWH 0.23 $/MWH 0.00 $/MVRH 14.65 $/MVRH Bus 3 1.00 PU Bus 4 12.00 PU 13.03 $/MWH 0. This is of course not unexpected. This transaction scenario is shown in Figure 4. the least expensive scenario for the sum of the areas is when an interchange of 65.02 $/MWH 0.1 Sample transaction scenarios Transaction [MW] None 50. 190 MW 25 MVR Area Two 100 MW 20 MVR 100 MW 20 MVR Bus 1 1.0 65.12 $/MVRH 100 MW 10 MVR 16. 37 .0 80. because the twoarea economic negotiations should yield the same solution that an OPF solution disregarding areas would.5 MW is undertaken.
Three of these transformers are load tap changing (LTC) transformers: buses 5 to 25.000 tap 1.44 $/MWH 1.98 PU Bus 25 Bus 4 13.27 $/MWH 300 MW 350 MW 1.16 $/MWH 0.000 tap 1.22 $/MWH 1.000 tap 75 MW 13. 10 generators.1 is on OPF control. For further details.u.51 $/MWH Bus 104 16. 204 to 224. and the tap ratios have been manipulated so as to minimize the objective function. This system has 23 buses.48 $/MWH 97 MVR 1.01 PU 1.57 $/MWH 13.03 PU Bus 204 657 MW 200 MW 300 MW 1. and 102 to 122.68 $/MWH Bus 2 0. and 5 transformers.49 $/MWH Bus 7 0.4 Example Transformer Tap Control In order to test the transformer tap control.25 $/MWH 14.1 TwentyThree Bus System on OPF Control with Tap Control Off 38 . The system as shown in Figure 4.99 PU 13. The system as in Figure 4.90 $/MWH 250 MW 432 MW 1.58 $/MWH 225 MW Bus 8 0.01 PU 13.99 PU Bus 103 13.03 PU Bus 203 15.03 PU Bus 23 1.97 PU Bus 201 Bus 202 300 MW 0.55 $/MWH Bus 122 1.02 PU Bus 105 543 MW 98 MVR 16.98 PU 450 MW 97 MVR 1.52 $/MWH Bus 3 1.00 PU Bus 6 150 MW 12. see Appendix F.50 $/MWH 200 MW 430 MW 106 MVR 340 MW Bus 5 13. 125 MW 700 MW 106 MVR 1.03 PU 16.02 PU Bus 9 400 MW 12.53 $/MWH 350 MW 1.02 PU Bus 1 Bus 101 464 MW 12. a more complex power system was used as in Figure 4.4.03 PU Bus 102 763 MW 17.64 $/MWH 0. but the ability to control the tap ratios is turned off and all tap ratios are set at 1 p.01 PU 13. 23 transmission lines.99 PU 100 MW 13.2 has activated the tap ratio control functions of the OPF.1.96 PU 15.01 PU 13.03 PU 200 MW 0.39 $/MWH 1.61 $/MWH 16.01 PU Bus 224 13.51 $/MWH Bus 28 1.86 $/MWH 106 MVR 1.49 $/MWH 13.86 $/MWH 175 MW Figure 4.98 PU 200 MW 359 MW 0.
tap ratio control will not have a huge effect.00 PU Bus 6 150 MW 12.54 $/MWH 350 MW 1.24 $/MWH 14.98 PU 450 MW 97 MVR 1. $ 50333.58 $/MWH 13.01 PU 1.05 PU Bus 23 1.44 $/MWH 1.03 PU Bus 204 657 MW 200 MW 300 MW 1.86 $/MWH 106 MVR 1. In reality.97 PU Bus 201 Bus 202 300 MW 0.125 MW 700 MW 106 MVR 1.03 PU Bus 203 15.99 PU Bus 103 13. system losses will only be reduced a small amount.25 per hour With tap ratio control ON 1.u.52 $/MWH Bus 28 1.98 PU Bus 25 Bus 4 13. therefore.62 $/MWH 16. $ 50330.03 PU Bus 102 761 MW 17. It should be noted.01 PU 13.99 PU 13.50 $/MWH 200 MW 430 MW 106 MVR 340 MW Bus 5 13.58 $/MWH 225 MW Bus 8 0.000 p.69 $/MWH Bus 2 0.965 tap 1.50 $/MWH Bus 7 0.u. 39 .2 TwentyThree Bus System with Tap Control On A summary of the tap ratios and resulting total system costs are shown in Table 4. tap ratio control allows control of the reactive power flow thus reducing losses. however.49 $/MWH 97 MVR 0.965 p.1 Summary of tap ratio control experiment Tap ratio from bus 5 to 25 Tap ratio from bus 102 to 122 Tap ratio from bus 204 to 224 Total System Cost With tap ratio control OFF 1.22 $/MWH 1. 1.u.01 PU 13.51 $/MWH Bus 3 1.000 p.86 $/MWH 175 MW Figure 4.02 PU Bus 1 Bus 101 463 MW 12.1.03 PU 200 MW 0.01 PU 13.28 $/MWH 300 MW 350 MW 1.91 $/MWH 250 MW 433 MW 1.u.17 $/MWH 0. 0. 1.63 $/MWH 0. Even given perfect reactive power control.03 PU 16.02 PU Bus 105 542 MW 98 MVR 16.010 tap 1.55 per hour As expected.50 $/MWH 13.02 PU Bus 9 401 MW 12.012 p. that the control of the tap ratios will normally not drastically reduce the system costs.010 p.000 p.98 PU 200 MW 360 MW 0.01 PU Bus 224 13.39 $/MWH 1.012 tap 75 MW 13.96 PU 15.51 $/MWH Bus 104 16. 1.54 $/MWH Bus 122 1.98 PU 100 MW 13.u. the ability to control tap ratios has lowered the total system cost. Table 4.u.
1 (data in Appendix G).00 PU 18.04 PU 50 MW 10 MVR 16.1 SevenBus.19 $/MWH 0.00 $/MVRH 1.07 $/MWH 0.00 $/MVRH 18.0000 tap Bus 4 1. a tap changing transformer can help lower generation costs both by reducing system losses and by allowing less expensive generation to operate when it could not have without tap changing control. it is also able to reduce system losses by directing the reactive power in a manner which reduces system currents.01 PU Bus 3 1. consider the sevenbus. onearea system shown in Figure 4. controlling reactive power can also help with real power control.51 $/MWH 0.02 $/MVRH Bus 6 1. it also is better for analyzing the potential abilities of tap changing transformers without the complication of the discrete jumps which would normally occur. Because reactive power and real power are not completely decoupled. a tap changing transformer is a reactive power controller. OneArea System with Taps Inactive 40 . Because it controls reactive power. In this thesis.4.01 PU 50 MW 10 MVR 19.99 PU Bus 2 1.28 $/MVRH 18. In this way.00 PU 150 MW 60 MVR 1.50 $/MWH 0.5 System MVAR Control Using Transformer Taps [26] As previously mentioned.32 $/MVRH 1. It is this ability which allows it to help control system voltages.75 $/MWH 0.11 $/MWH 0.04 PU Bus 7 50 MVR 100 MW 30 MVR 100 MW 30 MVR 136 MW 81 MVR Figure 4.18 $/MWH 0. While modelling the tap as continuous is much easier to implement. In order to demonstrate a transformer’s ability to control the system. 431 MW 59 MVR Bus 1 50 MW 10 MVR 0.36 $/MVRH 50 MW 10 MVR 20.21 $/MVRH 18.0000 tap Bus 5 1. the OPF has been configured to allow continuous tap values although in reality one encounters discrete tap values (often 33 steps).
Now consider the system with the tap changing abilities activated. This system has a total system cost of $8.23 $/MWH 0.90 $/MWH 0.80 $/MWH 0. and a comparison of the two systems is shown in Table 4.2.00 $/MVRH 1.21 $/MVRH 150 MW 60 MVR 1.059/hr $8.1.9497 tap Bus 5 1.22 MW.99 PU Bus 3 0. Comparing the data in this table.u.23 $/MWH 0.73 MW Tap at bus 24 1.2.u.61 $/MWH 0.u.9447 tap Bus 4 1.02 PU 18. OneArea.05 PU 50 MW 10 MVR 52 MVR 100 MW 30 MVR 100 MW 30 MVR 127 MW 32 MVR Figure 4.98 PU Bus 2 0.00 $/MVRH 18.032/hr.9497 p. the system is able to 439 MW 105 MVR Bus 1 50 MW 10 MVR 0. 0. Tap at bus 35 1.21 $/MVRH 0. System Losses 16.u.08 $/MVRH Bus 6 1. The OPF solution is shown in Figure 4. 41 .000 p.25 $/MVRH 50 MW 10 MVR 19. SevenBus.9447 p. System with Taps Active Table 4.059/hr and system losses of 16.04 PU 50 MW 10 MVR 16.22 MW 15.04 PU 18.93 $/MWH 0. it can be seen that by utilizing the variable tap settings. 0.04 PU Bus 7 18.The tap changing ability of the transformers in this system has been deactivated and the OPF solution is shown.1 Comparison of two systems Without Taps With Taps Bus 1 Generation 431 MW 439 MW 59 MVAR 105 MVAR Bus 7 Generation 136 MW 127 MW 81 MVAR 32 MVAR System Costs $8.60 $/MWH 0.25 $/MVRH 18.000 p.
04 PU Bus 7 50 MW 10 MVR Area Two 49 MVR 100 MW 30 MVR 100 MW 30 MVR 314 MW 4 MVR Figure 4.99 PU 24. The generation in area one is much less expensive than generation in area two.3.01 PU Bus 3 1.42 $/MWH 0.3. However.2 and Table 4. all voltages are still within 5% of nominal.04 $/MVRH 28.20 $/MWH 0.increase the cheaper generation at bus 1 and ship it towards the bottom of the system while decreasing system losses.39 $/MWH 0.3. Summaries of these results are shown in Table 4.22 $/MWH 0.56 $/MWH 0.00 $/MVRH 14.0000 tap Bus 4 0.14 $/MVRH Bus 6 0.00 PU Bus 2 1.00 $/MVRH 1.04 PU 50 MW 10 MVR 13. TwoArea System with TapChanging Transformers 42 .11 $/MVRH 14. This has been done at the expense of greater variability of system voltages.02 $/MVRH 1. SevenBus. Therefore it would be beneficial to both parties to undergo a megawatt transaction. While this example displays the use of the tap changing transformer as a reactive power controller. greater insight can be gained by considering the same system divided into two areas as shown in Figure 4.10 $/MVRH 28. Area One 50 MW 10 MVR 1.10 $/MWH 0.99 PU 253 MW 145 MVR Bus 1 150 MW 60 MVR 1.09 $/MWH 0.00 PU 50 MW 10 MVR 28. Various transactions were undertaken both with the TCUL ability active and with it inactive.0000 tap Bus 5 1.
582 5.] 253 104 314 36 1.032 8.u.2 Results for various interchanges with transformer taps left inactive Interchange Area 1.0000 2.491 Total Cost [$/hr] 9.543 5.79 16.73 27.84 +31 at 17.99 18.91 24.12 +19 at 17.0000 432 58 134 82 1.9448 467 105 102 38 0.0000 526 34 58 139 1.245 Losses Area 1 [MW] 2.0000 411 66 153 71 1.584 0.14 13.Table 4.80 11.45 4.0000 2.056 8.0000 2.84 +50 at 18.557 0.338 Losses Area 1 [MW] 3.0000 2.779 5.923 1.39 7.84 11.56 7.83 +25 at 17.] [$/hr] 1.2 [MW at $/MWhr] none +50 at 18.754 Cost at Area 2 [$/hr] 6.0000 Tap Cost at Cost at Bus 35 Area 1 Area 2 [p.667 1.9578 439 105 127 32 0.542 0.0466 2.664 5.0432 304 98 258 34 1.9411 2.072 8.12 +25 at 17.84 16.529 1.938 1.3 Results for various interchanges with transformer taps active Interchange Area 1.94 2.0000 304 114 258 17 1.46 +50 at 18.921 5.0000 468 48 104 103 1.716 0.15 13.] 253 145 314 4 1.] [$/hr] [$/hr] 1.949 6.9265 Tap Cost at Bus 35 Area 1 [p.44 3.500 1.0000 357 88 204 42 1.85 4.90 2.055 8.11 2.u.07 4.603 5.98 16.308 1.84 14.01 6.0137 357 97 204 32 0.u.718 5.9340 524 99 54 60 0.90 3.70 1. Gen Bus 24 [MW] [MVAR] [MW] [MVAR] [p.88 12.83 +50 at 17.51 33.46 17.u.86 11.84 +50 at 17.593 0.92 .475 5.04 10.45 +50 at 18.60 25.249 8.462 5.248 8.0000 2.0209 2.9499 2.237 8.96 21.2 [MW at $/MWhr] none +50 at 18.9933 2.0000 2.16 17.83 Losses Total Area 2 Losses [MW] [MW] 14.61 7.641 8.059 8.555 5.60 4.103 8.504 1.9851 411 101 152 32 0.9645 2.86 15.29 1.19 Losses Total Area 2 Losses [MW] [MW] 13.19 6.34 Table 4.37 Bus 1 Bus 1 Bus 7 Bus 1 Tap Gen Gen Gen.257 8.9280 2.637 8.559 Total Cost [$/hr] 9.514 5.83 +50 at 18.0000 2.27 13.14 10. Gen Bus 24 [MW] [MVAR] [MW] [MVAR] [p.299 5.31 Bus 1 Bus 1 Bus 7 Bus 1 Tap Gen Gen Gen.
the systems are at the same state found by running the OPF with no area constraints enforced. 4. From a total system cost point of view. and as a result. the generators’ reactive power injections are able to remain relatively constant. onearea power system of Appendix E again. the two areas should stop undergoing transactions after having reached the transactions shown in bold type (minimum total cost). the only elements of the system that have control over the reactive power flows are the two generators. and a transaction of 175 MW for the system with taps active. consider again the sixbus.3. As a simple example.A comparison of Table 4. Note that after undergoing these transactions.3 shows reduced costs at every step for the system with the tap changing transformer active.2 and Table 4. the variation in the injected reactive power from the generators is extremely large. With the taps active as in Table 4.2. This is after a transaction of 169 MW for the system with taps inactive. the transformers take over as the reactive power controllers. then this reactive power control by the tap changing transformers would free up the generator for more megawatt production. It should be noted that if a generator were to be treated as an MVA limited machine. Assume that the loads of the system vary as shown in 44 . The information in these tables illustrates the concept that the tap changing transformer is a reactive power controller. as opposed to an MW and MVAR (separately) limited machine. With taps inactive as in Table 4. As a result.6 Transmission Line Valuation by TimeDomain Simulation One of the most intriguing potential uses of an OPF is its use as a pricing tool by doing timedomain simulations of a power system.
00 $2.1.09% 0.1 Load Factor Curve for sixbus System Entering this load factor curve into the POWERWORLD simulation software and simulating the power system for a full 24 hours yield a total cost in dollars for operating the power system over that period of time. Table 4.2 0 0:00 Time 0:00 2:30 5:30 8:00 10:30 4:00 8:00 12:00 Time 16:00 20:00 0:00 Load Factor 0.8 Time 13:15 16:30 19:15 22:00 23:59 Load Factor 1.Figure 4.00 Percent Increase from No Change Configuration N/A 0.4 0. one can gain useful economic insights from the comparison of total system costs. The POWERWORLD simulation software utilizes trapezoidal rule integration to calculate the total system cost from the incremental costs at each time point. Load Factor vs.2 1 0. the sixbus system was simulated with various transmission lines removed.1 1. Note that the case was simulated at 600 times real time for the first 23 hours and at 60 times real time for the last hour.00 $392.40% 45 .2 1. By resimulating the system repeatedly with variations in system structure. The load at a given bus is then equal to its base value shown in Appendix D multiplied by the load factor.8 0.6 0.4 0.1 0.6 0. The results are summarized in Table 4.069.4 1.72% 1.071.1.377 $ 148.6 0.309 Difference from No Change Configuration N/A $137.1 Simulation data for sixbus system with transmission lines removed System Configuration No Changes Line from Bus 3 to Bus 5 Removed Line from Bus 3 to Bus 4 Removed Line from Bus 1 to Bus 2 Removed Line from Bus 4 to Bus 5 Removed Total Cost for 24 Hour Simulation $ 148.6 Load Factor Figure 4.8 0.26% 0. Time 1.632 $ 149.240 $ 148.311 $ 150.4 0. Using this technique.00 $1.
As an example. or even possibly in helping to determine where to build new transmission lines in the future. 4. however. consider capacitor bank valuation for the twentythree bus.2 1 0.100 Time 13:15 16:30 19:15 22:00 23:59 Load Factor 1. Regardless of this. It is very important to note. Load Factor vs Time 1.100 1.8 0.800 0.000 Load Factor Figure 4. one can find the approximate cost a company would incur per day while taking a line out for maintenance. In other words. that this does in no way take into account the possibility that another line may unexpectedly go out.4 0.1. onearea power system of Appendix E.950 1.350 1.800 0.1 Load Factor Curve for TwentyThree Bus System 46 .2 0 0:00 Time 0:00 2:30 5:30 8:30 10:30 4:00 8:00 12:00 Tim e 16:00 20:00 0:00 Load Factor 1.By comparing the total system cost with all lines connected to the cost with a line removed. the information is very valuable and could be used to help with maintenance scheduling.7 Capacitor Bank Valuation by TimeDomain Simulation The technique discussed in Section 4. in which case all lines may be needed to take up the slack.6 0.000 0. Assume that the loads of the system vary as shown in Figure 4.275 1.4 1. the pricing of redundancies in the transmission system is not considered in this methodology.275 1.5 can also be applied to the valuation of any transmission system component.
The results are summarized in Table 4.358. Table 4. the twentythree bus system was simulated with various capacitor banks removed.00 $1.609 $ 1.222 $ 1.186. The threebus system of Figure 4.1.423 $ 1.Again. Following the technique as shown before.00 Percent Increase from No Change Configuration N/A 0. entering this load factor curve into the POWERWORLD simulation software and simulating the power system for a full 24 hours yield a total cost in dollars for operating the power system over that period of time.1 Simulation data for sixbus system with transmission lines removed System Configuration No Changes Capacitor Bank at bus 103 Removed Capacitor Bank at bus 2 Removed Capacitor Bank at bus 104 Removed Total Cost for 24 Hour Simulation $ 1.359. from this information.359.626 Difference from No Change Configuration N/A $1. This simulation was run at 60 times real time for the entire simulation. the approximate cost a company would incur per day while removing the capacitor bank can be inferred from these calculations. and the line from bus 2 to bus 3. Area one consists only of the generator and load at bus 1 plus two tie lines.359. 47 .0886% Again.0873% 0. The OPF is able to take into account system security concerns including voltage limits and transmission line limits while calculating the ATC. Area two consists of the generator and load at bus 3 plus the load at bus 2.00 $799.8 Limit on Available Transfer Capability (ATC) Due to a Voltage Constraint [26] Another use for the OPF algorithm is for the calculation of the available transfer capability for a system.203.1 (data in Appendix H) can be used to illustrate and examine the value of reactive power support at a system load bus.0588% 0. 4. Area one has expensive generation.
Similarly. 20. the loads at all buses remain fixed at the values shown. Consider the situation where area one wants to purchase power at considerable savings from area two. This base case has zero scheduled megawatt transfer between the two areas. but 73 MW of area two’s generation flows through area one.1 ThreeBus Base Case with No Area Power Transfer Note that the incremental cost for reactive power at a generator bus is $0. For this study. This is true as long as no MVAR limits are reached because no cost is associated with generation of MVARs by a generator in the OPF formulation of this thesis.799/hr.00.1 has each area providing its own load plus a portion of the system losses.94 $/MVRH 0.04 PU 238 MW 55 MVR 100 MW 30 MVR Area Two Area One 72 MW 21 MVR Bus 1 1.18 $/MWH 0.22 $/MWH 0.while area two has cheaper generation. We first examine the available 48 .85 $/MWH 0.9710 PU Bus 2 72 MVR 50. For the indicated loads and no area transfers the cost of operation for the total system is $25.04 PU 75 MW 227 MW 22 MVR 73 MW 64 MVR 22. The base case shown in Figure 4. The unfilled pie charts on each line indicate that the lines are loaded at less than their MVA ratings.00 $/MVRH 73 MW 15 MVR 411 MW 70 MVR Bus 3 1.00 $/MVRH 504 MW 193 MVR 500 MW 100 MVR 300 MW 14 MVR 100 MVR Figure 4. area two wants to sell power to area one.
9600 PU Bus 2 112 MVR 38. Note that the incremental cost for MVARs is almost the same as for MWs at bus 2.2 for the case in which a nominal 15 MVAR capacitor bank is installed at bus 2 for reactive power support.04 p. the OPF solution will stop at that constraint. The voltage constraint for power quality used in this example is 0. The available transfer capability for these constraints is 244 MW. the OPF solution is given in Figure 4. When the transfer is such that the voltage reaches its lower limit (0. Both generators have their excitations set to give 1.04 PU 354 MW 55 MVR 100 MW 30 MVR Area Two Area One 201 MW 78 MVR Bus 1 1.36 $/MWH 0.00 $/MVRH 271 MW 290 MVR 500 MW 100 MVR 300 MW 14 MVR 100 MVR capability for these constraints is increased to Figure 4. The cost of operation for the total system is $21.458/hr. Thus this maximum power transfer has resulted in a reduction of total cost by $4.u. An OPF solution with bus voltage constraints is used to find the maximum power which can be transferred from area two to area one.341/hr.09 $/MWH 54.3.96 ≤ Vi ≤ 1.04 PU 28 MW 330 MW 16 MVR 30 MW 102 MVR 55. When the capacitor bank at bus 2 is increased from a nominal 15 MVAR rating to a nominal 30 MVAR rating. A global OPF solution which minimizes the total cost of providing the total load will attempt to increase areatwo generation and decrease areaone generation.00 $/MVRH 214 MW 26 MVR 668 MW 59 MVR Bus 3 1.2 ThreeBus Example at Maximum Power Transfer (15 MVAR Capacitor Support) 49 .55 $/MWH 0.96).40 $/MVRH 0. Additional transfers are not possible without violation of the voltage constraint at bus 2. voltage for serving their local area load within the voltage constraints. The available transfer 25. This power transfer will change the voltage at bus 2. The solution and associated marginal costs for both real and reactive power at each bus are shown in Figure 4.transfer capability of this system for transfer from area two to area one.04.
Therefore.849/hr.04 PU 278 MW 24 MVR dispatch. The cost of operation for the total system is $20. The cost of operation for the total system is $20. because there is a greater change in system megawatt 50 .3 ThreeBus Example at Maximum incremental cost for MVARs has reduced as Power Transfer 786 MWMVAR Capacitor (30 27.96).890/hr.963 > 0.04 PU 60 MW resulted in an additional reduction of total cost by $451/hr. the OPF solution is given in Figure 4. additional transfers are not possible without violation of the voltage constraint at bus 2.00 $/MWH 13.9631 PU Bus 2 122 MVR 33. Again. Thus the increase in available power transfer capability has 27.9600 PU Bus 2 122 MVR 34. Thus the increase in available power transfer capability has resulted in a further reduction of total cost by $41/hr.08 $/MWH 0.52 $/MWH 0.82 $/MWH 1. Note that the 364 MW 36 MVR 64 MW 109 MVR 38.37 $/MVRH 0.05 $/MVRH 0.325 MW. The savings between 15 and 30 MVAR capacitors was $451 while the savings between 30 and 45 MVAR capacitors was only $41.4.04 PU 393 MW 51 MVR 100 MW 30 MVR Area Two Area One 242 MW 101 MVR Bus 1 1. When the capacitor bank at bus 2 is increased from a nominal 30 MVAR rating to a nominal 45 MVAR rating.00 $/MVRH 261 MW 25 MVR 754 MW 56 MVR Bus 3 1.00 $/MVRH 198 MW 323 MVR 500 MW 100 MVR 300 MW 28 MVR 100 MVR Figure 4.04 PU 73 MW 377 MW 49 MVR 77 MW 107 MVR 32.00 $/MVRH 170 MW 332 MVR 500 MW 100 MVR 300 MW 42 MVR 100 MVR ThreeBus Example at Maximum capability for these constraints is increased to Figure 4.4 Power Transfer (45 MVAR Capacitor Support) 355 MW. no additional power transfers can lower the total system costs. The available transfer 408 MW 44 MVR 100 MW 30 MVR Area Two Area One 257 MW 110 MVR Bus 1 1.89 $/MWH 0. Note that the OPF has reached an optimal dispatch schedule without hitting the bus 2 voltage constraint (0.00 $/MVRH Bus 3 the OPF approaches a more economical 1.72 $/MWH 51 MVR Support) 0.
however. The SRMC of a power transaction can be estimated by solving the OPF for both the system with the transaction and the system without the transaction in place. These two topics are extensively discussed in both [27] and [28]. This cost would be charged to the two companies undergoing the transaction and would be paid to the operator of the transmission system. After these solutions are obtained. Two methods proposed for the determination of this cost is through shortrun marginal costing (SRMC) and longrun marginal costing (LRMC). the SRMC can be found as follows. The capital costs include the costs of future expansion to the transmission system as well as the generation capacity. albeit simple. 51 . explanation of these follows. A brief. The operating costs include fuel costs. 4. as well as system constraints such as transmission line limits. Define BMCi Pi. the SRMC technique can easily be implemented as discussed in [28]. losses.power generation at bus i before transaction + power generation at bus i after transaction) From these definitions a definition for SRMC can then be made [28].transaction = (marginal cost of power at bus i after the transaction) = the net power injection at bus i due to the transaction = (. includes both the operating costs and the capital costs of the power system. it may be necessary to determine the cost incurred to the transmission system due to a power transaction between two companies.9 Transmission System Pricing Through ShortRun Marginal Costing (SRMC) In the restructured environment of the future. This is also reflected in the incremental costs of MVAR at bus 2 for the various cases.dispatch when moving from 15 to 30 MVAR than from 30 to 45 MVAR. Through the use of the OPF. The LRMC. The SRMC takes into account only the operating costs of the power system.
27 755.86 122 14. but with the line between buses 4 and 5 doubled to 100 MVA from 50 MVA. the SRMC for this transaction is (.70 132. summarized in Table 4.13 Thus. 52 .71 116 16.77 10 SRMCtransaction = ∑ BMC i * Pi .25 432. As mentioned in [27]. twoarea system Bus Nu m 1 2 3 4 Generation BMC Generation BMC After Pi.5 MW transaction.13)/hour.transaction As a simple example consider the sixbus.3.93 39 133 12. twoarea system from Section 4.trans Before Before After Transaction Transaction Transaction Transaction [MW] [MW] [$/MWhr] [MW] [$/MWhr] 161 15.25 57 143 16.58 179 12. Figure 4.1.03 27 169 12.transaction = all buses in transaction This calculation is BMCi *Pi.5 MW undertaken between the two areas. Taking the data from these two figures. consider the same system undergoing the same 65. and Figure 4.81 127.1 on page 36 shows this system with no transactions. For comparison. The simulation of this system is shown in Figure 4.2 on page 37 shows the system with a transaction of 65. SRMC can be a negative value.trans [$/hr] 582.1 SRMC calculation for sixbus. Table 4.SRMCtransaction = all buses in transaction ∑ BMC i * Pi .$132.1.15 190 13. a calculation of the SRMC for the transaction can be made.
1 shows this system with no transactions.00 $/MWH 0. TwoArea System Undergoing Transaction with Line Limit Doubled Again.05 146.00 PU 125 MW 41 MVR Bus 5 0. twoarea system where no limit is encountered Bus Num 1 2 3 4 Generation BMC Generation Before Before Aafter BMC After Pi.15 165 12. Taking the data from these two figures.trans [$/hr] 540.71 113 15.transaction = all buses in transaction BMCi *Pi.86 125 15.23 $/MWH 0.2. Figure 4.50 463.97 PU Bus 6 14.trans Transaction Transaction Transaction Transaction [MW] [MW] [$/MWhr] [MW] [$/MWhr] 161 15.81 53 . and Figure 4.2 shows the system with a transaction of 65.95 30 169 12.1 Sixbus.2 SRMC calculation for sixbus.00 PU 50 MVR Area One 100 MW 20 MVR 113 MW 77 MVR Figure 4.23 35 SRMCtransaction = ∑ BMC i * Pi .5 MW undertaken between the two areas.00 $/MVRH 15.00 $/MVRH 15. This calculation is summarized in Table 4.77 32 143 16.00 408.00 $/MVRH 204 MW 35 MVR 0.23 $/MVRH Bus 3 100 MW 1. a calculation of the SRMC for the transaction can be made for the system where no line limit will be encountered.00 PU 165 MW 13 MVR Area Two 12.58 204 13.64 478.100 MW 20 MVR 100 MW 20 MVR Bus 1 1.30 $/MWH 0.10 $/MVRH 100 MW 10 MVR 15.77 $/MWH 100 MW 0.76 $/MWH 0.00 $/MVRH 20 MVR Bus 2 1.00 PU Bus 4 13. Table 4.00 36 133 12.95 $/MWH 0.97 PU 1.
This is not unexpected. 54 .e.81)/hour. the SRMC for this transaction is (.$146. therefore. it does display the basic use of the OPF to calculate the SRMC. The previous system encountered a transmission line constraint. the SRMC should be less. while this one has not.Thus. i. While this discussion is not entirely complete. More study will be done on this topic in the future. the incentive to do the transaction should be large..
real and reactive power pricing. Another problem which should be addressed is the modeling of discrete variables within the OPF framework. Newton’s method has proven to be very adept at solving the OPF problem. the use of the OPF in a power system simulation environment will be truly useful. 55 . These improvements to the software will allow the simulation of systems with thousands of buses. With this ability. Minimization of system costs. CONCLUSION The OFP written for this thesis has been very successful in achieving the goals set forth for an OPF. while maintaining system security. This will allow larger systems to be simulated. Many applications of the OPF have also been shown as in Chapter 4. Improvements will need to be made to this in order for the software to be truly useful. The author also plans to implement the partial refactorization schemes for the Hessian as seen in [25]. At present.5. was accomplished through the implementation of Newton’s method to the OPF problem. This can and will be done with relatively minor modifications. transmission system pricing. At present. The marginal cost data from the OPF were shown to aid in the available transfer capability (ATC) calculation. the OPF works very well for small systems (less than 200 buses). and transmission system component valuation. variables such as transformer tap ratios and phase shift angles are treated as continuous variables. The OPF performs generator control and transmission system control while taking into account system limits.5. the speed of the OPF solution will be increased. As discussed in Section 3. by keeping the previous OPF solution from time step to time step.
Also. the pricing schemes discussed do not take into account the need for future expansion of the power system. they do not take into account the possibility of random outages of transmission line components or generators. 56 .As a final note for this thesis. the author would like to reiterate that while the applications of the OPF as discussed in this thesis are extremely valuable.
m (equality constraints) Therefore. . ECONOMIC INTERPRETATION OF THE LAGRANGE MULTIPLIERS The Lagrange multipliers found in the optimal solution can be given an economic interpretation. 2 . the optimal solution (x*. The nonlinear programming problem that is being solved can be expressed in the following form (excluding inequality constraints). Evaluating this derivative yields ∇ u L(x . the Lagrange function may be written: L(x (u). Minimize f (x) (the objective function) subject to: hi (x) = u i . A proof of this as taken from [29] follows. The sensitivity to a change in u is desired. u) = ∂λ ( u ) ∂x ∂x ∂h (x( u)) ∇ x f (x(u)) + λ ( u) + h (x(u)) − u − λ ( u) ∂u ∂u ∂x ∂u [ ] Grouping terms of this functions gives 57 .APPENDIX A. u) = f (x(u)) + λ T (u) h (x(u)) − u with λ (u) is a vector of Lagrange multipliers [ ] The optimal solution for u = 0 is (x*. i = 1. One can show that the Lagrange multipliers are the negative of the derivative of the function which is being minimized with respect to the enforced constraint. This sensitivity can be expressed as the derivative of the Lagrange function with respect to the variable u. λ*). . . they are the marginal costs associated with enforcing the constraint.λ (u). the values of ui are constraints on some resource. In this problem formulation.λ . . λ*) would change. In other words. One can see that if this constraint value were varied.
the following is true. p.λ We also know that at the optimal solution. u) = ∂λ (u) ∂h (x(u)) ∂x λ (u) + h (x(u)) − u − λ (u) ∇ x f (x(u)) + ∂u ∂x 44 2 3 1∂u 4 444 144444 2444444 4 3 Second Term First Term [ ] Now.λ . 284]. * ) = −λ * ∇ u L(x * .λ * ) = ∇ u f (x * ) Combining these two equations. both the first and second terms must be zero at the optimal solution. λ * = −∇ u f (x * ) Thus the Lagrange multipliers are the negatives of the derivative of the function which is being minimized with respect to the enforced constraint.λ . u) ∇ x f (x(u)) + ∂x [h (x(u)) − u] = ∇ L(x .λ . and the following is left at the optimal solution ∇ u L(x * . u) + ∇ λ L(x . These Lagrange multipliers are thus often referred to as the shadow prices. study the first two terms of this more closely. u) λ Thus the following is written ∇ u L(x .λ . 58 . Therefore. u) must be zero at the optimal solution. it is seen that at the optimal solution. u) and ∇ λ L(x . then the λ* can be interpreted as the cost per unit of the resources associated with each constraint.∇ u L(x . both ∇ x L(x .λ . u) = ∂λ ( u ) ∂x ∇ x L(x .λ . If the objective function is considered a cost function.λ .λ . ∂h (x(u)) λ ( u) = ∇ x L(x . u) − λ (u) ∂u 2 144 44 3 1∂u 4 444 44 2 3 First Term Second Term By the KuhnTucker optimality conditions [20.
the derivative of the objective function with respect to them will be zero. they are not active). For all inequality constraints not being enforced (i.Although the previous derivation has not included inequality constraints.e. 59 . As for the inequality constraints that are being enforced. they will simply behave like equality constraints and the previous derivation can be applied to them. the derivation with them included is essentially the same.. A simple explanation will be proved here. because the problem solution will not be affected by changing the constraint (ui) since it is not being enforced anyway.
Appendix D contains a further Gradient term due to generator cost curves (the objective function) ∂L = bi + 2ci PGi ∂PGi Gradient terms due to the power flow equations ∂L = ∑ µ ∂Vi k ∂L = Pk ∂µ Pk ∂Qk ∂Pk + µ Qk Pk ∂Vi ∂Vi ∂L = ∑ µ ∂δ i k ∂Q ∂Pk + µ Qk k Pk ∂δ i ∂δ i ∂L = − µ Pk ∂PGi ∂L = Qk ∂µ Qk Gradient terms due to generator voltage set points ∂L = µ viset ∂Vi ∂L = Vi − VGi set ∂µ viset 60 . The listing is organized by those equations or constraints which result in the addition of the gradient term. Note: Many terms are surrounded by brackets. They are dependent on the inequality constraints being enforced. []. explanation of the calculation of these terms. (). CALCULATION OF THE GRADIENT OF THE LAGRANGIAN Appendix B contains a list of all the terms that make up the gradient of the Lagrangian for the OPF. are not always part of the gradient.APPENDIX B. Note: The terms that are shown surrounded by parenthesis.
Vm .δ k .Vm . or δm Gradient terms due to enforcement of hard voltage constraints ∂L = λ vih + − λ vil ∂Vi ∂L = + Vi − Vi max ∂λ vih ( ) ( ( ) ∂L = − Vi + Vi min ∂λ vil ( ) ) )) ∂L = + 2 k Vi − Vi max ∂Vi Gradient terms due to enforcement of soft voltage constraints ∂L = − 2 k − Vi + Vi min ∂Vi ( ( ( ( )) Gradient terms due to tap changing transformers 61 . or δm 2 ∂L 2 = S km − S km max ∂λ Skm Gradient terms due to enforcement of soft transmission line constraints 2 2 ∂ S ∂L km 2k S 2 − S = km max ∂x km ∂x where x = Vk . or δ m Gradient terms due to enforcement of area interchange constraints ∂L = ∑ {P } − Psched ∂µ int tie lines km ∂L ∂P = µ int ∑ km ∂x tie lines ∂x where x = Vk .Gradient terms due to maximum and minimum generator power output ∂L = − PGi + PGi min ∂λ PGil ∂L = (− λ ∂PGi ( ) ) ∂L = + PGi − PGi max ∂λ PGih ( ) PGil ) + (λ PGih Gradient terms due to enforcement of hard transmission line constraints ∂L = λ ∂x ∂ S 2 km Skm ∂x where x = Vk .Vm . δ k . δ k .
∂L ∂λ tkm max = t km − t km max ( ) ∂L ∂λ tkm min = − t km + t km min ( ) ∂L = µ ∂t km ∂Qk ∂Pm ∂Qm ∂Pk + µ Qm + µ Pk + µ Qk ∂t km ∂t km ∂t km ∂t km 2 2 λ ∂ S km + k ∂ S km + λ + tkm max + − λ tkm min Skm ∂t km ∂t km Pm ( ) ( ) Gradient terms due to phase shifting transformers ∂L ∂λ αkm max = α km− α km max ( ) ∂L ∂λ αkm min = − α km + α km min ( ) ∂L = ∂α km + λ µ Pm ∂Pm ∂Qm ∂Pk ∂Qk + µ Qm + µ Pk + µ Qk ∂α km ∂α km ∂α km ∂α km ∂ S 2 ∂ S 2 km + k km + λ αkm max + − λ αkm min Skm ∂α km ∂α km ( ) ( ) 62 .
Note: The terms that are shown surrounded by parenthesis. (). Note: Many terms are surrounded by brackets. CALCULATION OF THE HESSIAN OF THE LAGRANGIAN Appendix C contains a list of all the terms that make up the Hessian of the Lagrangian for the OPF. are not always part of the Hessian. explanation of the calculation of these terms. The listing is organized by those equations or constraints which result in the addition of the Hessian term. They are dependent on those inequality constraints being enforced.V j . Appendix E contains a further Hessian term due to generator cost curves (the objective function) ∂2 L = 2ci ∂PGi 2 Hessian terms due to the power flow equations ∂2 L = ∑ µ ∂x∂y k ∂ 2 Pk ∂ 2 Qk + µ Qk ∂x∂y where x and y = Vi . or δ j Pk ∂x∂y ∂P ∂2 L = k ∂Vi ∂µ Pk ∂Vi ∂P ∂2 L = k ∂δ i ∂µ Pk ∂δ i ∂2 L ∂PGk ∂µ = −1 Pk ∂Q ∂2 L = k ∂Vi ∂µ Qk ∂Vi ∂Q ∂2 L = k ∂δ i ∂µ Qk ∂δ i 63 .APPENDIX C.δ i . [].
or δm where x and y = Vk . or δ m Hessian terms due to area interchange constraints ∂2 L ∂P = ∑ km ∂µ int ∂x tie lines ∂x ∂ 2 Pkm ∂2 L = µ int ∑ ∂x∂y tie lines ∂x∂y where x and y = Vk . δ k . or δm Hessian terms due to hard voltage constraints 64 .Hessian terms due to generator voltage set points ∂2 L = +1 ∂Vi ∂µ viset Hessian terms due to maximum and minimum generator power output ∂2 L = (− 1) ∂PGi ∂λ PGil ∂2 L = (+ 1) ∂PGi ∂λ PGih Hessian terms due to hard transmission line constraints ∂ S 2 ∂2 L km = ∂x ∂λ Skm ∂x ∂2 L = λ ∂x∂y ∂ 2 S 2 km Skm ∂x∂y where x = Vk .Vm . or δm Hessian terms due to soft transmission line constraints 2 2 ∂ S 2 ∂ S 2 2 ∂ S 2 ∂2 L km km km = 2 k S km − S km max ∂x∂y + 2 k ∂x ∂y ∂x∂y where x = Vk .Vm .Vm . δ k .Vm . δ k . δ k .
depending on which end of the line is metered ∂t km Pk ∂ S 2 ∂2 L km or = ∂t km ∂λ Skm ∂t km ∂P ∂P ∂2 L = int = km or ∂t km ∂µ int ∂t km ∂t km ∂Pkm ∂t . depending on which end of the line is metered km Hessian terms due to phase shifting transformers 65 . or δ + m k m k Skm ∂t km ∂x ∂t km ∂x = (+ 1) ∂2 L ∂λ tkm min ∂t km = (− 1) ∂2 L ∂λ tkm max ∂t km ∂P ∂2 L = m ∂t km∂µ Pm ∂t km ∂2 L ∂t km∂µ ∂P = k ∂t km ∂Q ∂2 L = m ∂t km∂µ Qm ∂t km ∂Q ∂2 L = k ∂t km∂µ Qk ∂t km ∂ S 2 mk .δ .V .∂2 L = (+ 1) ∂Vi ∂λ vih ∂2 L = (− 1) ∂Vi ∂λ vil ∂2 L = (0) ∂Vi 2 Hessian terms due to soft voltage constraints ∂2 L = (2 k ) ∂Vi 2 Hessian terms due to tap changing transformers ∂2 L = µ ∂t km ∂x ∂ 2 Pm ∂ 2 Qm ∂ 2 Pk ∂ 2 Qk + µ Qk + µ Qm + µ Pk Pm ∂t km ∂x ∂t km ∂x ∂t km ∂x ∂t km ∂x 2 2 2 2 λ ∂ S km + k ∂ S km where x = V .
or δ m ∂2 L ∂λ αkm max ∂α ∂2 L ∂λ αkm min∂α = (− 1) km ∂2 L ∂α km∂µ ∂2 L ∂α km∂µ ∂2 L ∂α km ∂λ Pm ∂P = m ∂α km ∂P = k ∂α km ∂ S 2 km or = ∂α km ∂Qm ∂2 L = ∂α km∂µ Qm ∂α km ∂Qk ∂2 L = ∂α km∂µ Qk ∂α km ∂ S 2 depending on which end mk . of the line is metered ∂α km Pk Skm ∂P ∂P ∂P depending on which end ∂2 L = int = km or km .Vm . of the line is metered ∂α km ∂µ int ∂α km ∂α km ∂α km 66 . δ k .∂2 L = µ ∂α km ∂x ∂ 2 Pm ∂ 2 Qm + µ Qm +µ Pm ∂α km ∂x ∂α km ∂x 2 2 2 2 λ ∂ S km + k ∂ S km + Skm ∂α km ∂x ∂α km ∂x = (+ 1) km ∂ 2 Pk ∂ 2 Qk + µ Qk Pk ∂α km ∂x ∂α km ∂x where x = Vk .
the following partial derivatives can be found ∂Pk ∂Qk ∂Pk ∂Qk . . The listing is organized by the equations being differentiated. . SUMMARY OF DERIVATIVE CALCULATIONS In Appendices B and C. These partial derivatives are summarized in Appendix D. many partial derivatives left uncalculated. Calculation of partial derivatives of the net power and reactive power injections.1) where g km and bkm are elements of the real and imaginary parts of the network admittance matrix From Equation (D.1).and ∂V i ∂δ i ∂δ i ∂Vi are elements of the familiar power system Jacobian matrix ∂ 2 Pk 2 = 2 g kk ∂Vk ∂ 2 Pk ∂V 2 = 0 m ∂ 2 Pk = gkm cos δ k − δm + bkm sin δ k − δm ∂Vk ∂Vm ( ) ( ) ]] busses ∂ 2 Pk = −Vk ∑ Vn g kn cos(δk − δn ) + bkn sin(δk − δn ) 2 n =1 ∂δk all [ [ ∂ 2 Pk = VkVm g km cos δ k − δm + bkm sin δ k − δm ∂δ k ∂δm [ ( ) ( )] 67 . Pk = Vk ∑ Vm g km cos(δk − δm ) + bkm sin(δ k − δm ) − PGk + PLk m=1 N N [ [ ]] Qk = Vk ∑ Vm g km sin(δk − δm ) − bkm cos(δ k − δm ) − QGk + QLk m =1 [ [ ]] (D.APPENDIX D.
∂ 2 Pk 2 = VkVm − g km cos δ k − δm − bkm sin δ k − δm ∂δm [ ( ) ( )] )]} ∂ 2 Pk all busses = ∑ Vn − g kn sin δ k − δn + bkn cos δ k − δn n ∂Vk ∂δ k {[ ( ) ( ∂ 2 Pk = Vm g km sin δ k − δm − bkm cos δ k − δm ∂Vk ∂δ m [ ( ) ( )] )] ∂ 2 Pk = Vk − g km sin δ k − δm + bkm cos δ k − δm ∂Vm ∂δ k [ ( ) ( ∂ 2 Pk = Vk g km sin δ k − δm − bkm cos δ k − δm ∂Vm ∂δ m [ ( ) ( )] ∂ 2 Qk ∂V 2 = −2bkk k ∂ 2 Qk ∂V 2 = 0 m ∂ 2 Qk = g km sin δ k − δm − bkm cos δ k − δm ∂Vk ∂Vm ( ) ( ) ]] all busses ∂ 2 Qk Vn g kn sin(δk − δn ) − bkn cos(δk − δn ) 2 = −Vk ∑ n =1 ∂δk [ [ ∂ 2 Qk = VkVm g km sin δ k − δm − bkm cos δ k − δm ∂δ k ∂δm [ ( ) ( )] ∂ 2 Qk 2 = VkVm − g km sin δ k − δm + bkm cos δ k − δm ∂δm [ ( ) ( )] )]} ∂ 2 Qk all busses = ∑ Vn g kn cos δ k − δn + bkn sin δ k − δn n ∂Vk ∂δ k {[ ( ) ( 68 .
and bkm are the contributions to the network admittance matrix from the branch between bus k and bus m. ∂ 2 Qk = −Vm g km cos δ k − δm + bkm sin δ k − δm ∂Vk ∂δ m [ ( ) ( )] ∂ 2 Qk = Vk g km cos δ k − δm + bkm sin δ k − δm ∂Vm ∂δ k [ ( ) ( )] )] ∂ 2 Qk = −Vk g km cos δ k − δm + bkm sin δ k − δm ∂Vm ∂δ m [ ( ) ( Calculation of the partial derivatives of the power and reactive power flow between buses. Pkm = + Vk 2 g kk + VkVm g km cos(δk − δm ) + bkm sin(δk − δm ) Qkm = − Vk 2bkk k m km [ + V V [g sin(δ k − δm ) − bkm cos(δk − δm ] )] (D. bkk .2).2) where g kk . g km . From Equation (D. the following partial derivatives can be found ∂Pkm = 2Vk g kk + Vm g km cos δ k − δ m + bkm sin δ k − δ m ∂Vk [ ( ) ( )] ∂Pkm = Vk g km cos δ k − δ m + bkm sin δ k − δ m ∂Vm [ ( ) ( )] )] ∂Pkm = VkVm − g km sin δ k − δ m + bkm cos δ k − δ m ∂δ k [ ( ) ( ∂Pkm = VkVm g km sin δ k − δ m − bkm cos δ k − δ m ∂δ m [ ( ) ( )] ∂Qkm = −2Vk bkk + Vm g km sin δ k − δ m − bkm cos δ k − δ m ∂Vk [ ( ) ( )] 69 .
∂Qkm = Vk g km sin δ k − δ m − bkm cos δ k − δ m ∂Vm [ ( ) ( )] )] )] ∂Qkm = VkVm g km cos δ k − δ m + bkm sin δ k − δ m ∂δ k [ ( ) ( ∂Qkm = VkVm − g km cos δ k − δ m − bkm sin δ k − δ m ∂δ m [ ( ) ( ∂ 2 Pkm ∂V 2 = 2 g kk k ∂ 2 Pkm = g km cos δ k − δ m + bkm sin δ k − δ m ∂Vk ∂Vm ( ) ( ) ∂ 2 Pkm 2 = 0 ∂Vm ∂ 2 Pkm 2 = −Vk Vm g km cos δ k − δ m + bkm sin δ k − δ m ∂δ k [ ( ) ( )] ∂ 2 Pkm = VkVm g km cos δ k − δ m + bkm sin δ k − δ m ∂δ k ∂δ m [ ( ) ( )] ∂ 2 Pkm = −VkVm g km cos δ k − δ m + bkm sin δ k − δ m 2 ∂δ m [ ( ) ( )] ∂ 2 Pkm = Vm − g km sin δ k − δ m + bkm cos δ k − δ m ∂Vk ∂δ k [ ( ) ( )] ∂ 2 Pkm = Vm g km sin δ k − δ m − bkm cos δ k − δ m ∂Vk ∂δ m [ ( ) ( )] )] ∂ 2 Pkm = Vk − g km sin δ k − δ m + bkm cos δ k − δ m ∂Vm ∂δ k [ ( ) ( 70 .
Skm2 71 . ∂ 2 Pkm = Vk g km sin δ k − δ m − bkm cos δ k − δ m ∂Vm ∂δ m [ ( ) ( )] ∂ 2 Qkm ∂V 2 = −2bkk k ∂ 2 Qkm = g km sin δ k − δ m − bkm cos δ k − δ m ∂Vk ∂Vm ( ) ( ) ∂ 2 Qkm ∂V 2 = 0 m ∂ 2 Qkm 2 = −Vk Vm g km sin δ k − δ m − bkm cos δ k − δ m ∂δ k [ ( ) ( )] ∂ 2 Qkm = VkVm g km sin δ k − δ m − bkm cos δ k − δ m ∂δ k ∂δ m [ ( ) ( )] ∂ 2 Qkm 2 = −Vk Vm g km sin δ k − δ m − bkm cos δ k − δ m ∂δ m [ ( ) ( )] ∂ 2 Qkm = Vm g km cos δ k − δ m + bkm sin δ k − δ m ∂Vk ∂δ k [ ( ) ( )] )] )] )] ∂ 2 Qkm = Vm − g km cos δ k − δ m − bkm sin δ k − δ m ∂Vk ∂δ m [ ( ) ( ∂ 2 Qkm = Vk + g km cos δ k − δ m + bkm sin δ k − δ m ∂Vm ∂δ k [ ( ) ( ∂ 2 Qkm = Vk − g km cos δ k − δ m − bkm sin δ k − δ m ∂Vm ∂δ m [ ( ) ( Calculation of the partial derivatives of the square of the MVA flow on a line.
S km = Pkm 2 + Qkm 2
2
(D.3)
From Equation (D.3), the following partial derivative can be found
∂ S 2 ∂P ∂Q km = 2 Pkm km + 2Qkm km ∂x ∂x ∂x where x = Vk ,Vm ,δ k , or δm
Equation (D.3) can also be simplified to the following Skm = Vk 4 ykk 2 + Vk 2 Vm2 ykm2 + 2Vk 3 Vm ykm ykk cos(δ k − δ m+ δ kk − δ
2 km
)
where gkk + jbkk = ykk e j δ kk and gkm + jbkm = ykm e j δ km From Equation (D.4), the following partial derivatives can be found
∂2 S 2 km = − 2Vk 3 Vm y km y kk cos(δ k − δ m + δ 2 ∂δ k ∂2 S 2 km = 2Vk 3 Vm y km y kk cos(δ k − δ m + δ ∂δ k ∂δ m ∂2 S 2 km = − 2Vk 3 Vm y km y kk cos(δ k − δ m + δ 2 ∂δ m −δ
(D.4)
kk
km
)
kk
−δ
km
) ) )
kk
−δ
km
∂2 S 2 km = 12Vk 2 y kk 2 + 2Vm 2 y km 2 + 12Vk Vm y km y kk cos(δ k − δ m + δ kk − δ 2 ∂Vk ∂2 S 2 km = 4Vk Vm y km 2 + 6Vk 2 y km y kk cos(δ k − δ m + δ ∂Vk ∂Vm ∂2 S 2 km = 2Vk 2 y km 2 2 ∂Vm ∂2 S km ∂Vk ∂δ
2
km
kk
−δ
km
)
= − 6Vk 2 Vm y km y kk sin(δ k − δ m + δ k
kk
−δ
km
)
72
∂2 S km ∂Vk ∂δ
= 6Vk 2 Vm y km y kk sin(δ k − δ m + δ m
2
kk
−δ
km
)
∂2 S 2 km = − 2Vk 3 y km y kk sin(δ k − δ m + δ ∂Vm ∂δ k ∂2 S 2 km = 2Vk 3 y km y kk sin(δ k − δ m + δ ∂Vm ∂δ m
kk
−δ
km
)
kk
−δ
km
)
Calculation of partial derivatives with respect to tap ratio and phase shift variables Before beginning calculation of the partial derivative with respect to tap ratio and phase shift variables, first consider what equations they affect. In the modeling of power systems, tap ratios and phase shifts are typically represented as alterations to the network admittance matrix. Figure D.1 shows the model for a transformer with tap changing and/or phase shifting capabilities.
Ik + Vk _
r + jx Ycap 2 Ycap 2 1 : tkm e +jαkm
Figure D.1 Transformer Model
Im + Vm _
Using this model, the entries in the network admittance matrix due to the transformer can be found as in the book by Grainger [20, pp. 361362].
73
I k row k = I row m m
col k col m Ycap e − jαkm y V − y+ 2 k t km + jα where y = 1 = g + jb Ycap 1 e km r + jx y y+ − V t 2 m t km 2 4444244443 1 km
Y
From this model, the following derivatives are defined.
∂Ykk ∂[Y ] ∂t km = ∂t km ∂Ymk ∂t km ∂Ykm 0 ∂t km = ∂Ymm e + jα km y ∂t km t km 2 e − jαkm y 0 t km 2 = − 2 −1 Ymk y t km 3 t km −1 Ykm t km −2 Y t km mm
∂ 2Ykk ∂ 2 [Y ] ∂t km 2 = 2 ∂t km 2 ∂ Ymk ∂t 2 km
∂Ykk ∂[Y ] ∂α km = ∂α km ∂Ymk ∂α km ∂ 2Ykk ∂ 2 [Y ] ∂α km 2 = 2 ∂α km 2 ∂ Ymk ∂α 2 km
∂ 2Ykm 0 ∂t km 2 = ∂ 2Ymm − 2e + jαkm y ∂t km 2 t km 3
∂Ykm 0 ∂α km = ∂Ymm − e + jαkm j y ∂α km t km ∂ 2 Ykm 0 ∂α km 2 = ∂ 2 Ymm e + jα km y ∂α km 2 t km
− 2e − jαkm y 0 t km 3 = 6 2 Y 4 y t km 2 mk t km
j e − jα km y 0 t km = jY 0 mk
Ykm t km 6 Y t km 2 mm 2
2
− jYkm 0 = jg 0 mk − bmk − Ykm 0 = −g 0 mk − jbmk
− jg km + bkm 0 − g km − jbkm 0
e − jα km y 0 t km = 0 − Ymk
From these equations, the partial derivatives of elements of the network admittance matrix with respect to tap ratio and phase shift angle can be summarized as below.
74
∂b ∂g ∂w = u km + v km ∂t km ∂t km ∂t km ∂ 2 bkm ∂ 2 g km ∂2w =u +v ∂t km 2 ∂t km 2 ∂t km 2 where x = Vk . the following partial derivatives with respect to tap ratio can be found. Therefore. t km ) = u(•)bkm (t km ) + v(•) g km (t km ) . ∂Pk ∂Pkm Vk Vm g km cos(δ k − δ = =− t km ∂t km ∂t km [ m )+ b km sin(δ k − δ cos(δ k − δ m )] )] ∂Qk ∂Qkm Vk Vm g km sin (δ k − δ = =− t km ∂t km ∂t km [ m )− b km m 75 . the derivatives that need to be calculated may be found as follows. Vm .∂g kk =0 ∂t km g ∂g mk = − mk t km ∂t km ∂ 2 g kk =0 ∂t km 2 ∂ 2 g mk 2 g mk = ∂t km 2 t km 2 ∂g kk =0 ∂α km ∂g mk = − bmk ∂α km ∂ 2 g kk =0 ∂α km 2 ∂ 2 g mk = − g mk ∂α km 2 g ∂g km = − km t km ∂t km ∂g mm 2g = − mm t km ∂t km ∂ 2 g km 2 g km = t km 2 ∂t km 2 ∂ 2 g mm 6 g mm = ∂t km 2 t km 2 ∂g km = bkm ∂α km ∂g mm =0 ∂α km ∂ 2 g km = − g km ∂α km 2 ∂ 2 g mm =0 ∂α km 2 ∂bkk =0 ∂t km b ∂bmk = − mk t km ∂t km ∂ 2 bkk =0 ∂t km 2 ∂ 2 bmk 2bmk = ∂t km 2 t km 2 ∂bkk =0 ∂α km ∂bmk = g mk ∂α km ∂ 2 bkk =0 ∂α km 2 ∂ 2 bmk = − bmk ∂α km 2 b ∂bkm = − km t km ∂t km ∂bmm 2b = − mm t km ∂t km ∂ 2 bkm 2bkm = t km 2 ∂t km 2 ∂ 2 bmm 6bmm = ∂t km 2 t km 2 ∂bkm = − g km ∂α km ∂bmm =0 ∂α km ∂ 2 bkm = − bkm ∂α km 2 ∂ 2 bmm =0 ∂α km 2 The equations that have to be differentiated are of the form w(•. or δ k ∂g ∂2w ∂ ∂b = u km + v km ∂t km ∂x ∂x ∂t km ∂t km Using this process. δ k .
∂Pm ∂Pmk Vm 2Vm g mm + Vk g mk cos(δ m − δ = =− t km ∂t km ∂t km { [ k )+ b mk sin (δ m − δ k )]} k ∂Qm ∂Qmk Vm ∂t = ∂t = − t − 2Vm bmm + Vk g mk sin(δ m − δ km km km { [ k )−b mk cos(δ m − δ )]} ∂ 2 Pk ∂ 2 Pkm 2Vk Vm g km cos(δ k − δ 2 = 2 = t km 2 ∂t km ∂t km [ m )+ b km sin(δ k − δ m )] ∂ 2 Qk ∂ 2 Qkm 2Vk Vm g km sin(δ k − δ 2 = 2 = t km 2 ∂t km ∂t km [ m )− b km cos(δ k − δ m )] k ∂ 2 Pm ∂ 2 Pmk 2Vm V 2 = 2 = 2 3 m g mm + V k g mk cos(δ m − δ ∂t km ∂t km t km { [ k )+ b mk sin(δ m − δ )]} k ∂ 2 Qm ∂ 2 Qmk 2Vm V 2 = 2 = 2 − 3 m bmm + V k g mk sin (δ m − δ ∂t km ∂t km t km { [ k )− b mk cos(δ m − δ )]} ∂ 2 Pk ∂ 2 Pkm Vm g cos(δ k − δ = =− t km km ∂t km ∂Vk ∂t km ∂Vk [ m )+ b km sin (δ k − δ sin(δ k − δ m )] )] m ∂ 2 Pk ∂ 2 Pkm Vk g cos(δ k − δ = =− t km km ∂t km ∂Vm ∂t km ∂Vm [ m )+ b m km m ∂ 2 Pk ∂ 2 Pkm Vk V m − g km sin (δ k − δ = =− t km ∂t km ∂δ k ∂t km ∂δ k [ )+ b km cos(δ k − δ )] ∂ 2 Pk ∂t km ∂δ ∂ 2 Pkm Vk Vm g km sin(δ k − δ = =− t km m ∂t km ∂δ m [ m )− b km km cos(δ k − δ m )] ∂ 2 Qk ∂ 2 Qkm Vm g sin(δ k − δ = =− t km km ∂t km ∂Vk ∂t km ∂Vk [ m )− b cos(δ k − δ cos(δ k − δ m )] )] ∂ 2 Qk ∂ 2 Qkm Vk g sin (δ k − δ = =− t km km ∂t km ∂Vm ∂t km ∂Vm [ m )− b km m 76 .
Vm . ∂ 2 Qk ∂ 2 Qkm Vk V m g km cos(δ k − δ = =− t km ∂t km ∂δ k ∂t km ∂δ k [ m )+ b m km sin(δ k − δ m )] m ∂ 2 Qk ∂ 2 Qkm Vk Vm − g km cos(δ k − δ = =− t km ∂t km ∂δ m ∂t km ∂δ m [ )− b km sin(δ k − δ )] ∂ 2 Pm ∂ 2 Pmk Vm g cos(δ m − δ = =− t km mk ∂t km ∂Vk ∂t km ∂Vk [ k )+ b mk sin (δ m − δ k )] mk ∂ 2 Pm ∂ 2 Pmk 1 4Vm g mm + Vk g mk cos(δ m − δ = =− t km ∂t km ∂Vm ∂t km ∂Vm { [ k )+ b sin(δ m − δ k )]} ∂ 2 Pm ∂ 2 Pmk Vk Vm g mk sin(δ m − δ = =− t km ∂t km ∂δ k ∂t km ∂δ k [ k )− b k mk cos(δ m − δ k )] k ∂ 2 Pm ∂ 2 Pmk Vk Vm − g mk sin (δ m − δ = =− t km ∂t km ∂δ m ∂t km ∂δ m [ )+ b mk cos(δ m − δ )] ∂ 2 Qm ∂ 2 Qmk Vm − bmk cos(δ m − δ =− = t km ∂t km ∂Vk ∂t km ∂Vk [ k ) + g mk sin(δ m − δ k )] ) + g mk sin(δ m − δ k )]} k ∂ 2 Qm ∂ 2 Qmk 1 − 4Vm bmm + Vk − bmk cos(δ m − δ =− = t km ∂t km ∂Vm ∂t km ∂Vm { [ k ∂ 2 Qm ∂ 2 Qmk Vk Vm − g mk cos(δ m − δ = =− t km ∂t km ∂δ k ∂t km ∂δ k [ k )− b mk sin (δ m − δ )] ∂ 2 Qm ∂ 2 Qmk Vk Vm g mk cos(δ m − δ = =− t km ∂t km ∂δ m ∂t km ∂δ m [ k )+ b mk sin(δ m − δ k )] ∂ S 2 ∂Q ∂P km = 2 Pkm km + 2Q km km where x = Vk . δ k .δ k . or δ m ∂t km ∂t km ∂t km 77 . or δ m ∂t km ∂t km ∂t km ∂ S 2 ∂P ∂Q mk = 2 Pmk mk + 2Q mk mk where x = Vk .Vm .
Vm .Vm . the following partial derivatives can be found. δ k . ∂2 S 2 ∂ 2 Q km ∂Qkm ∂Qkm ∂ 2 Pkm ∂Pkm ∂Pkm km = 2 Pkm + 2 + 2Q km + 2 ∂t km ∂x ∂t km ∂x ∂x ∂t km ∂t km ∂x ∂x ∂t km where x = Vk . or δ m ∂2 S 2 ∂ 2 Q mk ∂Qmk ∂Qmk ∂ 2 Pmk ∂Pmk ∂Pmk mk = 2 Pmk + 2 + 2 + 2Q mk ∂t km ∂x ∂t km ∂x ∂x ∂t km ∂t km ∂x ∂x ∂t km where x = Vk . ∂Pk ∂Pkm = = Vk Vm bkm cos(δ k − δ ∂α km ∂α km [ m )− g km sin (δ k − δ m )] ∂Qk ∂Qkm ∂α = ∂α = Vk Vm bkm sin(δ k − δ km km [ m )+ g k km cos(δ k − δ m )] k ∂Pm ∂Pmk = = Vk Vm − bmk cos(δ m − δ ∂α km ∂α km [ )+ g mk sin (δ m − δ )] ∂Qm ∂Qmk ∂α = ∂α = Vk Vm − bmk sin(δ m − δ km km [ k )− g m mk cos(δ m − δ k )] m ∂ 2 Pk ∂ 2 Pkm 2 = 2 = −V k Vm g km cos(δ k − δ ∂α km ∂α km [ [ )+b km sin (δ k − δ cos(δ k − δ sin (δ m − δ )] )] )] )] ∂ 2 Qk ∂ 2 Qkm 2 = 2 = −V k Vm g km sin (δ k − δ ∂α km ∂α km ∂ 2 Pm ∂ 2 Pmk 2 = 2 = −V k V m g mk cos(δ m − δ ∂α km ∂α km m )− b km m [ [ k )+ b )− b mk k ∂ 2 Qm ∂ 2 Qmk 2 = 2 = −V k V m g mk sin (δ m − δ ∂α km ∂α km k mk cos(δ m − δ k 78 . or δ m Using the same process for phase shift angles. δ k .
∂ 2 Pk ∂ 2 Pkm = = Vm bkm cos(δ k − δ ∂α km ∂Vk ∂α km ∂Vk [ m )− g km sin (δ k − δ sin(δ k − δ m )] )] m ∂ 2 Pk ∂ 2 Pkm = = Vk bkm cos(δ k − δ ∂α km ∂Vm ∂α km ∂Vm [ m )− g m km m ∂ 2 Pk ∂ 2 Pkm = = Vk Vm − bkm sin(δ k − δ ∂α km ∂δ k ∂α km ∂δ k [ )− g km cos(δ k − δ )] ∂ 2 Pk ∂α km ∂δ ∂ 2 Pkm = m ∂α km ∂δ = Vk Vm bkm sin (δ k − δ m [ m )+ g km cos(δ k − δ m )] ∂ 2 Qk ∂ 2 Qkm = = Vm bkm sin(δ k − δ ∂α km∂Vk ∂α km∂Vk [ m ) + g km cos(δ k − δ m )] ∂ 2 Qk ∂α ∂V = km m ∂ 2 Qkm ∂α ∂V = Vk bkm sin (δ k − δ km m [ m )+ g m km cos(δ k − δ m )] m ∂ 2 Qk ∂ 2 Qkm = = Vk Vm bkm cos(δ k − δ ∂α km ∂δ k ∂α km ∂δ k [ )− g m km sin(δ k − δ )] m ∂ 2 Qk ∂α ∂δ km ∂ 2 Qkm = ∂α ∂δ m km = VkVm − bkm cos(δ k − δ m [ )+ g mk km sin(δ k − δ )] ∂ 2 Pm ∂α ∂V = km k ∂ 2 Pm ∂α ∂V = km m ∂ 2 Pmk ∂α ∂V = Vm − bmk cos(δ m − δ km k [ k )+ g sin(δ m − δ sin (δ m − δ k )] )] k ∂ 2 Pmk ∂α ∂V = Vk − bmk cos(δ m − δ km m [ k )+ g k mk k ∂ 2 Pm ∂ 2 Pmk ∂α ∂δ = ∂α ∂δ = Vk Vm − bmk sin (δ m − δ km k km k [ )− g mk cos(δ m − δ )] ∂ 2 Pm ∂α ∂δ km = m ∂ 2 Pmk ∂α ∂δ km = Vk Vm bmk sin(δ m − δ m [ k )+ g mk cos(δ m − δ k )] 79 .
δ k . or δ m ∂α km ∂α km ∂α km ∂2 S 2 ∂ 2 Q km ∂Qkm ∂Qkm ∂ 2 Pkm ∂Pkm ∂Pkm km = 2 Pkm + 2 + 2Q km + 2 ∂α km ∂x ∂α km ∂x ∂x ∂α km ∂α km ∂x ∂x ∂α km where x = Vk . δ k . or δ m ∂2 S 2 ∂ 2 Q mk ∂Qmk ∂Qmk ∂ 2 Pmk ∂Pmk ∂Pmk mk = 2 Pmk + 2 + 2Q mk + 2 ∂α km ∂x ∂α km ∂x ∂x ∂α km ∂α km ∂x ∂x ∂α km where x = Vk .δ k .Vm .Vm .Vm . ∂ 2 Qm ∂α ∂V = km k ∂ 2 Qm ∂α ∂V = km k ∂ 2 Qmk ∂α ∂V = Vm − bmk sin (δ m − δ km k [ k )− g )− g k mk cos(δ m − δ cos(δ m − δ k )] )] k ∂ 2 Qmk ∂α ∂V = Vk − bmk sin(δ m − δ km k [ k mk k ∂ 2 Qm ∂ 2 Qmk ∂α ∂δ = ∂α ∂δ = VkVm bmk cos(δ m − δ km k km k [ )− g k mk sin (δ m − δ )] k ∂ 2 Qm ∂α ∂δ km = m ∂ 2 Qmk ∂α ∂δ km = Vk Vm − bmk cos(δ m − δ m [ )+ g mk sin (δ m − δ )] ∂ S 2 ∂Q ∂P km = 2 Pkm km + 2Q km km where x = Vk .Vm . or δ m ∂α km ∂α km ∂α km ∂ S 2 ∂Q ∂P mk = 2 Pmk mk + 2Q mk mk where x = Vk . δ k . or δ m 80 .
The economic information of the system is shown in Table E. A oneline diagram of the system is shown in Figure E.] 0.u. The bus characteristics of the system are shown in Table E.04 0.1.04 0.85 $/MWH 0.08 0.02 0.u.1 OneLine Diagram of SixBus System The line characteristics of the system are shown in Table E. SIXBUS SAMPLE POWER SYSTEM Appendix E contains information on the sixbus sample power system discussed in the thesis.02 Line Limit [MVA] 100 100 100 100 100 50 100 81 .00 $/MVRH 20 MVR Bus 2 1.00 PU 100 MW 20 MVR Bus 6 15.] 0.02 0.u.08 0.23 $/MVRH Bus 3 1.02 0.08 0.02 0.25 $/MWH 0.19 $/MWH 0.18 $/MWH 0.08 0.11 $/MVRH 100 MW 10 MVR 15.00 $/MVRH Bus 4 13.2.85 $/MWH 0.97 PU 100 MW 50 MVR 1.00 $/MVRH 100 MW 20 MVR 100 MW 20 MVR Bus 1 1.04 0.00 PU 14.] 0.00 PU 91 MW 60 MVR Bus 5 0.08 0.02 0.04 0.1. 196 MW 27 MVR 13.37 $/MWH 100 MW 0.00 PU 86 MW 92 MVR Figure E.02 0.04 Reactance [p.3.08 Line Charging [p.08 0.04 0. Table E.1 Line characteristics for sixbus system From Bus 1 1 2 3 3 4 4 To Bus 2 5 4 5 6 5 6 Circuit 1 1 1 1 1 1 1 Resistance [p.00 $/MVRH 237 MW 23 MVR 0.97 PU 14.04 0.APPENDIX E.
0096 0.0 9.0130 0.6 13.Table E.3 Economic information for sixbus system Generator Bus 1 2 3 4 $ a hr 105 96 105 94 $ b MW hr 12.0094 82 .0120 0.2 Bus characteristics for sixbus system Bus Number 1 2 3 4 5 6 Load [MW] 100 100 100 100 100 100 Load [MVAR] 20 20 20 20 50 10 Min Generation [MW] 50 50 50 50 0 0 Max Generation [MVAR] 250 250 250 250 0 0 Table E.4 $ c 2 MW hr 0.0 9.
03 PU Bus 23 1.50 $/MWH 200 MW 430 MW 106 MVR 340 MW Bus 5 13.00 PU Bus 6 150 MW 12. The economic information of the system is shown in Table F.58 $/MWH 225 MW Bus 8 0.90 $/MWH 250 MW 432 MW 1.97 PU Bus 201 Bus 202 300 MW 0.000 tap 1.52 $/MWH Bus 3 1.44 $/MWH 1.99 PU Bus 103 13.APPENDIX F.48 $/MWH 97 MVR 1.27 $/MWH 300 MW 350 MW 1.02 PU Bus 1 Bus 101 464 MW 12. 125 MW 700 MW 106 MVR 1.03 PU Bus 102 763 MW 17.01 PU 13.98 PU 200 MW 359 MW 0.39 $/MWH 1.49 $/MWH Bus 7 0.3.98 PU Bus 25 Bus 4 13.01 PU 13.53 $/MWH 350 MW 1.02 PU Bus 9 400 MW 12.22 $/MWH 1.1 OneLine Diagram of the TwentyThree Bus System The line characteristics of the system are shown in Table F.03 PU 200 MW 0.02 PU Bus 105 543 MW 98 MVR 16.99 PU 100 MW 13.61 $/MWH 16.03 PU Bus 203 15.55 $/MWH Bus 122 1.96 PU 15.86 $/MWH 106 MVR 1.68 $/MWH Bus 2 0.57 $/MWH 13.25 $/MWH 14.000 tap 75 MW 13.98 PU 450 MW 97 MVR 1.51 $/MWH Bus 28 1.01 PU 13. A oneline diagram of the system is shown in Figure F.03 PU Bus 204 657 MW 200 MW 300 MW 1.49 $/MWH 13.99 PU 13.86 $/MWH 175 MW Figure F.03 PU 16.64 $/MWH 0. 83 .01 PU 1.000 tap 1. The bus characteristics of the system are shown in Table F.2. TWENTYTHREE BUS SAMPLE POWER SYSTEM Appendix F contains information on the twentythree bus sample power system discussed in the thesis.01 PU Bus 224 13.51 $/MWH Bus 104 16.16 $/MWH 0.1.1.
020 0.0000 0.0000 0.0005 0.070 0.010 0.0000 0.0005 0.060 0.0000 0.0000 0.0005 0.050 0.u.050 0.0000 0.020 0.] 0.Table F.0005 0.060 0.0000 0.009 0.0000 0.] 0.020 0.025 0.015 0.010 0.100 0.0008 0.070 0.0000 0.100 0.0006 0.002 0.0005 0.000 0.030 0.0008 0.0000 0.010 0.060 0.020 0.010 0.001 0.u.010 0.0000 0.050 0.0050 0.0005 0.007 0.040 0.020 0.000 Reactance [p.000 0.060 0.020 0.u.015 0.0000 0.0008 0.100 Line Charging [p.020 0.030 0.009 0.060 0.0005 0.050 0.002 0.060 0.010 0.040 0.040 0.030 0.0000 Line Limit [MVA] 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 500 500 1000 1000 1000 1000 1000 1000 1000 1000 1000 84 .0000 0.050 0.070 0.100 0.0005 0.100 0.015 0.] 0.050 0.030 0.030 0.030 0.1 Line characteristics for twentythree bus system From Bus 1 3 6 6 2 8 5 4 3 1 8 25 23 5 102 122 204 224 105 201 203 202 201 102 102 104 101 103 To Bus 2 23 7 9 6 9 7 7 4 3 28 28 25 25 122 23 224 28 201 203 204 204 202 101 104 105 103 105 Circuit 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Resistance [p.010 0.0000 0.
5255 7.008161 0.6570 7.9140 $ c 2 MW hr 0.8609 7.6570 7.90 64.28 58.3 Economic information for sixbus system Generator Bus 1 3 5 6 9 101 102 105 203 204 $ a hr 55.006897 0.24 65.9181 7.81 $ b MW hr 7.5102 7.007657 0.005279 0.2937 7.2 Bus characteristics for twentythree bus system Bus Number 1 2 3 4 5 6 7 8 9 23 25 28 101 102 103 104 105 122 201 202 203 204 224 Load [MW] 125 300 250 75 350 150 100 225 175 0 0 0 700 200 350 450 200 0 300 200 300 0 0 Load [MVAR] 50 100 50 10 100 35 10 75 50 0 0 0 200 20 100 75 20 0 100 20 100 0 0 Min Generation [MW] 100 0 100 0 150 100 0 0 100 0 0 0 100 200 0 0 100 0 0 0 50 100 0 Max Generation [MW] 1000 0 600 0 1000 400 0 0 800 0 0 0 600 1000 0 0 600 0 0 0 200 800 0 Table F.7141 7.004165 85 .59 65.47 72.94 74.004367 0.008285 0.009981 0.Table F.006428 0.91 70.5330 7.90 80.43 58.007657 0.
00 PU 18. 431 MW 59 MVR Bus 1 50 MW 10 MVR 0.u.00 0.APPENDIX G.3.00 Limit [MVA] 1000 1000 1000 1000 1000 1000 1000 86 .06 0.11 0.] 0.07 $/MWH 0.75 $/MWH 0.1.00 $/MVRH 18.11 $/MWH 0.u.04 PU Bus 7 18.05 0.02 $/MVRH Bus 6 1.51 $/MWH 0.1 OneLine Diagram of the SevenBus System The line characteristics of the system are shown in Table G. A oneline diagram of the system is shown in Figure G. The bus characteristics of the system are shown in Table G.00 0.00 0.07 0.04 PU 50 MW 10 MVR 16.0000 tap Bus 4 1.19 $/MWH 0.00 0.21 $/MVRH 1.u.10 0.00 PU 19.28 $/MVRH 18.1 From Bus 1 2 2* 3* 4 6 7 Line characteristics for sevenbus system To Bus 3 1 4* 5* 6 7 5 R [p.1. The economic information of the system is shown in Table G.01 PU Bus 3 1. SEVENBUS SAMPLE POWER SYSTEM Appendix G contains information on the twentythree bus sample power system discussed in the thesis. Table G.0000 tap Bus 5 1.] 0.01 PU 50 MW 10 MVR 50 MVR 100 MW 30 MVR 100 MW 30 MVR 136 MW 81 MVR Figure G.02 0.50 $/MWH 0.32 $/MVRH 150 MW 60 MVR 1.00 0.00 $/MVRH 1.05 0.07 X [p.00 0.36 $/MVRH 50 MW 10 MVR 20.03 0.02 0.99 PU Bus 2 1.18 $/MWH 0.] 0.11 C [p.01 0.00 0.03 0.2.
u.1.3 Economic information for sevenbus system a $ hr Generator Bus 1 7 b $ MW hr c $ MW 2 hr 90 115 9.0090 0. [MW] 0 0 0 0 0 0 0 Max Gen.0150 87 . [MW] 1000 0 0 0 0 0 1000 Load [MW] 150 50 50 50 50 100 100 * A 50 MVAR capacitor bank is at bus 6 Table G.1 p.2 Bus Number 1 2 3 4 5 6* 7 Bus characteristics for sevenbus system Load [MVAR] 60 10 10 10 10 30 30 Min Gen.0 0.0 15.* Note: Elements from buses 24 and 35 are transformers with tap ranges of 0.9 . Table G.
00 0.22 $/MWH 0.94 $/MVRH 0. Table H. 20.00 $/MVRH 504 MW 193 MVR 500 MW 100 MVR 300 MW 14 MVR 100 MVR Figure H.u.04 PU 238 MW 55 MVR 100 MW 30 MVR Area Two Area One 72 MW 21 MVR Bus 1 1.3.u.1 Line characteristics for threebus system From Bus 1 1 2 To Bus 2 3 3 R [p.02 X [p.] 0.00 $/MVRH 73 MW 15 MVR 411 MW 70 MVR Bus 3 1. A oneline diagram of the system is shown in Figure H.u.06 C [p.1. The bus characteristics of the system are shown in Table H.08 0.1 OneLine Diagram of the ThreeBus System The line characteristics of the system are shown in Table H.1.02 0.] 0.03 0. THREEBUS SAMPLE POWER SYSTEM Appendix H contains information on the twentythree bus sample power system discussed in the thesis.] 0.12 0.85 $/MWH 0.00 Limit [MVA] 500 500 600 88 .00 0.9710 PU Bus 2 72 MVR 50.2.APPENDIX H. The economic information of the system is shown in Table H.18 $/MWH 0.04 PU 75 MW 227 MW 22 MVR 73 MW 64 MVR 22.
0100 89 . [MW] 100 0 100 Max Gen.0 13.3 Economic information for threebus system Generator Bus 1 3 a $ hr b $ MW hr c $ MW 2 hr 100 150 25.0250 0.Table H.0 0. [MW] 800 0 800 Table H.2 Bus characteristics for threebus system Bus Number 1 2 3 Load [MW] 500 300 100 Load [MVAR] 100 100 30 Min Gen.
October 1968. Shuttleworth. J. Hunt and G. F. 33. “Optimal Power Flow Solutions. Vol. “Pioneering electricity reform in South America. September 1996. Federal Energy Regulatory Commission of the United States of America. Wood and B. H. G. New York. Vol. “Charting a new course in California. October 1984. M. 9. Power Generation Operation and Control. 2631. pp. July 1996. Ashley. Inc. No. R. pp. RM95800. 8.” IEEE Spectrum. Tinney. NY: John Wiley & Sons. Hughes and W. A.REFERENCES [1] [2] [3] [4] [5] [6] [7] S. J. Vol.6. 16.” IEEE Transactions on Power Apparatus and Systems. Tinney. B. April 2527. “Nurturing R&D. No. Hawk. R. D. Arizona. “WSCC Unfolds Causes of the July 2 Disturbance. 2025. pp. 7. No. “Unlocking the GRID. 8. pp. “Lessons from the UK and Norway. W. I. No. 3844. D. F. D. March 1995. No. Vol. 33. August 1996. Wollenberg. 33. pp. 4549. 1996.” IEEE Spectrum. Tabors. Vol.517.” IEEE Spectrum. 5. 39. 28642880. August 1962. Vol. 33. Brewer. Rudnick. 33. PAS103. Vol. Talukdar.” IEEE Spectrum. 50.” IEEE Power Engineering Review. The joint PSERC/EPRI workshop on underlying technical issues in electricity deregulation. Tucson.” IEEE Spectrum.33. pp. Vol. F. pp. [8] [9] [10] [11] [12] [13] 90 . PAS87. R. p.” IEEE Spectrum. “Optimal Power Flow by Newton Approach. Masiello. No.” Docket No.. 33. “Promoting wholesale competition through open access nondiscriminatory transmission services by public utilities.51. July 1996. August 1996. Dommel and W. “It’s put up or shut up for grid controls. B. H. Carpienter. “Contribution e l’étude do Dispatching Economique.” Bulletin Society Française Electriciens. 1996. 7. A. 32. 8. Morgan and S. August 1996. Vol. 3. Sun. 18661876.” IEEE Transactions on Power Apparatus and Systems. B. Barkovich and D. 7. No. July 1996. pp. pp. Vol.
G.” 1996 IEEE/PES Winter Meeting. J. J. D. “A UserFriendly Simulation Program for Teaching Power System Operations. No. Alsac.. Tinney and C. Sun. May 1991. “Assessing the Value of Reactive Power Services in Electric Power Systems. W. PWRS1. 10. 762770.” IEEE Transactions on Power Systems. No. “Partial Matrix Refactorization. J. T. T. November 1967. 5. Englewood Cliffs. Power System Analysis. Inc. MD. Hart. Weber.” 1993 IEEE Power Industry Computer Applications Conference. No. M. Stevenson. February 1986. M. 193200. D. M. J. 3. E. A. 1979. R. F. Chan and V. Sauer. B.J. Vol. 11. Bond. Sauer. S. 1996. A. “Challenges to Optimal Power Flow. pp. C. “Discrete Shunt Controls in A Newton Optimal Power Flow.” IEEE Transactions on Power Systems. 96 WM 3129 PWRS. 697711. pp. Luenberger. 2. 295392. Vol. Stott.” Conference on Unbundled Power Quality Services. Vol. D. 33. pp. B. Marsten. A. pp. Gross. F. Bright. Chapters 9 and 16. Vol.330. N. E. Key West. November 1992. P. November 1995. Vol. New York. 138145.: PrenticeHall. Gross. “Direct Nonlinear PredictorCorrector PrimalDual Interior Point Algorithm for Optimal Power Flows.” IEEE Transactions on Power Apparatus and Systems.” IEEE Transactions on Power Systems.423450. Florida. Tinney. M. “Further Developments in LPBased Optimal Power Flow. Grainger and W. Galiana. November 1719. D. W. Debs and R. Brandwajn. Computer Solution of Linear Algebraic Systems.A. Reading. No. S. Vol. 1967. M. J. Overbye. NY: John Wiley & Sons. NY: McGraw Hill. pp. Power System Analysis. G. August 1990. Y. Wu. 4. 15191528. J. D. pp. Linear and NonLinear Programming. D. 7. Moler. E.. Huneault and F. Liu. 1.[14] O. Koessler. 1984. A. PAS86. New York. Marzinzik and G. 1994. Overbye and P. No. 17251733. pp. D. MA: AddisonWesley Publishing Company. J. “A Survey of the Optimal Power Flow Literature. Prais and B. Papalexopoulos and W. 4. 1996. No. W. Papalexopoulos and P.” IEEE Transactions on Power Systems. Inc. [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] 91 . “Power Flow Solution by Newton’s Method. January 2125. S.” IEEE Transactions on Power Systems. Stott. J. Baltimore. J. E. pp. C. pp. M. Momah. Laufenberg. 18661876. Ristanovic. Forsythe and C. 6.
pp. Shirmohammadi.[27] R. 9. D. No. 206215. X. 11. D.V. NY: North Holland Publishing. Chattergy. D. No. Pereira. “Transmission System Management and Pricing: New Paradigms and International Comparisons. February 1994. “Some fundamental technical concepts about cost based transmission pricing. 1978. Vol. New York. B.” IEEE Transactions on Power Systems. 2. Tabors. [28] [29] 92 . Introduction to NonLinear Optimization. 10021008. May 1996. Filho. Gorenstin and M. 1. A. pp.” IEEE Transactions on Power Systems. A Problem Solving Approach. Vol. Wismer and R.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.