You are on page 1of 6

1

Leonardo Nepomuceno Member, IEEE, Aurélio R. L. Oliveira Member, IEEE,

Incorporating Voltage/Reactive Representation Takaaki Ohishi Member, IEEE, Secundino Soares Filho

to Short-Term Generation Scheduling Models



reactive power in STGS problems are discussed in [1], and
Abstract—This paper proposes a methodology to incorporate the quest for new models is also depicted.
voltage/reactive representation to Short Term Generation Computational effort is greatly enlarged when
Scheduling (STGS) models, which is based on active/reactive voltage/reactive constraints are incorporated to STGS
decoupling characteristics of power systems. In such approach
models. Thus, most STGS models proposed in the literature
STGS is decoupled in both Active (AGS) and Reactive (RGS)
Generation Scheduling models. AGS model establishes an initial totally ignore such reactive constraints or adopt some kind of
active generation scheduling through a traditional dispatch simplification in the representation of the transmission
model. The scheduling proposed by AGS model is evaluated system. In the classic economic dispatch [2] transmission
from the voltage/reactive points of view, through the proposed system is represented as a single bus. In the works [3]-[5] the
RGS model. RGS is formulated as a sequence of T nonlinear transmission system is represented by network flow or DC
OPF problems, solved separately but taking into account load load flow equations. Some models proposed in the literature
tracking between consecutive time intervals. This approach
represent active and reactive portions of the network. One of
considerably reduces computational effort to perform the
reactive analysis of the RGS problem as a whole. When the first formulation is presented by Bonaert et. all in [6]. In
necessary, RGS model is capable to propose active generation such work the hydroelectric as well as the transmission
redispatches, such that critical reactive problems (in which all system are modeled in detail. In the solution methodology
reactive variables have been insufficient to control the reactive adopted a decomposition is performed in terms of
problems) can be overcome. The formulation and solution hydroelectric and thermoelectric sub-problems. The works
methodology proposed are evaluated in the IEEE30 system in presented by Nanda [7] and Luo [8] deal basically with the
two case studies. These studies show that the methodology is same problem but using different solution methodologies.
robust enough to incorporate reactive aspects to STGS
problem.
Another class of problems trying to improve the
representation of the transmission system in the STGS
Index Terms--Short term Generation Scheduling, Economic problem is Hydrothermal Optimal Power Flow (HOPF)[9],
Dispatch, Generation Dispatch, Optimal Power Flow. which is Optimal Power Flow (OPF) incorporating dynamic
constraints (such as generation targets established by
I. INTRODUCTION medium/long term models). Some other extensions of this

S hort Term Generation Scheduling (STGS) aims to approach have been successfully studied recently [10][11]. A
determine a power production planning for active power large scale HOPF is discussed in [12].
generation of thermal and hydraulic units over a short term This work presents a new approach to the STGS problem
period (typically one day to one week). For dominant hydro aiming to incorporate detailed active and reactive aspects of
systems this short term power production planning is the transmission and generation systems. In the proposed
concerned with optimal discretization of generation targets approach, STGS problem is decoupled in two sub-problems:
proposed by higher level (medium/long term) planning Active (AGS) and Reactive (RGS) Generation Scheduling
models. models. AGS is a traditional dispatch model in which
The optimal generation scheduling provided by STGS reactive aspects are totally neglected. All dynamic
models is a set point for on-line operation. On-line constraints such as ramp rate and generation targets (for
applications are concerned with both security and hydraulic units) are introduced in AGS model. These
optimization of the system. In on-line studies operational constraints greatly enlarge computational effort to solve the
aspects of the electric system are critical, especially those problem, and so, are left to AGS model, which is simpler and
associated with the system security, such as voltage/reactive less time consuming
aspects. Thus, it is highly desirable to include in STGS The incorporation of all voltage/reactive aspects are
models as many aspects associated to the transmission system performed by RGS model here proposed. RGS is a multi-
as possible. Aspects associated with the representation of objective problem and is concerned with minimization of
generation and transmission losses and also with minimum
This research was partially sponsored by FAPESP (Research Supporting
 deviation from a previous dispatch policy established by
Foundation of the State of São Paulo, Brazil) through grant no. 97/00414-5 AGS model solution. RGS is formulated as a sequence of T
and CNPq (Brazilian Council for the Development of Science and
Technology). (number of time intervals) OPF independent problems solved
L. Nepomuceno is with State University of São Paulo – UNESP, Bauru, separately, but taking into account load tracking between
SP, Brazil (e-mail: leo@feb.unesp.br ) consecutive time intervals; This approach considerably
A. R. L. Oliveira is with State University of São Paulo – USP, São Carlos,
SP, Brazil (e-mail: aurelio@densis.fee.unicamp.br ) reduces computational effort to solve RGS model.
T. Ohishi is with State University of Campinas – UNICAMP, Campinas, The solution process for the STGS model here proposed is
SP, Brazil (e-mail: taka@densis.fee.unicamp.br ) based on interaction between AGS and RGS solutions (see
S. Soares Filho is with State University of Campinas – UNICAMP,
Campinas, SP, Brazil (e-mail: dino@densis.fee.unicamp.br ) Figure 1). Initially RGS model seeks for a purely reactive
2

dispatch maintaining unaltered the generation scheduling  F max , F min : minimum and maximum active power
proposed by AGS model. If this reactive dispatch is found to flows ;
be inadequate for some time interval , RGS model is capable  P t : are the active power generation at time interval t;
to propose a new generation scheduling that, for such
interval, would be feasible from both active and reactive  P max , P min :minimum and maximum active generation
aspects. It is not common in the literature to re-program the ;
active generation to cope with reactive problems; those  Lt : are the active power loads at time interval t;
problems are generally solved using reactive controls.  A : is the nodal incidence matrix;
However, the situation evaluated in this paper is a critical  X : is the reactance matrix;
one, in which the use of all reactive controls (calculated by  T : set of time intervals
OPF model) failed to provided a feasible reactive scheduling. Constraints (1.1) represent the generation targets (it is
The re-programming of active generation scheduling to cope written in terms of active power generation instead of
with problems associated with reactive power flows is energy). Constraints (1.2) represent lower and upper active
analyzed in the context of deregulated electric systems in power flow limits in all system branches. Constraints (1.3)
[14]. If RGS proposes a new dispatch for some time interval, represent lower and upper active power generation limits for
the dynamic constraints, which are completely neglected in all units. Constraints (1.4) and (1.5) represent load flow
RGS model, may be violated. Thus, it is necessary to equation. The criterion adopted is the loss minimization for
redispatch all other time intervals to guarantee that these hydraulic generation and the transmission system. Such
constraints will not be violated; this task is once more criterion was proposed in [13].
performed by AGS model. Thus, in the methodology The solution methodology adopted in this paper to solve
proposed here, the solution to the STGS problem is obtained AGS is based on interior point methods. This methodology is
beyond the scope of this work. This paper focuses on the
by interacting the solutions obtained by AGS and RGS
reactive aspects of the STGS problem and on its inter-
models. The results presented in this work show that
relation with purely active power dispatch models.
voltage/reactive aspects can easily be incorporated to
traditional purely active STGS models through the decoupled III. REACTIVE GENERATION SCHEDULING (RGS) MODEL
approach proposed.
This work is organized as follows. In section II a typical RGS model is concerned with the evaluation of reactive
AGS model is described. In section III the proposed RGS capabilities for generation and transmission systems. The
model is formulated. Section IV describes the proposed prime objective of such model is to determine whether the
active power dispatch fixed by AGS models is adequate from
solution technique for RGS model. The decoupled solution
the reactive power perspective. The first task for RGS model
methodology, involving AGS and RGS is described in
is to fix the generation scheduling proposed by AGS models
section V. Section VI presents numerical results and final
and calculate a daily optimal reactive dispatch for all reactive
conclusions are depicted in section VII. variables (voltages on controllable buses, transformer taps,
reactive generation, etc.). For a given period of the day, this
II. ACTIVE GENERATION SCHEDULING PROBLEM task may be performed by Optimal Power Flow (OPF)
A typical Active Generation (AGS) problem is concerned models. If the dispatch proposed by the AGS model is found
with thermal and hydraulic generations and their respective to be inadequate, the second task for RGS model is to seek
operational constraints. The transmission system for a new daily generation dispatch which would be feasible
representation is generally simplified so that only the active with respect to both active and reactive power flows in the
generation aspects are taken into account. A simple network transmission and generation systems. This second task may
flow based AGS model is adopted in this work. As already also be performed through an OPF model in which the active
discussed, the focus of this paper is on voltage/reactive power for all generating units are delivered. RGS model is
aspects associated with transmission system, which is thus generically formulated as T (number of time intervals of
detailed in RGS model (in the next section). It is worth the STGS problem) OPF problems, as follows.
noting that the proposed active reactive methodology is
Min   RGS  x t 
independent on specific AGS formulation. The typical AGS 

s.t . :
tT


gt xt 0 t T ( 2.1)
RGS 
here adopted is formulated as follows: 
 x min  x t  x max t T ( 2.2)

 
Qimin  hit x t Q max
i i   gen t T ( 2.3)

Mi n

  f t
, P t

tT

where:
 s.t . :

  Pi  M i i   hyd (1.1)
 tT
AGS 
 F min
 f t
 F max
t T (1.2)

x t   V,Θ, Tp, P, Bsh t :


 min t max
 P  P  P t T (1.3)

 is the vector of
 Af t
 P t
 Lt t T (1.4)
 t

 Xf  0 t T (1.5)

optimization variables at time interval t.


where:
 V : vector of voltage magnitudes;
 Pi t : active power generation of unit i at time interval t;  Θ : vector of voltage angles;
 M i : fixed generation target for hydraulic unit i;  Tp : vector of transformer taps;
  hyd : set of hydraulic units;  P : vector of active power generation for all units;
 Bsh : vector of shunt devices (capacitors and reactors);
 f t : are the active power flows at time interval t;
3

  
hit x t reactive power generation of unit i at time
Constraints in optimization variables are handled through


interval t
 gen : set of all generating units 
penalty functions Pen x min , x, x max , given by equation
(6).
 x min , x max : vectors of lower and upper limits for x.
 Qimin , Qimax : minimum and maximum reactive power 
Pen k x min , x, x max  2
xj xj 
  k 1* lim 2
(6) 
j xviol
generation at bus i
RGS is a multi-objective problem.  RGS x t is a linear    xviol : set of violated variables in the solution to sub-

combination of the following criteria: nonlinear transmission problem MOPF k -1


lim
system losses, quadratic generation losses for hydraulic units  x j is the upper or lower limit in variable x j
  is a weighting factor
[13], quadratic generation costs for thermal units and also a
quadratic weighted deviation on active power generation
(hydro and thermal units) from specified values. In each time The initial subproblem MOPF 0 is relaxed with respect
interval AC load flow equations for the transmission system to functional constraints and is described in (7)
are represented by (2.1). The limits on variables are
established by equation 2.2 in each time interval. The  Min

C ( x )  Pen k x min
, x, x max

MOPF 0  s.t . :

nonlinear functional constraint 2.3 establishes the limits on 



g ( x)  0

reactive power generation for all reactive sources. (7)


It is worth noting that RGS model does not present any The solution to the modified problem given by (4)
dynamic coupling. This means that the objective function is constitutes a series of parameterized problems starting from
additive separable in time domain. Thus, RGS problem an arbitrary point. In each iteration infeasible reactive
described above can be formulated as a sequence of T generation constraints are incorporated to the problem
independent OPF problems. If the power generation is fixed through homotopy function and infeasible constraints in
at a certain interval, only a reactive dispatch is calculated variables are incorporated through quadratic penalty
(first task). If the generation is delivered, an active/reactive functions. The parameterized problems are solved varying
dispatch is calculated (second task). the parameter  from 0.0 to 1.0. This is equivalent to follow
IV. SOLUTION TECHNIQUE FOR RGS MODEL the path defined by the points x t , such that:  
RGS model is composed of T independent OPF problems. H k ( x,  )  0
Although these OPF problems are independent, RGS is For a given  the Lagrangian function of the problem is
solved as a unique problem. A hybrid solution technique is written as:
adopted to solve each OPF model. In the OPF solution
process, constraints in variables are treated through penalty
 
L x   C ( x )  Pen k x min , x, x max  λ g g  x   λ h H k ( x,  )
functions [15][16] and functional constraints associated with  λ is the vector of Lagrange multipliers associated with
g
reactive power generation are treated using parameterizations equality constraints (4.1)
techniques [17]. Thus, for a specific time interval t one OPF
problem is solved. The OPF problem is solved through a Medium Term
basic Newton algorithm applied to a sequence of modified Model
OPF problems ( MOPF k k  1,2  ) This sequence of Generation Targets
problems is defined by (4) as follows.
 Min  x min max

AGS
C ( x) Pen k , x, x

MO P F k
   s.t . :

 g (x)  0 (4.1)

 H k ( x,  )  0 (4.2)

(4)
reprogramming Generation program
In (4) inequality constraints are handled through
parameterization techniques, where: RGS
lim
H k (x,  )  h 
 1    h x  k 1 *
(5) 
 k 1 * is the solution to sub-problem MOPF k -1
DECOUPLED Active/reactive generation
 x MODEL dispatch
   are the reactive generation functions calculated
h x  k 1 *
for the set of infeasible functional constraints detected in
the solution for sub-problem MOPF k -1 On-line Operation
  is a parameter varying from 0.0 to 1.0 Figure 1- Decoupled AGS/RGS Solution
 h lim are the reactive generation limits (maximum or
minimum values) for infeasible functional constraints  λ h is the vector of Lagrange multipliers associated with
(detected in the solution for sub-problem MOPF k -1 ) equality constraints (4.2)
Necessary optimality conditions correspond to the
4

stationary point to the Lagrangian function. The first order study, the daily load profile is less heavy. In the second study
necessary conditions for the sequence of problems is solved such load profile is made heavier such that critical load
by the Newton method. General algorithm is detailed in [17]. conditions can be studied. The basic objectives of these
studies are to evaluate the coupling details between active
V. DECOUPLED SOLUTION METHODOLOGY APPLIED TO STGS and reactive
MODEL dispatch, to analyze the adequacy of the adopted
decomposition, and also to highlight the importance of the
In the approach proposed STGS problem is solved
reactive aspects in STGS models.
interacting the solutions provided by AGS and RGS models
In the first case, a less heavy daily profile is adopted.
as shown is figure 1. The basic algorithm for such approach
Following the solution methodology algorithm described in
is summarized on the following steps:
section VI an initial AGS study was performed. The active
generation dispatch provided by this study is presented in the
1. AGS is solved taking into account generation targets.
column "Generation MW" of Table 1. The generation targets
The solution of such model provides an initial generation
for each unit are also depicted in the bottom of each
scheduling: active generation dispatches are calculated
generation column.
for each thermal and hydraulic unit at each time interval. TABLE I
2. The active generation scheduling calculated in 1. is fixed SOLUTION TO AGS/RGS – CASE I
and RGS model is solved so that a purely reactive
dispatch is calculated for each time interval. If this
reactive dispatch is feasible (no active or reactive
constraint is violated in any time interval) then stop. This
is a feasible solution to STGS; If otherwise the dispatch
is unfeasible in any time interval go to 3.
3. A new active generation scheduling must be set to the
time intervals presenting unfeasible dispatches
(calculated in item 2.). This scheduling is obtained
releasing the active power generation in the RGS model
solution. In other words, RGS model tries to calculate a
new active/reactive generation dispatch such that a
feasible solution is obtained. This procedure is repeated
for all time intervals presenting unfeasible solutions.
These critical time intervals are generally the ones
presenting heavy load conditions (voltage levels tend to 30
become down). Go to 4. If RGS model is not capable to 25
Newton Iteration Number

obtain such feasible solution, the methodology can not


20
obtain a solution to STGS problem. Stop.
4. The rescheduling calculated in 3. by RGS model (for 15
certain time intervals) may have violated dynamic 10
constraints (generation target or ramp rate constraint).
5
AGS model is once more used to deal with these
constraints. In this solution process the active generation 0
of all critical time intervals re-programmed in step 3. are -5
0 5 10 15 20 25 30
fixed and the active generation of all other time intervals Interval
are re-programed, so that the dynamic constraints are
satisfied. After this rescheduling we get back to step 2. to
Figure 2 – RGS Evolution
obtain a new reactive power dispatch and the process is
repeated.
Note that by the solution methodology proposed the
voltage/reactive evaluation is introduced in STGS only in
specific critical time intervals. For such intervals the
generation dispatch is performed in a coupled way using
RGS model as an active/reactive power dispatch tool. This
approach reduces computational effort to solve the whole
STGS problem.

VI. NUMERICAL RESULTS


The proposed methodology was applied to IEEE 30 bus
system. A hydroelectric generation system has been adapted
to such system. Two case studies are presented as follows
differing only by the daily load profile adopted. In the first
5

Violations Binding Constraints Generation MW as the starting point to the OPF problem for the next time
t Mw V C Tp Mw V C Tp Mvar Gen1 Gen2 Gen5 Gen8 Gen11 Gen13
1 0 0 0 0 0 0 2 0 0 26,19 41,33 49,6 44,59 46,95 28,06 interval (2:00) the number of iteration is greatly reduced. The
2 0 0 0 0 0 0 1 0 0 25,86 40,64 47,8 43,48 45,83 27,39 following intervals are consecutively adjusted in the same
3 0 0 0 0 0 0 1 0 0 24,53 37,85 40,6 39,08 41,35 24,69
4 0 0 0 0 0 0 1 0 0 23,2 35,07 33,4 34,71 36,87 21,99 manner such that the overall time to solve RGS model is
5 0 0 0 0 0 0 2 0 0 26,19 41,33 49,6 44,59 46,95 28,06 reduced considerably. It can be noticed that time intervals
6 0 0 0 0 0 0 2 0 1 29,3 47,96 66,72 55,16 57,63 34,42
7 0 0 0 0 0 0 2 0 1 30,98 51,45 75,74 60,79 63,29 37,73 18:00, 19:00 and 20:00 required higher number of iteration.
8 0 0 0 0 0 0 2 0 1 32,02 53,55 81,15 64,19 66,7 39,7 In such time intervals the variation of the solution points
9 0 0 0 0 0 0 2 0 2 32,2 53,9 82,06 64,75 67,27 40,03
10 0 0 0 0 0 0 2 1 3 32,56 54,6 83,86 65,89 68,41 40,69 from one time interval to the other is greater due to active set
11 0 0 0 0 0 0 2 1 3 32,75 54,95 84,77 66,46 68,98 41,01 changes.
12 0 0 0 0 0 0 2 1 3 32,94 55,3 85,67 67,03 69,55 41,34
13 0 0 0 0 0 0 2 1 3 33,13 55,65 86,57 67,6 70,12 41,67
14 0 0 0 0 0 0 2 1 3 33,52 56,34 88,38 68,74 71,27 42,32 TABLE II-SOLUTION TO AGS/RGS (REACTIVE EVALUATION) - CASE II
15 0 0 0 0 0 0 2 1 4 33,7 56,69 89,28 69,31 71,84 42,65 Violation Binding Const. Generation (MW)
16 0 0 0 0 0 0 2 1 4 34,23 57,74 91,99 71,02 73,56 43,62 t Mw V C Tp Mw V C Tp Mvar Gen1 Gen2 Gen5 Gen8 Gen11 Gen13
17 0 0 0 0 0 0 2 1 4 34,61 58,44 93,8 72,17 74,71 44,27 1 0 0 0 0 0 0 2 0 0 46.13 34.58 18.77 51.02 40.00 47.60
18 0 0 0 0 0 0 2 1 5 35,47 59,84 97,42 74,46 77,02 45,57 2 0 0 0 0 0 0 2 0 0 45.80 33.88 16.95 49.91 38.87 46.97
19 0 0 0 0 0 0 2 1 5 35,71 60,19 98,32 75,04 77,6 45,89 3 0 0 0 0 0 0 1 0 0 44.43 31.06 9.68 45.46 34.39 44.44
4 0 0 0 0 0 0 1 0 0 42.83 27.56 4.00 40.83 29.55 41.71
20 0 0 0 0 0 0 2 1 4 34,05 57,4 91,09 70,46 72,98 43,3
5 0 0 0 0 0 0 2 0 0 46.13 34.58 18.77 51.02 40.00 47.60
21 0 0 0 0 0 0 2 1 3 33,13 55,65 86,57 67,6 70,12 41,67 6 0 0 0 0 0 0 2 0 1 49.34 41.29 36.07 61.73 50.71 53.48
22 0 0 0 0 0 0 2 0 1 30,63 50,75 73,93 59,66 62,16 37,07 7 0 0 0 0 0 0 2 0 2 51.05 44.84 45.19 67.43 56.38 56.52
23 0 0 0 0 0 0 2 0 1 29,3 47,96 66,72 55,16 57,63 34,42 8 0 0 0 0 0 0 2 0 2 52.12 46.97 50.67 70.87 59.81 58.33
24 0 0 0 0 0 0 2 0 0 27,18 43,42 55 47,91 50,32 30,08 9 0 0 0 0 0 0 2 0 2 52.30 47.32 51.58 71.45 60.38 58.63
743,37 1228 1800 1449,8 1509,1 897,63 10 0 0 0 0 0 0 2 1 3 52.68 48.03 53.41 72.60 61.52 59.22
11 0 0 0 0 0 0 2 1 3 53.07 48.74 55.24 73.75 62.67 59.82
In the AGS solution these targets are satisfied. Fixing this 12 0 0 0 0 0 0 2 1
13 0 0 0 0 0 0 2 1
3
4
52.88 48.39 54.32 73.18 62.10 59.52
53.25 49.10 56.15 74.33 63.25 60.12
active generation dispatch, a RGS study was performed. As 14 0 0 0 0 0 0 2 1 4 53.60 49.81 57.98 75.49 64.40 60.71
15 0 0 0 0 0 0 2 1 4 54.16 50.87 60.72 77.23 66.13 61.60
the generation dispatch is fixed, only a reactive dispatch is 16 0 0 0 0 0 0 2 1 4 54.74 51.94 63.47 78.97 67.86 62.48
provided by the RGS model. This reactive study is also 17 0 0 0 0 0 1 2 1 4 55.40 53.01 66.21 80.71 69.60 63.36
18 0 1 0 0 0 3 2 3 5 56.81 54.79 70.79 83.64 72.51 64.81
depicted in Table 1. The violated constraints on the solution 19 0 4 2 3 0 4 2 3 5 58.23 56.62 75.50 86.69 75.44 66.24
20 0 0 0 0 0 0 2 1 4 54.16 50.87 60.72 77.23 66.13 61.60
provided by RGS model are shown in columns "MW" (active 21 0 0 0 0 0 0 2 1 4 53.25 49.10 56.15 74.33 63.25 60.12
generation), "V" (voltage magnitude), "C" (capacitors and 22 0 0 0 0 0 0 2 0 2 50.71 44.13 43.37 66.29 55.25 55.92
23 0 0 0 0 0 0 2 0 1 49.34 41.29 36.07 61.73 50.71 53.48
reactor banks) and "TP" (transformer taps). Those violations 24 0 0 0 0 0 0 2 0 0 47.15 36.70 24.23 54.39 43.37 49.47
are shown for each time interval "t". The same symbols 1229.56 1075.48 1086.01 1620.26 1354.24 1353.74

above are used to show the number of binding constraints.


TABLE III – SOLUTION TO RGS (WITH REDISPATCHES) – CASE II
The reactive generation binding constraints given by the Violation Binding Constr. Generation (MW)
column "Mvar" are also shown in the Table. As is shown is 1
t Mw V C Tp
0 0 0 0
Mw V C Tp Mvar
0 0 2 0 0
Gen1
46.13
Gen2
34.58
Gen5
18.77
Gen8
51.02
Gen11 Gen13
40.00 47.60
Table 1. the active generation dispatch has been evaluated by 2 0 0 0 0 0 0 2 0 0 45.80 33.88 16.95 49.91 38.87 46.97
3 0 0 0 0 0 0 1 0 0 44.43 31.06 9.68 45.46 34.39 44.44
RGS model and no active or reactive limit is violated. From 4 0 0 0 0 0 0 1 0 0 42.83 27.56 4.00 40.83 29.55 41.71
the proposed methodology this initial generation dispatch is a 5 0 0 0 0 0 0 2 0 0 46.13 34.58 18.77 51.02 40.00 47.60
6 0 0 0 0 0 0 2 0 1 49.34 41.29 36.07 61.73 50.71 53.48
feasible final solution to the STGS model. Thus the active 7 0 0 0 0 0 0 2 0 2 51.05 44.84 45.19 67.43 56.38 56.52
dispatch provided by Table 1. are the final dispatches to be 8 0 0 0 0 0 0 2 0 2 52.12 46.97 50.67 70.87 59.81 58.33
9 0 0 0 0 0 0 2 0 2 52.30 47.32 51.58 71.45 60.38 58.63
implemented in the generating units. In this less heavy load 10 0 0 0 0 0 0 2 1 3 52.68 48.03 53.41 72.60 61.52 59.22
11 0 0 0 0 0 0 2 1 3 53.07 48.74 55.24 73.75 62.67 59.82
condition there has not been the need to establish iteration 12 0 0 0 0 0 0 2 1 3 52.88 48.39 54.32 73.18 62.10 59.52
between AGS and RGS models. 13 0 0 0 0 0 0 2 1 4 53.25 49.10 56.15 74.33 63.25 60.12
14 0 0 0 0 0 0 2 1 4 53.60 49.81 57.98 75.49 64.40 60.71
In the second case study the load profile was made 15 0 0 0 0 0 0 2 1 4 54.16 50.87 60.72 77.23 66.13 61.60
heavier. Following the proposed methodology algorithm an 16 0 0 0 0 0 0 2 1 4 54.74 51.94 63.47 78.97 67.86 62.48
17 0 0 0 0 0 1 2 1 4 55.40 53.01 66.21 80.71 69.60 63.36
initial AGS study was performed. The active generation 18 0 0 0 0 0 1 2 1 4 13.89 48.23 127.37 111.21 61.62 38.88
19 0 0 0 0 2 4 2 3 4 3.99 4.00 154.16 177.74 40.26 37.71
dispatch is shown in the columns "Generation MW" for each 20 0 0 0 0 0 0 2 1 4 54.16 50.87 60.72 77.23 66.13 61.60
unit in Table 2. Fixing this active generation dispatch, a RGS 21 0 0 0 0 0 0 2 1 4 53.25 49.10 56.15 74.33 63.25 60.12
22 0 0 0 0 0 0 2 0 2 50.71 44.13 43.37 66.29 55.25 55.92
study was performed. As the generation dispatch is fixed, 23 0 0 0 0 0 0 2 0 1 49.34 41.29 36.07 61.73 50.71 53.48
only a reactive dispatch is provided by the RGS model. The 24 0 0 0 0 0 0 2 0 0 47.15 36.70 24.23 54.39 43.37 49.47
1132.41 1016.29 1221.25 1738.89 1308.19 1299.27
solution to the RGS model is also provided in Table 2. The
symbols used in Table 2. are the same used in Table I. It is Once that the RGS solution present some violations in
clear from the table that some voltage magnitude and tap critical time intervals (18:00 and 19:00)it is necessary to
transformer values are violated in time intervals 18:00 and establish a new active dispatch for such intervals (step 3).
19:00. Thus, the active generation is released and a new RGS is
As described in section IV, the RGS solution process was performed for these time intervals. The solution of such
implemented so that the solution for a certain time interval is process is shown in Table 3. The RGS model has calculated a
used as a hot-start for the solution process of the next time new generation scheduling for these time intervals. It is clear
interval. The impact of such strategy is clearly shown in that in these new scheduling the active and reactive violation
Figure 2 as follows. In this figure the number of iterations problems are overcome. However, the generation targets
necessary to solve RGS model in each time interval is established are unbalanced because of the redispatches in
depicted. It is noticeable that the first time interval (1:00) has time intervals 18:00 and 19:00. It is also worth noting that
taken 27 iterations to provided an active/reactive generation these redispatches proposed by RGS model have
dispatch. Using the solution obtained for such time interval considerably altered the initial dispatch provided by AGS
6

model. As the generation targets are violated it is necessary The work also proposes a methodology that reduces
to re-program all other generation scheduling so as to considerably computational effort to solve RGS model. The
distribute the variations on active power generation produced approach can be easily implemented in dispatch centers
by the new scheduling established for time intervals 18:00 having a traditional (active) dispatch tool and an OPF tool.
and 19:00. The approach was applied to a test system in two case
To compensate these deviations in generation targets an studies. Results show that the methodology is robust enough
AGS study is again performed. As described in section V, the to calculate active/reactive generation dispatches that are
active dispatch in time intervals 18:00 and 19:00 are fixed feasible with respect to the most important operational
and all other time intervals are re-dispatched (step 4). AGS aspects associated with generation and transmission for
model solution is shown in Table 4. Comparing the hydrothermal systems.
generation dispatches provided by Tables 3 and 4 it is worth
noting that the deviations in generation targets are distributed VIII. REFERENCES
along the time intervals of the day so that the generation [1] Chattopadhyay, D. "Daily Generation Scheduling: Quest for New
scheduling of Tables 3 and 4 are similar. Models", IEEE Ttrans. On Power Syst., vol 13 no. 2, pp 624 629, May
1998.
TABLE IV- FINAL SOLUTION AGS/RGS - CASE II [2] H. H. Happ, “Optimal power dispatch - a comprehensive survey”, IEEE
Violation Binding Constraints Generation (MW) Trans. Power Apparatus and Syst., vol PAS-96, n. 3, pp 841-854, 1977.
t Mw V C Tp Mw V C Tp Mvar Gen1 Gen2 Gen5 Gen8 Gen11 Gen13 [3] Ohishi, T. and Soares, S. and Carvalho, M. F. H. “A Short-Term
1 0 0 0 0 0 0 2 0 0 50.52 37.49 12.31 45.57 42.33 50.27 Hydrothermal Scheduling Approach for Dominantly Hydro Systems”,
2 0 0 0 0 0 0 2 0 0 50.18 36.78 10.49 44.45 41.21 49.64 IEEE Trans on Power Syst. vol. 6,. n.2, May, 1991.
3 0 0 0 0 0 0 1 0 0 48.69 33.62 4.00 39.88 36.55 47.06
4 0 0 0 0 1 0 1 0 0 46.25 27.65 4.00 34.46 30.46 43.74
[4] Ohishi, T. and Soares, S. "Hydro-Dominated Short-Term Hydrothermal
5 0 0 0 0 0 0 2 0 0 50.52 37.49 12.31 45.57 42.33 50.27 Scheduling Via a Hybrid Simulation – Optimisation Approach: a Case
6 0 0 0 0 0 0 2 0 1 53.74 44.22 29.65 56.31 53.01 56.08 Study", IEE Proc. – C, vol. 142, no. 6, pp 569-575, November, 1995.
7 0 0 0 0 0 0 2 0 2 55.45 47.78 38.79 62.03 58.68 59.08 [5] Irissari, G. and Kimball, L. M. and Clements, K. A. and Bagchi, A. and
8 0 0 0 0 0 0 2 0 2 56.53 49.91 44.28 65.49 62.09 60.86
Davis, P. W., "Economic dispatch with network and ramping constraints
9 0 0 0 0 0 0 2 0 2 56.72 50.27 45.19 66.07 62.66 61.15
10 0 0 0 0 0 0 2 1 3 57.10 50.98 47.02 67.23 63.80 61.74 via interior point methods", IEEE Trans on Power Syst. vol. 13,. n.1,
11 0 0 0 0 0 0 2 1 4 57.48 51.69 48.86 68.38 64.95 62.33 236-242, 1998.
12 0 0 0 0 0 0 2 1 3 57.29 51.34 47.94 67.80 64.37 62.03 [6] Bonaert, A. P., and El-Abiad, A.H., Koivo and A. J. “Optimal
13 0 0 0 0 0 0 2 1 4 57.65 52.05 49.77 68.96 65.52 62.62 Scheduling of Hydrothermal Power Systems”, AIEE Trans. on Power
14 0 0 0 0 0 0 2 1 4 58.00 52.76 51.60 70.13 66.67 63.20
15 0 0 0 0 0 0 2 1 4 58.56 53.83 54.35 71.87 68.39 64.08
Apparatus and System, vol. PAS-91, pp.263-271, 1971.
16 0 0 0 0 0 0 2 1 4 59.16 54.90 57.10 73.62 70.12 64.95 [7] Nanda, J. and Bijwe, P. R. “Optimal Hydrothermal Scheduling With
17 0 0 0 0 0 2 2 1 5 59.83 55.97 59.85 75.37 71.85 65.81 Cascaded Plants Using Progressive Optimality Algorithm”, IEEE Trans
18 0 0 0 0 0 1 2 1 4 13.89 48.23 127.37 111.21 61.62 38.88 on Power Apparatus and Systems, Vol. PAS-100, n. 4,2093-2099, 1981.
19 0 0 0 0 0 4 2 3 4 3.96 4.00 154.17 177.75 40.27 37.72 [8] Luo, G. X.; Habibollahzadeh, H. and Semlyen, A. “Short-Term Hydro-
20 0 0 0 0 0 0 2 1 4 58.56 53.83 54.35 71.87 68.39 64.08
21 0 0 0 0 0 0 2 1 4 57.65 52.05 49.77 68.96 65.52 62.62
Thermal Dispatch Detailed Model and Solutions”, IEEE/PES Winter
22 0 0 0 0 0 0 2 0 2 55.10 47.06 36.96 60.89 57.54 58.48 Meeting, WM 169-4 PWRS, New York, 1989.
23 0 0 0 0 0 0 2 0 1 53.74 44.22 29.65 56.31 53.01 56.08 [9] El-Hawary, M. E. and Tsang D. H. “The Hydrothermal Optimal Load
24 0 0 0 0 0 0 2 0 0 51.54 39.61 17.78 48.94 45.70 52.12 Flow, a Practical Formulation and Solution Techniques Using Newton’s
1228.09 1077.75 1087.58 1619.13 1357.04 1354.89
Approach”, IEEE trans. On Power Syst., vol. PWRS-1, no. 3, pp. 157-
167, August, 1986.
This new generation profile was again submit to a reactive [10] Habibollahzadeh, H. ; Luo, G. X. ; Semlyen A. "Hydrothermal Optimal
Power Flow Based on a Combined Linear and Non-linear Programming
evaluation through the RGS model solution. The solution of Methodology", IEEE/PES Summer Meeting , 1988.
such study is also synthesized in Table 4. In this final [11] Mbamalu, G. A. N., El Hawary , F. , El-Hawary, M. E , "Effects of load
solution, Modeling on Minimum Loss, Mininmum Emission, and Multi-Objective
obtained after iterative process between AGS and RGS Optimal Hydrothermal Power Flow", Electric Power Systems Research,
34 (1995) pp 97-108.
models solution, there is no active/reactive violation. This [12] Wei, H. , Sasaki, H. , Kubokawa, J. and Yokoyama, R. "Large Scale
generation scheduling is a feasible solution to the problem Hydrothermal Optimal Power Problems Based on Interior Point
STGS. Nonlinear Programming", IEEE Trans. on Power Syst., vol. 15, no. 1,
February, 2000.
[13] Soares, S. and Salmazo, C. T. “Minimum Loss Predispatch Model for
VII. CONCLUSIONS Hydroelectric Power Systems” , IEEE Trans. On Power Syst., vol. 12,
no. 3, pp. 1220 – 1228, August, 1997.
This paper proposes a new approach to the Short Term [14] Bell, K. R. W. and Kirschen, D. S. “Improved Sensitivities for MW
Generation Scheduling (STGS) problem in which the active Dispatch for Control of Voltage” , International Conference on Power
and reactive aspects of the transmission systems are Industry Computer Applications – PICA 1999.
[15] Sun, D.I.; Ashley B.; Brewer, B.; Hugues, A. and Tinney , W. F.
represented in detail. The approach is based on an decoupling "Optimal Power Flow by Newton Approach" , IEEE Trans. on Power
of STGS problem into the Active Generation Scheduling Apparatus and Syst., vol. PAS 103, no. 10, October, 1984.
(AGS) and the Reactive Generation Scheduling (RGS) [16] Sun,, D. I. B.; Hugues, A.; Tinney, W. F.; Bright, J. M. and Lamont, J.
"Optimal Power Flow Solution by Newton's Method", IEEE Tutorial
problems. Such decomposition scheme takes advantage of Course Reactive Problems: Basics, Problems and Solutions,35-48,1987.
two basic aspects: the active-reactive decoupling [17] Nepomuceno, L. and Santos Jr., A. Equivalent Optimization Model For
characteristics of power system; and the incorporation of Loss Minimization: Suitable Analysis Approach ; IEEE Trans. On
dynamic constraints (that enlarges computational effort) only Power Syst., vol. 12, no. 4, pp. 1403 – 1412, November, 1997.
into AGS model, which is computationally less time
consuming. One of the most important characteristics of the
approach proposed is its ability to take into account
voltage/reactive problems. Such problems are pointed out in
the literature as responsible for on-line systems instabilities.

You might also like