POWER FLOW ANALYSIS

 Power flow analysis assumption

 steady-state
 balanced single-phase network
 network may contain hundreds of nodes and
branches with impedance X specified in per unit on
MVA base
 Power flow equations
 bus admittance matrix of node-voltage equation is
formulated
 currents can be expressed in terms of voltages
 resulting equation can be in terms of power in MW

Energy Conversion Lab

 BUS ADMITTANCE MATRIX

 Nodal solution
 nodal solution is based on
the Kirchhoff’s current law
 impedance is converted to
admittance
1 1
yij = =
Z ij rij + jxij
 Bus admittance equations
 the impedance diagram:
see Fig.6.1

Energy Conversion Lab

 BUS ADMITTANCE MATRIX

 Bus admittance
equations
 the admittance is
based on bus-to-
bus: see Fig.6.2
 if no connection
between bus-to-
bus, leave as zero
 node voltage
equation is in the
form
I bus = YbusVbus
Energy Conversion Lab
 BUS ADMITTANCE MATRIX
 Node-voltage matrix
 Ibus=YbusVbus

 Y11 Y12  Y1i  Y1n 

 I1    V1 
 I   Y21 Y22  Y2i  Y2n  V 
 2          2 
  =    
   Yi1 Yi2  Yii  Yin   
 I i          Vi 
 I n    Vn 
 Yn1 Yn2  Yni  Ynn 
 Ibus is the vector of injected currents
 Vbus is the vector of the bus voltage from reference
node
 Ybus is the bus admittance matrix
Energy Conversion Lab
 BUS ADMITTANCE MATRIX
 Node-voltage matrix
 diagonal element Yii: sum of admittance connected to bus i
n
Yii = ∑ yij j≠i
j =0

 off-diagonal matrix Yij: negative of admittance between nodes I

and j
Yij = Y ji = − yij
 when the bus currents are known, bus voltages are unknown,
bus voltage can be solved as
−1
Vbus = Ybus I bus
 inverse of bus admittance matrix is known as impedance matrix
Zbus −1
Z bus = Ybus

 if matrix of Ybus is invertible, Ybus should be non-singular

Energy Conversion Lab
 BUS ADMITTANCE MATRIX
 Node-voltage matrix
 admittance matrix is symmetric along the leading diagonal,
which result in an upper diagonal nodal admittance matrix
 a typical power system network, each bus is connected by a few
nearby bus, which cause many off-diagonal elements are zero
 many zero off-diagonal matrix is called sparse matrix
 the bus admittance matrix in Fig.(6.2) by inspection is

− j8.5 j 2.5 j5.0 0 

 j 2.5 − j8.75 j5.0 0 
Ybus = 
 j5.0 j5.0 -j 22.5 j12.5 
 
 0 0 j12.5 − j12.5 

Energy Conversion Lab

 SOLUTION OF NONLINEAR ALGEBRA EQUATIONS
 Techniques for iterative solution of non-linear
equations
 Gauss-Seidal
 Newton-Raphson
 Quasi-Newton
 Gause-Seidal method
 consider a nonlinear equation f(x)=0
 rearrange f(x) so that x=g(x), f(x)=x-g(x) or
f(x)=g(x)-x
 guess an initial estimate of x = x(k)
 use iteration, obtain next x value as x(k+1) = g(x(k))
 criteria for stop iteration: |x(k+1)-x(k)| ≤ ε
 ε is the desired accuracy
Energy Conversion Lab
 GAUSE-SEIDAL METHOD
 Nature of Gause-Seidal method
 see Ex.(6.2) and Fig.(6.3)
 Gause-Seidal method needs many iterations to
achieve desired accuracy
 no guarantee for the convergence, depend on the
location of initial x estimate

Energy Conversion Lab

 GAUSE-SEIDAL METHOD
 Nature of Gause-Seidal method
 solution: if initial estimate x is within convergent
region, solution will converge in zigzag fashion to one
of the roots
 no solution: if initial estimate x is outside convergent
region, process will diverge, no solution found
 in some case, an acceleration factor α is added to
improve the rate of convergence:
 x(k+1) = x(k) +α[g(x(k))-x(k)], where α>1
 acceleration factor should not too large to produce
overshoot
 see Ex.(6.3) for the acceleration factor used
Energy Conversion Lab
 GAUSE-SEIDAL METHOD
 Extend one variable to n variable equations using Gause-
Seidal method
 consider the system of n equations in n variables and solving for
one variable from each equation in one time of iteration
f1 ( x1 , x2 , , xn ) = c1 x1 = c1 + g1 ( x1 , x2 , , xn )
f 2 ( x1 , x2 , , xn ) = c2 x2 = c2 + g 2 ( x1 , x2 , , xn )
 
f n ( x1 , x2 , , xn ) = cn xn = cn + g n ( x1 , x2 , , xn )
 the updated variable x1(k+1) calculated in first equation in
Eq.(6.12) is used in the calculation of x2(k+1) in the second
equation
 Ex: in the 2nd iteration x2(k+1) = c2+g2(x1(k+1)+x2(k)+x3(k)+…+xn(k))
 at n iteration to complete n variables, the x1(k+1),…,xn(k+1) is
tested against x1(k),…,xn(k) for accuracy criterion
Energy Conversion Lab
 POWER FLOW SOLUTION
 Power Flow (Load Flow)
 operating condition: balanced, single phase model
 quantities used in power flow equation are: voltage
magnitude |V|, phase angle δ, real power P, and
reactive power Q
 system bus classification:
 slack bus (swing bus): taken as reference where |V| and
∠V are specified. It makes up the loss between generated
power and scheduled loads
 load bus (PQ bus): P and Q are specified, |V| and ∠V are
unknown
 regulated bus (PV bus): P and |V| are specified, ∠V and Q
are unknown

Energy Conversion Lab

 POWER FLOW EQUATION
 Power flow formulation
 consider bus case in Fig.(6.7)
 current flow into bus i:
n n
I i = Vi ∑ yij − ∑ yijV j j≠i
j =0 j =1

 express Ii in terms of P,Q:

Pi − jQi
Ii =
Vi *
 the power flow equation becomes
Pi − jQi n n

*
= Vi ∑ yij − ∑ yijV j j≠i
Vi j =0 j =1

 the power flow problem results in algebraic nonlinear equations

which must be solved by iteration methods
Energy Conversion Lab
 GAUSS-SEIDEL POWER FLOW EQUATION
 Gauss-Seidel power flow solution
 solving Vi: for PQ bus, assume P,Q are known
Pi sch − jQisch
Vi *( k )
+ ∑ y V
ij k
(k )

Vi ( k +1) = j≠i
∑ yij
 solving Pi: for slack bus, assume V is known
  
 n n

Pi ( k +1) = Re Vi *( k ) Vi ( k ) ∑ yij − ∑ yijV j( k )   j≠i
 
 
j =0 j =1
 
  j ≠ i 
 solving Qi: for PV bus, assume |V| is known
  
 *( k )  ( k ) n n
(k ) 
 i ∑
Qi( k +1) = − Im Vi V yij − ∑ yijV j  j≠i

 
j =0 j =1
 
 j ≠i

Energy Conversion Lab

 GAUSS-SEIDEL POWER FLOW EQUATION
 Instructions for Gauss-Seidel solution
 there are 2(n-1) equations to be solved for n bus
 voltage magnitude of the buses are close to 1pu or
close to the magnitude of the slack bus
 voltage magnitude at load buses is lower than the slack
bus value
 voltage magnitude at generator buses is higher than
the slack bus value
 phase angle of load buses are below the reference
angle
 phase angle of generator buses are above the
reference angle
Energy Conversion Lab
 INSTRUCTIONS FOR G-S SOLUTION
 Instructions for PQ bus solution
 real and reactive power Pisch, Qisch are known
 starting with an initial estimate of voltage using Vi
equation P sch − jQ sch i i
+ ∑ yijV j( k )
( k +1) Vi *( k )
Vi = j≠i
∑ yij
 Instructions for PV bus solution
 Pisch, |Vi| are specified
 assume Vi = |Vi|∠0o, solve the Qi equation as below
  
 n n

Qi( k +1) = − Im Vi *( k ) Vi ( k ) ∑ yij − ∑ yijV j( k )   j≠i
 
 
j =0 j =1
 
 j ≠i
Energy Conversion Lab
 INSTRUCTIONS FOR G-S SOLUTION
 Instructions for PV bus solution
 when Qi(k+1) is available, solve Vi using equation below
Pi sch − jQi( k )
*( k )
+ ∑ yijVk( k )
Vi
Vi ( k +1) = j≠i
∑ yij
 since |Vi| is specified, keep imaginary part of Vi,
calculate real part of Vi
Re{Vi ( k +1) } = − (imag {Vi ( k +1) })
2 2
Vi
 solve Vi
Vi ( k +1) = Re{Vi ( k +1) }+ j Im{Vi ( k +1) }
 stopping criteria
Re{Vi ( k +1) }− Re{Vi ( k ) } ≤ ε , Im{Vi ( k +1) }− Im{Vi ( k ) } ≤ ε
Energy Conversion Lab
 INSTRUCTIONS FOR G-S SOLUTION
 Instructions for PV bus solution
 to accelerate the convergence, using the following
approximation after new Vi is obtained

Vi ( k +1) = Vi ( k ) + α (Vi (cal

k)
− Vi ( k ) )

 α is in the range between 1.3 to 1.7

 voltage accuracy in |Vi| and ∠δ is in the range between
0.00001 to 0.00005

Energy Conversion Lab

 INSTRUCTIONS FOR G-S SOLUTION
 Instructions for V, ∠δ slack bus solution
 solve Pi
  
 *( k )  ( k ) n n
(k ) 
 i ∑
Pi ( k +1) = Re Vi V yij − ∑ yijV j  j≠i

 
j =0 j =1
 
 j ≠i
 solve Qi
  
 *( k )  ( k ) n n
(k ) 
 i ∑
Qi( k +1) = − Im Vi V yij − ∑ yijV j  j≠i

 
j =0 j =1
 
  j ≠ i  

 accuracy: the largest ΔPΔQ is less than the

specified value, typically is about 0.001pu
Energy Conversion Lab
 G-S Power flow Homework
For the one-line diagram shown below, using the G-S method
to determine all bus voltages (magnitude and phase) and
show the power flow solution between the buses assume the
regulated bus (#2) reactive power limits are between 0 and
600Mvar.
 NEWTON RAPHSON METHOD
 Newton Raphson method for solving one variable
 consider the solution of one-dimensional equation f(x)=c
 assume x = x(0)+∆x(0)
 f(x)=f(x(0)+∆x(0))=c
 use Taylor’s series expansion
( 0)
1  d2 f 
( 0)
( 0) ( 0) ( 0)  df 
f ( x + ∆x ) = f ( x ) +   ∆x +  2
( 0)
 (∆x )
( 0) 2
+ = c
 dx  2!  dx 
 assume ∆x(0) is very small, higher order terms of expansion can
be neglected, Taylor series becomes
( 0)
 df 
f (x ( 0)
+ ∆x ) = f ( x ) +   ∆x ( 0 ) = c
( 0) ( 0)

 dx 
 assume f(x(0))=c-∆c(0), the equation becomes ∆c(0)≅(df/dx)(0)∆x(0)
 the new approximation of x x = x +
( 1) ( 0 ) ∆ c ( 0)

( 0)
 df 
 
 dx 
Energy Conversion Lab
 NEWTON RAPHSON METHOD
 Newton Raphson method for solving one variable
 the new approximation of x
∆c ( 0 )
x =x +
(1) ( 0)
( 0)
 df 
 
 dx 
 Newton Raphson algorithm
 ∆c ( k ) = c − f ( x ( k ) )
∆c ( k )
∆x (k )
= (k )
 df 
 
 dx 
x ( k +1) = x ( k ) + ∆x ( k )
 for more information, see Ex.(6.4)
 Newton’s method converges faster than Gauss-Seidal, the root
may converge to a root different from the expected one or
diverge if the starting value is not close enough to the root
Energy Conversion Lab
 NEWTON RAPHSON METHOD FOR n VARIABLES

 Newton Raphson method for solving n variables

( 0) ( 0) ( 0)
 ∂f1   ∂f1   ∂f1 
f1 ( x + ∆x ) = f1 ( x ) +   ∆x1 + 
( 0) ( 0) ( 0) ( 0)
 ∆x2 +  + 
( 0)
 ∆xn ( 0 ) = c1
 ∂x1   ∂x2   ∂xn 
( 0) ( 0) ( 0)
 ∂f 2   ∂f 2   ∂f 2 
f 2 ( x + ∆x ) = f 2 ( x ) +   ∆x1 + 
( 0) ( 0) ( 0) ( 0)
 ∆x2 +  + 
( 0)
 ∆xn ( 0 ) = c2
 ∂x1   ∂x2   ∂xn 

( 0) ( 0) ( 0)
 ∂f   ∂f   ∂f 
f n ( x ( 0 ) + ∆x ( 0 ) ) = f n ( x ( 0 ) ) +  n  ∆x1 +  n  ∆x2 +  +  n  ∆xn = cn
( 0) ( 0) ( 0)

 ∂x1   ∂x2   ∂xn 

Energy Conversion Lab

 NEWTON RAPHSON METHOD FOR n VARIABLES

 Rearrange in matrix form

 ∂f ( 0 )  ∂f1  
( 0) ( 0)
 ∂f1 
 1       
 ∂x1   ∂x2   ∂xn   ( 0 )
c1 − f1  
( 0)
( 0) ( 0) ( 0) 
∆x1 
   
 ∂f 2   ∂f 2   ∂f 2    ( 0 ) 
− ( 0)
 2 2  =  
c f       ∆x2 
    ∂x1   ∂x2   ∂xn     
         ( 0) 
cn − f n  
( 0)
( 0) ( 0)
∆xn 
( 0)  
 ∂f n   ∂f n   ∂f n  
 ∂x       
 1   ∂x2   ∂xn  

 The matrix can be written as

 ΔC(k) = J(k) ΔX(k)
Energy Conversion Lab
 NEWTON RAPHSON METHOD FOR n VARIABLES

 The Newton-Raphson algorithm for n-

dimensional case is
 X(k+1) = X(k) +ΔX(k) = X(k) + [J(k)]-1ΔC(k)
 where
 ∂f ( k )  ∂f1  
(k ) (k )
∆x1( k )   ∂f1 
 (k )   1       
∆X ( k )
∆x
= 2   ∂x1   ∂x2   ∂xn  
    (k ) (k ) (k ) 
 (k )   ∂f 2   ∂f 2   ∂f 2  
∆xn       
J (k ) =  ∂x1   ∂x2   ∂xn  
c1 − f1( k )  
 (k )       
−
= 2 2   (k ) 
c f
∆C ( k ) (k ) (k )
   
 n ∂f   ∂f n   ∂f n  
 
 ∂x       
cn − f n  ∂  ∂xn  
(k )
 1   2
x

Energy Conversion Lab

 NEWTON RAPHSON METHOD FOR n VARIABLES

 The Newton-Raphson algorithm

 J(k) is called the Jacobian matrix
 solution to X(k+1) is inefficient because it involves
inverse of J(k) , a triangular factorization is used to
facilitate the computation
 in MATLAB, the operator “\” (i.e., ΔX=J\ΔC) is used to
apply the triangular factorization
 Newton-Raphson method converge to solution
quadratically when near a root
 The limitation is that it does not generally converge to
a solution from an arbitrary starting point

Energy Conversion Lab

 LINE FLOWS AND LOSSES
 Complex power flow between bus i,j
 for line model, see Fig. 6.8
 current flow from bus i to bus j
I ij = I l + I i 0 = yij (Vi − V j ) + yi 0Vi
 current flow from bus j to bus i
I ji = − I l + I j 0 = yij (V j − Vi ) + y j 0V j
 complex power Sij from bus i to j and Sji from j to i
Sij = Vi I ij* S ji = V j I *ji
 power loss in the line i-j
S L ( i − j ) = Sij + S ji
 for more Gauss-Seidel method examples, see Ex. (6.7)
and Ex. (6.8)
Energy Conversion Lab
 NEWTON-RAPHSON POWER FLOW
 Real power flow in terms of Vi , ∠δ, and Yij
V j Yij cos(θ ij − δ i + δ j )
n
Pi = ∑V
j =1
i

 Reactive power flow

Qi = − ∑ Vi V j Yij sin (θij − δ i + δ j )
n

j =1

 Newton-Raphson matrix form: ΔC(k) = J(k) ΔX(k)

 ∆P   J 1 J 2   ∆δ 
 ∆Q  =  J  
J 4   ∆ V 
   3
 diagonal and off-diagonal elements of J1
∂Pi
= ∑V V j Yij sin (θij − δ i + δ j )
∂δ i
i
j ≠i

∂Pi
= − Vi V j Yij sin (θ ij − δ i + δ j ) j≠i
∂δ j
Energy Conversion Lab
 NEWTON-RAPHSON POWER FLOW
 Newton-Raphson matrix form: ΔC(k) = J(k) ΔX(k)
 ∆P   J 1 J 2   ∆δ 
 ∆Q  =  J  
J 4   ∆ V 
   3
 diagonal and off-diagonal elements of J2
∂Pi
= 2 Vi Yii cos θii + ∑ V j Yij cos(θij − δ i + δ j )
∂ Vi j ≠i

∂Pi
= Vi Yij cos(θ ij − δ i + δ j ) j≠i
∂Vj

 diagonal and off-diagonal elements of J3

∂Qi
= ∑V V j Yij cos(θ ij − δ i + δ j )
∂δ i
i
j ≠i

∂Qi
= − Vi V j Yij cos(θ ij − δ i + δ j ) j≠i
∂δ j
Energy Conversion Lab
 NEWTON-RAPHSON POWER FLOW
 Newton-Raphson matrix form: ΔC(k) = J(k) ΔX(k)
 ∆P   J1 J 2   ∆δ 
 ∆Q  = J  
J 4   ∆ V 
   3
 diagonal and off-diagonal elements of J4
∂Qi
= −2 Vi Yii sin θ ii − ∑ V j Yij sin (θ ij − δ i + δ j )
∂ Vi j ≠i

∂Qi
= − Vi Yij sin (θ ij − δ i + δ j ) j≠i
∂Vj

 power residuals ΔPi(k) ΔQi(k)

∆Pi ( k ) = Pi sch − Pi ( k ) , ∆Qi( k ) = Qisch − Qi( k )
 new estimates for bus voltages
δ i( k +1) = δ i( k ) + ∆δ i( k ) , Vi ( k +1) = Vi ( k ) + ∆Vi ( k )
Energy Conversion Lab
 NEWTON-RAPHSON POWER FLOW
 Procedure for Newton-Raphson method:
 PQ bus: set |Vi
(0)|=1.0, δ (0)=0.0
i
 PV bus: set δi
(0)=0.0

 set PQ bus equation for J matrix elements:

V j Yij cos(θ ij − δ i + δ j )
n
Pi = ∑V
j=
i
n1
Qi = − ∑ Vi V j Yij sin (θij − δ i + δ j )
j =1

∆Pi ( k ) = Pi sch − Pi ( k ) , ∆Qi( k ) = Qisch − Qi( k )

 set PV bus equation for J matrix elements:

V j Yij cos(θ ij − δ i + δ j )
n
Pi = ∑V
j =1
i

∆Pi ( k ) = Pi sch − Pi ( k )

Energy Conversion Lab

 NEWTON-RAPHSON POWER FLOW
 Procedure for Newton-Raphson method:
 use above equation to calculate Jacobian matrix (J1, J2,

J3, J4)
 solve Δ|V| and Δδ using Newton-Raphson matrix

 ∆P   J1 J 2   ∆δ 
 ∆Q  = J  
J 4   ∆ V 
   3
 update Δ|V| and Δδ by
δ i( k +1) = δ i( k ) + ∆δ i( k ) , Vi ( k +1) = Vi ( k ) + ∆Vi ( k )
 repeat the calculation until
∆Pi ( k ) ≤ ε , ∆Qi( k ) ≤ ε
 for example: see Ex.(6.10)

Energy Conversion Lab

 FAST DECOUPLED POWER FLOW
 Fast decoupled power flow solution:
 the algorithm is based on Newton-Raphson method

 when transmission lines has a high X/R ratio, the

Newton-Raphson method could be further simplified

 Consider the Newton-Raphson power flow
equation

 ∆P   J1 J 2   ∆δ 
 ∆Q  = J  
J 4   ∆ V 
   3
 ΔP are less sensitive to |V| and most sensitive to Δδ
 ΔQ is less sensitive to Δδ and most sensitive to |V|
 we can reasonably eliminate J2 and J3 elements in
Jacobian matrix
Energy Conversion Lab
 FAST DECOUPLED POWER FLOW
 Consider the Newton-Raphson power flow equation
 the power flow equation reduces to
 ∆P   J1 0   ∆δ 
 ∆Q  = 0  
J 4   ∆ V 
  
 ΔP = J1Δδ = [∂P/∂δ]Δδ, ΔQ = J4Δ|V| = [∂Q/∂|V|]Δ|V|

 ∂Pi/∂δi = -Qi - |Vi|2Bii, Bii = |Yii|sinθii is the imaginary part of the

diagonal elements

 since Bii >> Qi, ∂Pi/∂δi (diagonal elements of J1) can be further
reduced to ∂Pi/∂δi = - |Vi|Bii (|Vi|2 ≈|Vi| )

 off diagonal element of J1: ∂Pi/∂δi = - |Vi||Vj|Yijsin(θij-δi+δj), since δj-δi

is quite small, θij-δi+δj = θij, J1 = ∂Pi/∂δj = - |Vi||Vj|Bij

 since |Vj|≈1, off diagonal elements of J1 = ∂Pi/∂δj = - |Vi|Bij

Energy Conversion Lab
 FAST DECOUPLED POWER FLOW
 Consider the Newton-Raphson power flow
equation
 similarly, diagonal elements of J4: ∂Qi/∂|Vi| = - |Vi|Bii
 off diagonal elements of J4: ∂Qi/∂|Vj| = - |Vi|Bij
 therefore, ΔP and ΔQ has the following forms
∆P ∆Q
= − B ' ∆δ , = − B '' ∆ V
Vi Vi
 B’ and B” are the imaginary part of Ybus
 the updated Δδ and Δ|V| can be obtained from
∆P ∆Q
∆δ = −[B'] ∆ V = −[B"]
−1 −1
,
V V
 to calculate PQ bus, use simplified J1 and J4 to obtain
solution
 to calculate PV bus, J4 can be further eliminated, only
J1 is used to obtain solution
Energy Conversion Lab
 FAST DECOUPLED POWER FLOW
 Comparison between fast decouple power flow
solution and Newton Raphson power flow solution
 fast decoupled solution requires more iterations than
Newton Raphson solution
 fast decoupled solution requires less time per iteration
 since decoupled solution needs less time for iteration,
the overall computation time may be less than using
the Newton Raphson method
 fast decoupled solution often used in fast computation
of power flow, for example, contingency analysis or on-
line control of power flow
 see Ex. 6.12

Energy Conversion Lab