You are on page 1of 22

EECE 471

Power Flow Analysis

POWER FLOW ANALYSIS


10.0 INTRODUCTION

The power flow analysis problem is solved to obtain the steady-state voltages and powers in a transmission system. The calculation is analogous to solving for the steady-state voltages and currents in a circuit. Consider the one-line diagram shown in Fig. 10.1. The load powers (SDi) represent the power loads supplied to large scale industrial or residential consumers. The powers (SGi) are injected at the generator nodes to supply the loads and losses incurred in the transmission system. A figure similar to Figure 10.1 might represent a sub-transmission network supplying various distribution substations and fed from the transmission substations. The power system must deliver the load powers in a reliable and economic manner with the voltage and frequency being within acceptable limits. In this chapter we are concerned mainly with the voltages being kept within limits. Thus the magnitudes and phases of the nodal voltages Vi are the variables that will be calculated in the power flow analysis problem. Once the voltages are calculated, the power flows or currents in the lines may be calculated and compared to their thermal or stability limits. Modifications in the operation such as changing SGi or in the design such as adding new lines may then be proposed to rectify the detected anomalies in the system condition.
G G G

SG1 V1 SD1 V4 SD4

V2

SG2

V3

SG3

SD3 V5 SD5

Figure 10.1 System one-line diagram 10.1 POWER FLOW EQUATIONS

The per unit complex power injected at node i is defined as:


S i = S Gi S Di

Si can be calculated as the sum of complex powers sent to particular nodes in the network. For example bus 3 in Figure 10.1, S3 the net complex power sent to nodes 2 and 5 as visualized in Fig. 10.2.

EECE 471

Power Flow Analysis

V3 S32

SG3 S32 SD3

V3 S3 S35

V3

SG3

SD3

S35

Figure 10.2 Significance of S3

In general, using conservation of complex power for the ith bus,


Si =
k =1

S ik
n

i=1, 2 n

(10.1)

We also define the injected node or bus current at bus i in per unit as
I i = I Gi I Di =
k =1

I ik

i=1, 2 n

Using I= Ybus V we get for the ith component:


Ii =
k =1

YikVk

i=1, 2 n

(10.2)

Next we calculate the ith bus power as: Si = ViIi*


* * S i = Vi Yik Vk k =1 n

(10.3)

Suppose we let
Vi = Vi e jVi = Vi e j1

ik = i k
Yik = Gik + jB ik

So using a polar representation of the complex voltages and a Cartesian one for the admittances then (10.3) becomes as follows:
Si = =
k =1 n

Vi Vk e jik (Gik jBik )


(10.4)

k =1

Vi Vk (cos ik + j sin ik )(Gik jBik ) i=1, 2 n

EECE 471

Power Flow Analysis

Resolving (10.4) into real and imaginary parts:


Pi = Qi =
k =1 n

Vi Vk (Gik cos ik + Bik sin ik )


k =1

Vi Vk (Gik sin ik Bik cos ik )

(10.5)

10.2 THE POWER FLOW PROBLEM

The power flow problem is based on mathematical as well as operational consideration. At a generator bus the active power PGi and the voltage magnitude |Vi| are specified. At all buses the complex power demands SDi are specified. Buses with generators are known as generator or PV buses and buses with no generators are load or PQ buses. For example in Figure 10.1 buses 1, 2, and 3 are generator buses and buses 4 and 5 are load buses. However, the power at all generator buses cannot be specified because the losses are not known prior to solving the power flow equations. Hence the power at one of the generator buses, selected arbitrarily, is left open to take the slack and balance the active power. This bus, known as the slack or swing bus, is selected as Bus 1 to simplify the presentation and, instead of specifying the active power, the phase 1 at this bus is specified in addition to the voltage magnitude |V1|. In summary there are three types of sources at the different buses: 1. A voltage source, assumed at Bus 1. 2. P, |V| sources, at the other generator buses. 3. P, Q sources, at the load buses. In developing the methodology for solving the power flow problem we present two cases. Case I is a simpler problem to solve; it has one slack bus and the remaining buses are PQ buses: Case I: Given V1, S2, S3, , Sn find S1, V2, V3,, Vn Case II is more typical with PV and PQ buses. The PV buses are numbered from 2 to m, and the PQ buses are from m+1 to n. Case II: Given V1, (P2, |V2|), , (Pm, |Vm|), Sm+1, ..., Sn find S1, (Q2, 2), , (Qm, m), Vm+1, ..., Vn Before we discuss the mathematical formulation note the following points:

EECE 471

Power Flow Analysis

Case I corresponds to the one-generator case. Case II is the more typical case. In both cases we assume that two out of four variables at each bus are given (complex Vi specifies |Vi| and i, while complex Si specifies Pi and Qi) and are asked to find the remaining two variables. Once the stated problem has been solved, we know all the Vi and can then solve for power flows or currents on individual transmission lines. In Case II we may specify Qimin Qi Qimax, i = 2, 3, , m.
10.3 SOLUTION BY GAUSS ITERATION

Consider Case I of the power flow problem. The equations to be solved are a form of (10.3) repeated for convenience as:
S1 = V1 Y1* Vk* k
k =1
n * S i = Vi Yik Vk* k =1

(10.3a) i= 2 n (10.3b)

Note that when the voltages are known we can solve for S1 from (10.3a). The key issue now is to solve the n-1 equations given in (10.3b). Taking complex conjugates of (10.3b) and dividing by Vi* and separating the Yii term we obtain:
S i* Vi* = YiiVi + Yik Vk
k =1 k i n

i= 2 n

(10.6)

By rearranging
1 S i* n Vi = V * Yik Vk i=2, 3, , n Yii i k =1 k i

(10.8)

Thus we get n 1 implicit nonlinear algebraic equations of a general form: ~ V2 = h2 (V2 ,V3 ,L,Vn ) ~ V3 = h3 (V2 ,V3 ,L,Vn ) (10.9) M ~ Vn = hn (V2 ,V3 ,L,Vn ) ~ where the functions hi are given by (10.8). In vector notation (10.9) takes the form x = h(x) (10.10)

EECE 471

Power Flow Analysis

Equation (10.10) is solved by iteration. In the simplest case we use the formula: xv+1 = h(x v) v=0, 1, (10.11) where the superscript v indicates the iteration number. Thus starting at x0 the iterative process generates a sequence x0, x1, x2, If the series converges (i.e., xv x*), then x* = h(x*) In practice we stop the iterations when the changes in x v become very small. Thus defining x v = x v+1 x v , we stop when the norm of is smaller than arbitrarily small positive number :
x v

The norm may be a sup norm: x v = max xi )


i

or an Euclidian norm: x

N = (xi ) 2 i =1

1/ 2

Using a scalar version of (10.11) the Gaussian iteration with v= 0, 1, 2 may be written as:
v v v v x1 +1 = h1 ( x1 , x 2 , L, x N ) v v v v x2 +1 = h2 ( x1 , x2 , L, x N )

M
v v v v x N+1 = hN ( x1 , x2 , L, x N )

(10.12)

An improved version is the Gauss-Seidel iteration given by:


v v v v x1 +1 = h1 ( x1 , x 2 , L, x N ) v v v v x2 +1 = h2 ( x1 +1 , x2 , L, x N ) v v v v x3 +1 = h2 ( x1 +1 , x2 +1 , L, x N )

(10.12)

M
v v v v 1 v x N+1 = hN ( x1 +1 , x2 +1 , L, x N+1 , x N )

Example 10.2

Use Gauss iteration to solve the system shown in Figure E10.2, i.e. find S1 and V2.

EECE 471

Power Flow Analysis

SG1 V1=10 SD1 S1 ZL=j0.5

jQG2=j1.0 V2 S2 SD2= 0.5+j1.0


Figure E10.2

Solution

We iterate on V2 using (10.8). With n= 2 we have only equation:


V2v+1 = 1 Y22
* S2 v * Y21V1 (V2 )

We next calculate the matrix Ybus:

Y j 2 j 2 Y Ybus = 11 12 = Y21 Y22 j 2 j 2 Thus putting the values for S2, V1, Y11 and Y21 we get:
V2v+1 = j
0.25 + 1.0 (V2v )*

Starting with a guess V20 = 10 we get convergence in about 6 steps. The results are listed in Table E10.2.
Table E10.2

Iteration

V2

0 1 2 3 4 5 6

1.00 1.030776-14.036243 0.970143-14.036249 0.970261-14.931409 0.966235-14.931416 0.966236-14.995078 0.965948-14.995072

EECE 471

Power Flow Analysis

A simple modification of the procedures developed to handle Case I works for Case II. For buses i= 2 to m the reactive power Qiv is not specified and thus to evaluate (10.8) the reactive power at node i, at iteration v is estimated by:
n ~ * Qiv = ImViv Yik (Vkv )* k =1

i=2, 3, , n

(10.14)

Using this estimate the voltages at buses i= 2 to m are estimated by:


~v n ~ v +1 1 Pi jQi v Yik Vk Vi = Yii (Viv )* k =1 k i

i=2, 3, , n

(10.15) and

~ Since |Vi| is specified for the generator buses, we then replace Vi v+1 by Vi

spec

obtain Viv+1.
Example 10.3

Now consider the simple example shown in Figure E10.3 to illustrate Gaussian iteration for Case II.
G

SG1 S1 ZL=j0.5 S2

SG2=0.25+jQG2 V2=12 SD2= 1.0+j0.5


Figure E10.3

V1=10 SD1
Solution

The iteration formula (10.15) with Yii= Y22= -j2, and Yik= Y21= j2 gives ~ ~ 0.75 + jQiv 1 0.75 jQiv ~ v +1 j 2 = 1 + V2 = j 2 (V2v )* j 2(V2v )*
~ where Qiv is the estimate of the reactive power at the bus at iteration v given by (10.14) as ~v * * Q2 = Im V2v Y21V1* + Y22 (V2v )*

{ [

]}

2 = Im j 2V2v + j 2 V2v

= 2(1 ReV2v )

Starting with a guess V20 = 10 we obtain the iterations in Table E10.3.

EECE 471

Power Flow Analysis

Table E10.2

Iteration
0 1 2 3 4

V2

~v Q2
0 0.1273 0.1446 0.1459 0.1459

~ V2v+1
1.0680-20.5560 1.0003-21.9229 1.0000-22.0169 1.0000-22.0238

1.00 1.0-20.5560 1.0-21.9229 1.0-22.0169 1.0-22.0238

To complete the solution we can solve for S1 using (10.3a).


* * S1 = V1 Y11V1* + Y12V2*

= j 2 ( j 2)(122.0238o ) = 0.75 + j 0.1459


10.4 NEWTON-RAPHSON ITERATION

The Newton-Raphson iteration method to solve a problem of the f(x) = 0. It is based on the Taylor series expansions for f(x + x) given by f(x + x) = f(x) + f(x) x + h.o.t. (10.21) where h.o.t. stands for higher-order terms. When n scalar equations are to be solved they are denoted by the following:
f(x) = 0

(10.22)
f1 (x) f (x) x1 + ... + 1 xn + h.o.t. xn x1 f 2 (x) f (x) x1 + ... + 2 xn + h.o.t. x1 xn

In this case, the Taylor series of the n scalar equations are given by:
f1 ( x + x) = f1 (x) + f 2 (x + x) = f 2 ( x) + M

(10.23)

f n (x) f (x) x1 + ... + n xn + h.o.t. x1 xn Using matrix notation we get f n (x + x) = f n (x) +

f (x + x) = f(x) + J(x) x + h.o.t.

(10.24)

where

EECE 471

Power Flow Analysis

f1 (x) f1 (x) L x xn 1 O M J ( x) = M f n (x) L f n (x) x1 xn

x1 x x = 2 M xn

(10.25)

J(x) is called the Jacobian matrix of f evaluated at x. By neglecting the h.o.t. and if the inverse of J(x) exists we get the general N-R iteration formula:
x v +1 = x v J (x v )

f (x v )

(10.26)

In some places we will use the simpler notation Jv rather than J(xv). The iterative procedure defined by (10.26) will generate a series of values x0, x1 xv that will make f(xv) progressively smaller as v increases. The iterative process would stop when the norm f (x) is smaller than some small number . Note that in (10.26) we need to update matrix J at every iteration and invert it. In large systems calculating inverses is computationally expensive. Instead we define
x v = x v+1 x v

(10.27)

And we write (10.26) as


J (x v ) x v = f (x v )
(10.28)

We then for xv by putting J in its upper and lower factors and doing forward and backward substitution as discussed in Section 9.2.
Example 10.5

Given the direct-current (dc) system shown in figure E10.5, use the NewtonRaphson method to find the (dc) bus voltages V1 and V2 and find PG1.

R V1 PD1 PD2 PD3

= 0.01 = 1.0 = 0.5 = 1.0 = 0.5

Figure E 10.5

EECE 471 Solution

Power Flow Analysis

10

Note that the considerations and techniques developed for ac systems hold as well for dc systems. For simplicity we are using the same notation, although all the variables are real quantities. Forming Ybus we have
Ybus 2 1 1 = 100 1 2 1 1 1 2

In what corresponds to (10.3), we write the power flow equations P1 = 200V12 100V1V2 100V1V3 P2 = -100V2V1 +200V22 100V2V3 P3 = -100V3V1 100V3V2 + 200V32 Bus 1 is the slack bus, with V1 known and P1 unknown. We (temporarily) strip away the first equation and solve the remaining two for the unknown V2 and V3. In using the Newton-Raphson the next-step is to put the equations in the form:
f(x) = 0.

The simplest way is to subtract the left sides from the right sides, or vise-versa. Subtracting left from right and putting in the known values of P2 = -PD2 = -1.0, P3 = -PD3 = -0.5, V1 = 1.0, we get f1(x) = 1.0 100V2 + 200V22 100V2V3 = 0 f2(x) = 0.5 100V3 100V2 V3 + 200V32 = 0 The first component of x is V2, the second is V3. Next we find the Jacobian matrix,
f1 x J ( x) = 1 f 2 x1 f1 1 + 4V2 V3 x2 = 100 f 2 V2 x2 1 V2 + 4V3 V2

In the 22 case it is easy to find the inverse and we will solve using (10.26). In higher-dimensional cases we would almost certainly use (10.28). Starting with a flat-profile (i.e. V2 = 1, V3 = 1), we get
2 1 J 0 = 100 1 2

(J )

0 1

1 2 1 300 1 2

1.0 f (x 0 ) = 0.5

Then using (10.26) yields

EECE 471

Power Flow Analysis

11

1 1 2 1 1.0 0.991667 x1 = = 1 300 1 2 0.5 0.993333

Continuing to the next iteration gives us 1.973333 0.991667 J 1 = 100 J1 0.99333 1.981667
0.00843 f (x1 ) = 0.00323

( )

1 1.981667 0.991667 292.54 0.99333 1.973333

Note that x1 is really a very good estimate. We want f(x) = 0 and already f(x1) is almost zero! Note the great improvement from f(x0) to f(x1) in just one step. Continuing the iteration we have
0.991667 1 1.981667 0.991667 0.00843 0.991599 x2 = 292.54 0.993333 1.973333 0.00323 = 0.993283 0.993333

Noting that the change between x1 and x2 is less than 0.0005, we stop the iteration. Alternatively, we can compute f(x2). We get
0.000053 f (x 2 ) = 0.000040

which is small compared to the power levels given for this problem. Finally we find P1 by using the first equation in (10.29). We find P1 = 1.511800. Note that since PD1 + PD2 = 1.5, the I2R loss in the transmission system is 0.011800.
10.5 APPLICATION TO POWER FLOW EQUATIONS

The power flow equations that are solved in the Newton-Raphson (N-R) iterations are obtained from (10.5) and repeated here as
Pi = Qi =
k =1 n

Vi Vk (Gik cos ik + Bik sin ik )


k =1

Vi Vk (Gik sin ik Bik cos ik )

i=1, 2 n

(10.5)

The first equations corresponding to the slack bus are removed since at this bus P1 and Q1 are not specified. The left hand sides of (10.5) are specified values given by Pi = PGi PDi and Qi = QGi QDi . The right hand sides represent the calculated power flowing to the network from the bus. They are functions of |Vi| and i, i= 2, 3 n. It is convenient to define the vectors , |V| and their composite vector x as follows:

EECE 471

Power Flow Analysis

12

2 = M n

V2 V = M Vn

x= V

(10.31)

At the solution point the calculated active and reactive powers are equal to the powers specified at the bus. We then represent the RHS of (10.5) with a simpler notation, remove the equations of the slack bus, and rearrange to get
Pi (x) Pi = 0 Qi (x) Qi = 0

i= 2, 3 n

(10.34)

Equations (10.34) have 2n-2 components and correspond to the vector function f(x) in (10.22). In expanded matrix notation (10.34) take the following form
P2 (x) P2 M Pn (x) Pn f ( x) = = 0 Q2 (x) Q2 M Q ( x) Q n n

(10.35)

Next we need to identify the Jacobian matrix J(x) of f(x) given in a partitioned format as follows:
P2 (x) 2 M Pn (x) 2 J 12 = J 22 Q (x) 2 2 M Qn (x) 2 L O L L P2 (x) n M Pn (x) n P2 ( x) | V2 | M Pn (x) | V2 | L O L P2 ( x) | Vn | M Pn (x) | Vn | Q2 ( x) | Vn | M Qn (x) | Vn |

J J (x) = 11 J 21

(10.36)

Q2 (x) n O M Qn (x) L n

Q2 ( x) L | V2 | M O Qn (x) L | V2 |

Each partition of matrix J(x) has dimensions (n-1)(n-1). The terms of The Jacobian matrix J(x) can be evaluated explicitly using the RHS of (10.5). Finally to determine the form of N-R iterations we define the so called mismatch vectors:

EECE 471

Power Flow Analysis

13

P2 (x) P2 M P (x) = Pn (x) Pn P (x) f ( x) = Q(x)

Q2 (x) Q2 M Q(x) = Qn (x) Qn

(10.37)

The f(x) can be expressed in terms of the mismatch vectors as follows: (10.38)

So finally by using (10.31), (10.36), and (10.38) in (10.28) we get:


v J 11 v J 21 v P (x v ) J 12 v = v J v | V |v Q(x ) 22

(10.39)

Starting from an initial guess x0= [0; |V|0] at every step (v) of the N-R iteration we evaluate the mismatch vectors P(xv) and Q(xv) and the Jacobian matrix J(xv) and solve (10.39) to obtain xv = [v; |V|v] and then use xv+1= xv + xv. As the iterations proceed the mismatch vectors are expected to go to zero. Next we calculate the elements of the Jacobian matrices as follows:
11 J ii =

Pi (x) n = Vi Vk (Gik sin ik + Bik cos ik ) i k =1


k i

11 J ik = 12 J ii = 12 J ik =

Pi (x) = Vi Vk (Gik sin ik Bik cos ik ) k Pi (x) = Vk (Gik cos ik + Bik sin ik ) + Vi Gii Vi k =1
n

(10.40a)

Pi (x) = Vi (Gik cos ik + Bik sin ik ) Vk Qi (x) n = Vi Vk (Gik cos ik + Bik sin ik ) i k =1
k i

21 J ii =

21 J ik = 22 J ii = 22 J ik =

Qi (x) = Vi Vk (Gik cos ik + Bik sin ik ) k Pi (x) = Vk (Gik sin ik Bik cos ik ) Vi Bii Vi k =1
n

(10.40b)

Qi (x) = Vi (Gik sin ik Bik cos ik ) Vk

EECE 471

Power Flow Analysis

14

From these equations we can deduce some properties of the Jacobian matrix J. The J matrix is real but has a dimension of (2n-2)(2n-2). The off-diagonal elements in each submatrix involve only elements directly connecting nodes in the network. If there is no direct connection between two buses, the corresponding entry in the submatrix is zero. Thus each partition of J has a structure similar to that of Ybus and as such is called a sparse matrix. For large networks the number of zero elements could be higher than 90%! In a Case II problem, at the generator or PV buses we do not solve for |Vi| because these are specified and hence we have less variables to solve for in this case. Hence for every generator bus we remove the reactive power equation involving Qi since it is not specified and hence cannot be equated to the calculated power Qi (x). Given that n and m are the number of buses and the number of generators, the number rows and columns of the Jacobian matrix J is n+m-2. This is illustrated by the structure of J(x) matrix for the system shown in Figure 10.1 as follows:
P2 (x) 2 P3 (x) 2 P4 (x) 2 J ( x) = P (x) 5 2 Q4 (x) 2 Q5 (x) 2 P2 (x) 3 P3 (x) 3 P4 (x) 3 P5 (x) 3 Q4 (x) 3 Q5 (x) 3 P2 (x) 4 P3 (x) 4 P4 (x) 4 P5 (x) 4 Q4 (x) 4 Q5 (x) 4 P2 (x) 5 P3 (x) 5 P4 (x) 5 P5 (x) 5 Q4 (x) 5 Q5 (x) 5 P2 ( x) V4 P3 (x) V4 P4 ( x) V4 P5 (x) V4 Q4 ( x) V4 Q5 (x) V4 P2 (x) V5 P3 (x) V5 P4 (x) V5 P5 (x) V5 Q4 (x) V5 Q5 (x) V5

The vector of variables x is given by:


2 3 4 x= 5 V4 V5

It could happen that at the end of the N-R power flow analysis one of the generators is providing reactive power outside its capability limits discussed in Chapter 6. In this case, the reactive power generated is set at the limit that was

EECE 471

Power Flow Analysis

15

violated and the bus is treated as a PQ bus. The voltage magnitude at the bus is reintroduced as a variable and the Jacobian matrix and the mismatch vector are augmented to reflect the change in the bus type, then the solution process is repeated.
Example 10.6

Use the Newton-Raphson method to find 2, 3, |V3|, then calculate SG1 and QG12 for the system shown in Figure E10.6. The shunt elements of transmission lines are equal to j0.02 and their series impedances are equal to j0.1.
G SG1 G

PG2=0.6661 |V2|=1.05

V1=10

V3 SD3=2.8653+j1.2244
Figure E10.6 Solution

The Ybus matrix of the system is given by:


Ybus j10 j10 j19.98 = j10 j19.98 j10 j10 j19.98 j10

Bus 1 is the slack bus, Bus 2 is a PV bus, and Bus 3 is a PQ bus. We have 3 variables to determine during the N-R iterations and the Jacobian is a 33 matrix. We then write the expression for P2(x), P3(x), and Q3(x) using the RHS of (10.5):
P2 (x) = V1 V2 B21 sin( 2 1 ) + V2 V3 B23 sin( 2 3 ) = 10.5 sin 2 + 10.5 V3 sin( 2 3 ) P3 (x) = V3 V1 B31 sin( 3 1 ) + V3 V2 B32 sin( 3 2 ) = 10 V3 sin 2 + 10.5 V3 sin( 3 2 )
2

(10.42a)

(10.42b)

Q3 ( x) = [ V3 V1 B31 cos( 3 1 ) + V3 V2 B32 cos( 3 2 ) + V3 B33 ] = [10 V3 cos 3 + 10.5 V3 cos( 3 2 ) 19.98 V3 ] (10.42c)
2

EECE 471

Power Flow Analysis

16

The vector of unknown x and the Jacobian J(x) matrix are given by:
P2 2 P J ( x) = 3 2 Q3 2 P2 3 P3 3 Q3 3 P2 V3 P3 V3 Q3 V3

2 x = 3 V3

(10.43)

We can now calculate the elements of the Jacobian matrix:


P2 = V1 V2 B21 cos( 2 1 ) + V2 V3 B23 cos( 2 3 ) 2 = 10.5 cos 2 + 10.5 V3 cos( 2 3 ) P2 = V2 V3 B23 cos( 2 3 ) 3 = 10.5 V3 cos( 2 3 )

P2 = V2 B23 sin( 2 3 ) V3 = 10.5 sin( 2 3 )


P3 = 10.5 V3 cos( 3 2 ) 2

P3 = 10 V3 cos 3 + 10.5 V3 cos( 3 2 ) 3


P3 = 10 sin 3 + 10.5 sin( 3 2 ) V3

Q3 = 10.5 V3 sin( 3 2 ) 2 Q3 = 10 V3 sin 3 + 10.5 V3 sin( 3 2 ) 3


Q3 = [10 cos 3 + 10.5 cos( 3 2 ) 39.96 V3 V3

EECE 471

Power Flow Analysis

17

Let us note the scheduled powers at the buses: P2= PG2= 0.6661 P3= -PD3= -2.8653 Q3= -QD3= -1.2244 We start the N-R iterations by making an initial guess for the unknowns:
2 0 0 x = 3 = 0 V3 1
0

We then calculate the mismatch vector:


0 P2 (x 0 ) P2 0 0.6661 0.6661 P2 P = P (x 0 ) P = 0 + 2.8653 = 2.8653 3 3 3 Q (x 0 ) Q 0.52 + 1.2244 0.7044 Q3 3 3 As expected, at the initial guess the mismatch is large.

Iteration 1:

We start the iteration by calculating the Jacobian matrix J0 using (10.43):


10.5 0 21 10.5 20.5 0 J (x ) = 0 0 19.46 The inverse of the Jacobian is given by:
0

[J(x )]
0

J = 11 0

0 J 22

1 J 11 = 0

0 0.0640 0.0328 0 = 0.0328 0.0656 0 1 J 22 0 0 0.0514


0.6661 0.0513 2.8653 = 0.01660 0.7044 0.0362

The correction on the initial guess is obtained:


0 0.0640 0.0328 x 0 = 0.0328 0.0656 0 0 0 0.0514

Next we update the values of the variables:


1 2 0 0.0513 0.0513 1 1 0 0 x = 3 = x + x = 0 + 0.01660 = 0.01660 V 1 1 0.0362 0.9638 3

EECE 471

Power Flow Analysis

18

We calculate the mismatch vector using the new values of variables:


1 1 P2 P2 (x ) P2 0.0463 P = P (x1 ) P = 0.1145 3 3 3 Q3 Q3 (x1 ) Q3 0.2251

The elements of the mismatch vector has have been reduced by a factor of 10. But we judge that they are not small enough and we continue iterating.
Iteration 2:

We then calculate the Jacobian matrix J1 using (10.43):


20.5396 10.0534 1.2017 J (x ) = 10.0534 19.5589 2.8541 1.1582 2.7508 18.2199
1

The inverse of the Jacobian is given by:

[J(x )]

1 1

0.0651 0.0336 0.0010 = 0.0336 0.0696 0.0087 0.0009 0.0084 0.0561 0.0463 0.0524 0.1145 = 0.1744 0.2251 0.9502

The new values of the variables are given by:


0.0513 0.0651 0.0336 0.0010 x = 0.1660 0.0336 0.0696 0.0087 0.9638 0.0009 0.0084 0.0561
2

We then calculate the mismatch vector:


0.0019 P2 P = 0.0023 3 0.0031 Q3
2

The mismatch is judged small enough and on this basis we may stop iterating. We finally calculate SG1 and QG12 using (10.5):
PG1 = P1 = V1 V2 B12 sin(1 2 ) + V1 V3 B13 sin(1 3 )

= 10.5 sin(0.0524) + 9.502 sin(0.1744) = 2.1987

EECE 471

Power Flow Analysis


2

19

QG1 = Q1 = [ V1 V2 B12 cos(1 2 ) + V1 V3 B13 cos(1 3 ) + V1 B11 ]

= [10.5 cos(0.0524) + 9.502 cos(0.1744) 19.98 ] = 0.1365


QG 2 = Q2 = [ V2 V1 B21 cos( 2 1 ) + V2 V3 B23 cos( 2 3 ) + V2 B22 ]
2

= [10.5 cos(0.0524) + 9.977 cos(0.122) 22.028 ] = 1.6395


10.6 DECOUPLED POWER FLOW

In Example 10.6 we notice that the elements of the off-diagonal submatrices J12 and J21 are much smaller than elements of matrices J11 and J22. The reason may seen by looking at some terms. For example consider a typical term in J12:
12 J ik =

Pi (x) = Vi (Gik cos ik + Bik sin ik ) Vk

Transmission lines are mostly reactive and Gik are small, and under normal operating conditions the angle ik= i - k is relatively small making the term involving Gik also quite small. The same reasoning applies to elements of J21:
21 J ik =

Qi ( x) = Vi Vk (Gik cos ik + Bik sin ik ) k

However typical terms in the diagonal submatrices J11 and J22 are not small as can concluded by examining typical terms of these matrices in (10.40). Note that in these terms the dominant Bik is being multiplied by cos ik. The implication is that changes in active power Pi depends mostly on i (and not so much on |Vi|), and changes in reactive power depends mostly on |Vi| (and not so much on i). Based on this discussion the Jacobian matrix can be approximated as follows:
v J11 J 0 v

0 Jv 22

The N-R iterative equations (10.39) take the following form:


v J 11 v = P (x v )

(10.51a) (10.51b)

J v V = Q(x v ) 21

Again, another observation in Example 10.6 is that matrices J0 and J1 did not significantly change and as such further simplification can be made to equations 10.51. This is due to the fact that in transmission systems under normal operating conditions the phase angle differences across lines are small (i.e. ik 0) and the

EECE 471

Power Flow Analysis

20

voltage magnitudes are around 1 per unit (i.e. |Vi||Vk|1). Under these conditions matrices J11 and J22 are approximated by matrices H and L, respectively. The elements of these matrices are given by:
H ii = Bik
k i

and and

H ik = Bik

(10.52a) (10.52b)

Lii = Bik 2 Bii


k i

Lik = Bik

Therefore equations (10.51a) and (10.51b) are replaced by the following:


H v = P (x v ) L V = Q(x v )
v

(10.53a) (10.53b)

It is important to observe that matrices H and L are constant matrices that are independent of the iteration count, which may be approximated by the negative of the B matrix, the imaginary part of Ybus. Thus matrices H and L are formed and inverted (or factorized) only once at the onset of the iterative process. These equations are known as the fast-decoupled power flow equations. In programming the fast-decoupled power flow usually one full iteration of the active power equation (10.53a) is carried out followed by one full iteration of the reactive power equation (10.53b) using the updated values of phase angle as follows:
v = H 1P ( v , V )
v v +1

= +
v

(10.54a)
v

V = L1Q( v+1 , V )
v

v +1

= V +V

(10.54b)

The above equations (10.54a and 10.54b) imply a kind of block successive displacement which allows for a faster convergence and is illustrated in the following example.
Example 10.7

Use the fast-decoupled Newton-Raphson method to find 2, 3, |V3|, of the system of Example 10.6 then calculate SG1 and QG12. The system diagram and data are repeated in Figure 10.7 for convenience. The shunt elements of transmission lines are equal to j0.02 and their series impedances are equal to j0.1.

EECE 471

Power Flow Analysis

21

G SG1

PG2=0.6661 |V2|=1.05

V1=10

V3 SD3=2.8653+j1.2244
Figure E10.7 Solution

The Ybus matrix of the system was found to be:


Ybus

j10 j10 j19.98 j10 = j10 j19.98 j10 j10 j19.98

Thus (10.54a) for this problem is:


v P ( v , V v ) 19.98 10 2 2 10 19.98 = v v P3 ( , V ) 3

In this simple example we will use matrix inversion to determine the variable updates as follows:
v v 2 0.0668 0.0334 P2 ( v , V ) = 0.0334 0.0668 v v P3 ( , V ) 3

The updated phase angles are then updated as:

v +1

= 2 3
v

v +1

= 2 + 2 3 3
v v

And (10.54b) for this problem is:


19.98 V3 = Q3 ( v +1 , V ) V3 = 0.0501 Q3 (x v )
v

Note that the calculation of P2(v, |V|v), P3(v, |V|v), and Q3(v, |V|v) are based on equations (10.42a), (10.42b), and (10.42c), respectively. The results of the iterations are given in Table E10.7.

EECE 471

Power Flow Analysis

22

Table E10.7

Iteration v 0 1 2 3

2
0 -0.0513 -0.0522 -0.0523

3
0 -0.1691 -0.1737 -0.1744

|V3| 1 0.9540 0.9505 0.9501

P2 -0.6661 -0.0272 -0.0050 -0.0007

P3 2.8653 0.0829 0.0131 0.0019

Q3 0.9198 0.0683 0.0094 0.0014

You might also like