Chapter 7

Optimal Dispatch of Generation Part III

7.5 Economic Dispatch Neglecting Losses and

Including Generator Limits
In the real world, generators have limits on their generation, both minimum and
maximum generation. Thus we have the inequality constraints
Pi ( min ) ≤ Pi ≤ Pi ( max ) i = 1,2,K , ng
which are in addition to the equality constraint
ng
∑ Pi = PD
i =1
Now the Kuhn-Tucker conditions compliment the Lagrangian conditions to include the
inequality constraints as additional terms. The necessary conditions for optimal dispatch
in lossless systems but including generator limits become:
dCi
=λ for Pi ( min ) ≤ Pi ≤ Pi ( max )
dPi
dCi
≤λ for Pi = Pi ( max )
dPi
dCi
≥λ for Pi = Pi ( min )
dPi
The numerical solution is the same as before. Thus for an estimated value of lambda, the
powers are found for each participating generator. If the power is outside the limits of a
generator, that generator is "pegged" at that limit (either Pmin or Pmax), and the
generator is no longer a participating generator in the optimization of dispatched power
(since its power is pegged at either the high or low limit).

Example 7.6

Find the optimal dispatch and the total cost in $/h for the thermal plants of Example 7.4
when the total load is 975 MW with the following generator limits (in MW):
200 ≤ P1 ≤ 450
150 ≤ P2 ≤ 350
100 ≤ P3 ≤ 225
Assume the initial value of λ (1) = 6 . From the coordination equations we compute the
powers as follows:
 (1) 6 − 5.3
P1 = = 87.5
2 ( 0.004 )
(1) 6 − 5.5
P2 = = 41.67
2 ( 0.006 )
(1) 6 − 5.8
P3 = = 11.11
2 ( 0.009 )
∆P (1) = 975 − ( 87.5 + 41.67 + 11.11) = 834.72
and we compute change in lambda as
(1) 834.72
∆λ = = 3.1632
1 1 1
+ +
2 ( 0.004 ) 2 ( 0.006) 2 ( 0.009 )
Therefore the new value of lambda is
(2 )
λ = 6 + 3.1632 = 9.1632
Continuing for a second iteration we have
( 2) 9.163 − 5.3
P1 = = 482.89
2 ( 0.004 )
( 2) 9.163 − 5.5
P2 = = 305.26
2 (0.006 )
( 2) 9.163 − 5.8
P3 = = 186.84
2 (0.009 )
and
∆P ( 2 ) = 975 − ( 482.89 + 305.26 + 186.84 ) = 0
Thus the equality constraint is met, but the power of generator No. 1 exceeds its upper
limit. Thus we change it from 482.89 to 450, thus:
∆P ( 2 ) = 975 − 450 + 305.26 + 186.84 = 32.89
( )
Now we continue, but we have only two generators participating in the optimal dispatch
since the first generator is "pegged" at 450:
( 2) 32.89
∆λ = = 0.2368
1 1
+
2 ( 0.006 ) 2 ( 0.009 )
and the new value of lambda is
( 3)
λ = 9.1632 + 0.2368 = 9.4
And the powers are now as follows:

2
 (3 )
P1 = 450
(3) 9.4 − 5.5
P2 = = 325
2 ( 0.006 )
(3) 9.4 − 5.8
P3 = = 200
2 ( 0.009 )
and
∆P ( 3) = 975 − ( 450 + 325 + 200 ) = 0
Thus the equality constraint is met and the above is the optimal generation dispatch. The
total generation cost is found using the powers found above thus:
C t = 500 + 5.3 × 450 + 0.004 × 4502 + 400 + 5.5 × 325 + 0.006 × 3252
+ 200 + 5.8 × 200 + 0.009 × 2002 = 8,236.25 $/h

We could solve this using the following Matlab commands:

cost=[500 5.3 0.004

400 5.5 0.006
200 5.8 0.009];
mwlimits=[200 450
150 350
100 225];
Pdt=975;
dispatch
gencost

Incremental cost of delivered power (system lambda) = 9.400000 $/MWh

Optimal Dispatch of Generation:

450.0000
325.0000
200.0000

Total generation cost = 8236.25 $/h

7.6 Economic Dispatch Including Losses

When the loads and generators are in a small geographical region, transmission losses
may be neglected and the optimal dispatch of generation is achieved with all plants
operating at equal incremental productions cost. However, in a large system over long
distances, transmission losses become significant and need to be considered. We shall
use Kron's loss formula which is
ng ng ng
PL = ∑∑ PB
i ij Pj + ∑ B0 i Pi + B00
i =1 j =1 i =1

3
The coefficients Bij are called loss coefficients or B-coefficients. It is assumed —with
little error— that these coefficients are constant (as long as operation is near the value
where these coefficients are computed).

The economic dispatch problem is to minimize the overall generation cost Ci , which is a
function of the plant outputs
ng
C t = ∑ Ci
i =1

( )
n
= ∑ αi + β Pi + γ Pi 2
i =1
Note: If all generators participate in this minimization, then n = ng and either equation
above may be used. If only ng < n generators are participating, then the first equation is
used.

This minimization is to be performed subject to the constraint that the total generation
must equal the demand plus the losses thus
ng
∑ Pi = PD + PL
i =1
satisfying the inequality constraints:
Pi ( min ) ≤ Pi ≤ Pi ( max ) i = 1, 2, K , ng
where Pi ( min ) and Pi ( max ) are the minimum and maximum generating limits,
respectively, for plant i.

Using the Lagrange multiplier and adding additional terms to include the inequality
constraints, we have
  ng
( ) ( )
ng ng
L = Ct + λ  PD + PL − ∑ Pi  + ∑ µi ( max ) Pi − Pi ( max ) + ∑ µi ( min ) Pi − Pi ( min )
 
 i =1  i =1 i =1
The constraints should be understood to mean that µi ( max ) = 0 when Pi < Pi ( max ) and
µi ( min ) = 0 when Pi > Pi ( min ) . Thus if the power is not beyond the limits both µi ( min ) = 0
and µi ( max ) = 0 thus L has only the equality constraint. If a generator's power is beyond
one of the limits, then that limit is the power to which the generator is "pegged", and the
remaining generators participate in the optimization of the load dispatch. These
statements are clear from the necessary conditions for a minimum which are:

4
 ∂L
=0
∂Pi
∂L
=0
∂λ
∂L
= Pi − Pi ( max ) = 0
∂µi ( max )
∂L
= Pi − Pi ( min ) = 0
∂µi ( min )
The last two equations imply that the power should not go beyond its limit, and when it is
within its limits then µi ( min) = µi ( max ) = 0 and the Kuhn-Tucker function becomes the
same as the Lagrangian one.

The first equation results in

∂Ct  ∂P 
+ λ  0 + L − 1 = 0
∂Pi  ∂Pi 
Since Ct = C1 + C2 + L + Cng the above equation becomes
∂Ci ∂P
+ λ L =λ i = 1,2,K, ng **
∂Pi ∂Pi
∂PL
The term is known as the incremental transmission loss. The second condition is
∂Pi
given by
ng
∑ Pi = PD + PL
i =1
which is precisely the equality constraint imposed by the losses and load demand. The
classical form of the previous equation is
 
 1  ∂C
  i =λ i = 1,2,K , ng
∂
 1 − L  ∂Pi
P
 ∂Pi 

or
∂C
Li i = λ i = 1,2,K , ng
∂Pi
 
 1 
where Li is known as the penalty factor of plant i and is given by   . Thus the
 1 − ∂PL 
 ∂Pi 

effect of transmission losses is to introduce a penalty factor with a value that depends on
the location of the plant with respect to the loads. In the above equation it is clear that the

5
most economic cost of dispatch is obtained when the incremental cost multiplied by the
corresponding penalty factor are equal for all the participating plants.

The incremental transmission loss can be found from the loss equation and is given by
ng
∂PL
= 2 ∑ Bij Pj + B0i
∂Pi j =1
and the incremental production cost is given by
dCi
= 2γ i Pi + βi
dPi
Using these two equations in the equation marked ** (reproduced below for convenience)
∂Ci ∂P
+ λ L =λ i = 1,2,K, ng **
∂Pi ∂Pi
we have
ng
βi + 2γ i Pi + 2λ ∑ Bij Pj + B0i λ = λ
j =1
or (divide by 2λ and pull out βii from the summation)
ng
 γi  1 βi 
 + Bii  Pi + ∑ Bij Pj = 1 − B0i −  *
λ  j =1, j ≠i 2 λ 
Expanding the above equation for all plants and detailing it into full matrix form we have
γ 1   β1 
 λ + B11 B12 L B1ng   1 − B01 − λ 
   P1   
 B γ     β2 
22 L
2 + B B2ng P 1 1 − B02 −
 21
λ  2  =  λ 
   M  2 
M M O M M
  P   
 γ ng   ng   βng 
 Bng1 Bng 2 L + Bngng  1 − B0ng − 
 λ   λ 
or in short matrix form
EP = D
To find the optimal dispatch first an estimated guess of λ , say λ (1) is made. The
simultaneous linear equations above are solved using Matlab (P = E\D). Then the
iterative process is continued using the gradient method. Thus, from equation * above we
have Pi at the kth iteration expressed as
λ ( k ) (1 − B0i ) − β i − 2λ (k ) ∑ j ≠ i Bij Pj( )
k
(k)
Pi =
( )
***
(k )
2 γ i + λ Bii
and using this equation in the equality constraint gives

6
 λ ( k ) (1 − B0i ) − β i − 2λ ( k ) ∑ j ≠ i Bij Pj( )
k
ng ng
(k )
∑ Pi =∑ = PD + PL( )
k

i =1 i =1 (
2 γi + λ (k)
Bii )
or
f ( λ )( ) = PD + PL( )
k k

Expanding the left side above in a Taylor series and keeping only first order terms results
in
( k )  df ( λ )  ( k ) (k )
f (λ ) +   ∆λ = PD + PL
 dλ 
or
ng
PD + PL − ∑ Pi( )
(k )
k

∆λ ( k ) = i =1
(k )
 df ( λ ) 
 
 dλ 
∆P ( )
k
=
(k )
 df ( λ ) 
 
 dλ 
∆P ( )
k
= #
(k )
∑  d λi 
dP

where
ng
∆P ( k ) = PD + PL( ) − ∑ Pi( )
k k
++
i =1
and
(k )
ng
 ∂P 
(k ) ng γ i (1 − B0i ) − Bii β i − 2γ i ∑ j ≠ i Bij Pj
∑  ∂λi  =∑ +++
( (k )
)
2
i =1 i =1 2 γi + λ Bii
This last equation is derived by differentiation with respect to λ of *** above and
  u  v∂u − u∂v 
summing the result  recall that ∂   = .
 v v2 
Now we have
( k +1) (k ) (k )
λ =λ + ∆λ
The process is repeated till ∆P ( ) is less than a specified accuracy.
k

If we simplify the loss equations and use the simpler

7
 ng
PL = ∑ Bii Pi 2
i =1
(where we assume Bij = 0 and B00 = 0 ), the solution of the simultaneous equations
simplifies to
λ ( ) − βi
k
(k)
Pi =
( )
##
2 γ i + λ ( ) Bii
k

which are known as the coordination equations and equation +++ simplifies to
ng ( k ) ng
 ∂Pi  γ −B β
∑  ∂λ  = ∑ i (iik ) i 2 ###
i =1
(
i =1 2 γ + λ
i Bii )
The iteration process remains the same using these simpler functions.

Example 7.7

The fuel cost in $/h of three thermal plants of a power system are
C1 = 200 + 7.0 P1 + 0.008 P12
C 2 = 180 + 6.3P2 + 0.009 P22
C 3 = 140 + 6.8P3 + 0.007 P32
where the powers are in MW. The plant outputs are subject to the following limits
10 MW ≤ P1 ≤ 85 MW
10 MW ≤ P2 ≤ 80 MW
10 MW ≤ P3 ≤ 70 MW
For this example, assume the power loss is given by the simplified expression
PL ( pu ) = 0.0218P12( pu ) + 0.0228 P22( pu) + 0.0179 P32( pu)
where the loss coefficients are specified per unit on a 100-MVA base. Determine the
optimal dispatch of generation when the total system load is 150 MW.

First we change the power loss equation above to actual units instead of pu:
PL = 0.000218 P12 + 0.000228 P22 + 0.000179 P32
where the powers are now in MW.

Assume an initial value of λ (1) = 8.0 . From the coordination equations we have:

8
 (1) 8.0 − 7.0
P1 = = 51.3136 MW
2 ( 0.008 + 8.0 × 0.000218 )
(1) 8.0 − 6.3
P2 = = 78.5292 MW
2 ( 0.009 + 8.0 × 0.000228)
(1) 8.0 − 6.8
P3 = = 71.1575 MW
2 ( 0.007 + 8.0 × 0.000179 )
The real power loss is
P ( ) = 0.000218 51.3136
( )2 + 0.000228 (78.5292 )2 + 0.000179 ( 71.1575 )2 = 2.886
1
L
Since PD = 150 MW , the error is (from equation ##)
∆P (1) = 150 + 2.8864 − (51.3136 + 78.5292 + 71.1517 ) = −48.1139
and from equation ### :
(1)
3
 ∂Pi  0.008 + 0.000218 × 7.0 0.009 + 0.000228 × 6.3
∑  ∂λ  = 2 0.008 + 8.0 × 0.000218 2 + 2 0.009 + 8.0 × 0.000228 2 +
i =1 ( ) ( )
0.007 + 0.000179 × 6.8
= 152.4924
2 ( 0.007 + 8.0 × 0.000179 )
2

From equation # we have:

−48.1139
∆λ ( ) =
1
= −0.31552
152.4924
Hence the new estimate for λ is
(2 )
λ = 8.0 − 0.31552 = 7.6845
Repeating the above steps it is found at the fourth step that the change in lambda becomes
very small and convergence is declared at
( 4)
λ = λ = 7.6789
And the optimal dispatch for λ = 7.6789 is given by
( 4) 7.6789 − 7.0
P1 = = 51.3136 MW
2 ( 0.008 + 7.6789 × 0.000218 )
( 4) 7.6789 − 6.3
P2 = = 64.1317 MW
2 ( 0.009 + 7.6789 × 0.000228)
( 4) 7.6789 − 6.8
P3 = = 52.4767 MW
2 ( 0.007 + 7.6789 × 0.000179 )
The real power loss is
P ( ) = 0.000218 35.0907
( )2 + 0.000228 ( 64.1317 )2 + 0.000179 (52.4767 )2 = 1.699
4
L
and the total fuel cost is
Ct = 200 + 7.0 (35.0907 ) + 0.008 ( 35.0907 ) + 180 + 6.3 ( 64.1317 ) +
2

0.009 ( 64.1317 ) + 140 + 6.8 (52.4767 ) + 0.007 (52.4767 ) = 1592.65 $/h

2 2

9
The "dispatch" program can be used to solve example 7.7. The program is designed so
the loss coefficients are specified in pu. The loss coefficients are arranged in a matrix
form with the reserved name "B". The base MVA must be specified in the reserved
variable name "basemva". If the base MVA is not specified, it is set to 100 MVA. The
following Matlab commands are then used to solve example 7.7:

cost=[200 7.0 0.008

180 6.3 0.009
140 6.8 0.007];
mwlimits=[10 85
10 80
10 70];
Pdt=150;
B=[0.0218 0 0
0 0.0228 0
0 0 0.0179];
basemva=100;
dispatch
gencost

Incremental cost of delivered power (system lambda) = 7.678935 $/MWh

Optimal Dispatch of Generation:

35.0907
64.1318
52.4767

Total generation cost = 1592.65 $/h

It is very impressive how much less time it takes to solve example 7.7 using
the Matlab code…

Example 7.8

Figure 7.7 page 295 of the text shows the one-line diagram of a 5-bus 3-generator power
system. Given the loss coefficients (pu on a 100 MVA base) are as follows:
0.0218 0.0093 0.0028
B = 0.0093 0.0228 0.0017 
 
0.0028 0.0017 0.0179 
B0 = [0.0003 0.0031 0.0015]
B00 = [0.00030523]
Cost functions, generator limits, and total loads are as given in example 7.7. Use
"dispatch" program to find the optimal dispatch of generation.

10
The following Matlab commands are used:

cost=[200 7.0 0.008

180 6.3 0.009
140 6.8 0.007];
mwlimits=[10 85
10 80
10 70];
Pdt=150;
B=[0.0218 0.0093 0.0028
0.0093 0.0228 0.0017
0.0028 0.0017 0.0179];
B0=[0.0003 0.0031 0.0015];
B00=0.00030523;
basemva=100;
dispatch
gencost

Incremental cost of delivered power (system lambda) = 7.767785 $/MWh

Optimal Dispatch of Generation:

33.4701
64.0974
55.1012

Total generation cost = 1599.98 $/h

11