ISBN 978-93-5156-328-0

International Conference of Advance Research and Innovation (ICARI-

2014)

Economic Load Dispatch Problem and Mat

lab Programming of Different Methods
Rahul Dogra a,*, Nikita Gupta b, Harsha Saroa b
a b
Siemens Limited India, India Department of Power Systems, Delhi
Technological University, Delhi, India

Lambda Iteration Crankshaft Abstract

Article Info Economic Load Dispatch (ELD) is one of an important optimization tasks
Article history: and operational decision which provides an economic condition for
Received 29 December 2013 power systems. This paper presents overview of economic load dispatch
Received in revised form problems and solution methodologies. MATLAB programming of
10 January 2014 different methods used for solving economic load dispatch problem was
Accepted 20 January 2014 done and results are given in tabular form. Lambda iteration method is
Available online 1 February 2014 implemented for three cases of
Without generation limits and losses
Keywords
With generation limits and without losses and
Economic Load Dispatch
With generation limits and losses. Newtons and the dynamic
Fuel Cost
programming methods were put into action for coding.
1. Introduction optimally out of the available generating sources to operate to
meet the expected load and provide a specified margin of
The efficient and optimum economic
operation of electric power systems has
always occupied an important position in Corresponding Author,
electric power industry. In recent decades, it is E-mail address: [email protected] All rights
becoming very important for utilities to run reserved: http://www.ijari.org operating reserve over a
their power systems with minimum cost while specified period of time. The second aspect of ELD
satisfying their customer demand all the time is on-line economic dispatch where in it is required
and trying to make profit. With limited to distribute the load among the generating units
availability of generating units and the large actually parallel with the system in such a manner as to
increase in power demand, fuel cost and supply minimize the total cost of supplying power. In case of
limitation, the committed units should serve ELD, The generations are not fixed but they are allowed
the expected load demand with the changes in to take values again within certain limits so as to meet a
fuel cost and the uncertainties in the load particular load demand with minimum fuel
demand forecast in all the different time consumption [5], [6].
intervals in an optimal manner[1],[2],[3]&[4].
The basic objective of ELD of electric 2. Problem Formulation
power generation is to schedule the The ELD problem is defined as to minimize the total
committed generating unit outputs, so as to meet operating cost of a power system while meeting the total
the load demand at minimum operating cost load plus transmission losses within generator limits.
while satisfying all unit and system equality and Mathematically the problem is defined as (including losses) by
inequality constraints. The ELD problem has equation (1) given as
been tackled by many researchers in the past. [1]
ELD problem involves different problems. The Minimize:
first is Unit Commitment or pre-dispatch Fi (Pi) = ( + + ) (1)
problem where in it is required to select

ICARI
 ISBN 978-93-5156-328-0

International Conference of Advance Research and Innovation (ICARI-2014)

Subject to (1) the energy balance equation ( ) ( ) (3)
= PD+PL Where , , are cost coefficients , PD is load
(2) demand, Pi is real power generation, PL is power transmission
(2) the inequality constraints loss and is number of generation busses.

202
One of the most important, simple but 2 + = (1 2 2 )
approximate method of expressing transmission loss (i=1, 2, 3 )
as function of generator powers is through
Bcoefficients. This method uses the fact that under 2( + ) +2 =( )
normal operating condition, the transmission loss is (i=1, 2, 3 ) (13)
quadratic in the injected bus real power. The general The above linear eqn(13) can be solved to obtain the value of
form of the loss formula using B-coefficient is given Pi if is known.
by eqn(4)
= MW (4) 3. Lambda Iteration Method
th th
Where Pi, Pj are real power injection at the i , j The solution to this problem can be approached by
buses and Bij are loss coefficients which are constant considering a graphical technique for solving the
under certain assumed conditions [8], [9], & [11]. The problem and then extending this into the area of
above loss formula is known as the Georges formula. computer algorithms. The lambda-iteration procedure
Another more accurate form of transmission loss converges very rapidly for this particular type of
expression, frequently known as the Kronlossf optimization problem. The actual computational
ormula is eqn (5) procedure is slightly more complex [1], [12].
We use following MATLAB
= + + MW
code formulated for
(5) (Rs/MW h) Equation (9) is known as the exact
where, B00, Bi0, and Bij are the loss coefficient which coordination equation and
are constant under certain assumed conditions. The above
constrained optimization problem is converted
into an unconstrained optimization problem. For inclusion of losses:
Lagrange multiplier method is used in which a function ( ) = PD + PL - =0 (9)
minimized (or maximized) is subjected to side conditions in By differentiating the transmission loss eqn (5)
the form of equality constraints. Using Lagrange
with respect to Pi, the incremental transmission
multipliers, an augmented function is defined by eqn
(6) loss can be obtained as

L(Pi ) = F(Pi) + (PD - PL - ) (6) = 2 (10)

Where, is the Lagrangian multiplier. Necessary conditions for and by differentiating the cost function of Eqn
the optimization problem are (1) with respect to Pi , the incremental cost can
( ) ( ) be obtained as
= + ( 1) = 0, (i =1, 2 ) (7) ( )

=2 + , (i=1, 2, 3 ) (11)
Rearranging the above equation
( ) to find the solution 2 + = (1 2 )
= (1 , (i=1, 2, 3 ) (8) (i=1, 2, 3 ) (12)
Rearranging the eqn(12) to get P i, i.e.
th
( ) no losses and no generation losses as:
Where, : Incremental cost of the i generator

203
ICARI
 ISBN 978-93-5156-328-0

International Conference of Advance Research and Innovation (ICARI-2014)

for i=1:3 n(i)=(f(i,2)/(2*f(i,3))); Now, if the g function is the gradient vector , then
p(i)=(1/(2*f(i,3))); end m=sum(n);
q=sum(p); lambda = x = - . L
((Demand+m)/q); disp(lambda) for For economic dispatch problem this takes the form
i=1:3 eqn(16):
P(i)=((lambda-f(i,2))/(2*f(i,3))); end P =()+() (16) and is as it was
for i=1:3 defined before. The Jacobian matrix now becomes
totalcost=(f(i,1)+(f(i,2)*P)+(f(i,3)*P*P)); end one made up of second derivatives and is called the
Ploss = 0.00003*P Hessian matrix:
(1,1)^2+0.00009*P(1,2)^2+0.00012*P(1,3)^2
for j=1:4 for i=1:3 n(i)=f(i,2)/(2*f(i,3));
w(i)=(1-(2*PL(1,i)*P(1,i)))/(2*f(i,3));
end m=sum(n);
q=sum(w);

lambda=(PD+Ploss+m)/q
=
for i=1:3

P(i)=(lambda*(1-2*PL(1,i)*P(1,i))-f(i,2))/(2*f(i,3));
end j=j+1; end

disp(P)

4. Newtons Method
The economic dispatch can also be solved by Generally, Newton's method will solve for the
observing that the aim is to always drive correction that is much closer to the minimum
=0 (14) generation cost in one step than would the gradient
Since this is a vector function, the problem can be method.
formulated as one of finding the correction that exactly drives We use following MATLAB code formulated as:
the gradient to zero (i.e., to a vector, all of whose elements
are zero). Newton's method can be used to find this. Newton's
method for a function of more than one variable is developed Alpha=X;
as follows [1], [13]&[14]. Beta =Y;
Suppose for the function g(x) to be driven to zero. The Gamma =Z;
function g is a vector and the unknowns, x, are also vectors. Lambda=input ('Enter the value of lambda :');
Then, to use Newton's method eqn(15): P_load=A
g(x+ x) = g(x) + [g'(x)]x = 0 (15) If the P_sum=sum(P_initial(1)+P_initial(2)+P_initial(3)
function is defined as: ) %......Formation of Hessian Matrix...... for i=1:3
( , , ) L(i)=beta(i)+2*gamma(i)*P_initial(i); end
dellambda=[L(1)-Lambda;L(2)-
()= ( , , )
Lambda;L(3)Lambda;P_load-P_sum]
( , , ) G=2*gamma;
then H= [G(1) 0 0 -1;0 G(2) 0 -1;0 0 G(3) -1;-1 -1 -1 0]
M= inv(H) delp_Power=-(M*dellambda)
Lambda_optimal=delp_Power(4)
P_optimal=(P_initial')+delp_Power;

5. Dynamic Programming Method

Which is the familiar Jacobian matrix. The adjustment at each The application of digital methods to solve a wide
step is then x = - [g'(x)]-1 g(x) variety of control and dynamics optimization
problems in the late 1950s led Dr. Richard Bellman

204
ICARI
 ISBN 978-93-5156-328-0

International Conference of Advance Research and Innovation (ICARI-2014)

and his associates to the development of dynamic Then the units must be scheduled to minimize the cost
programming. These techniques are useful in solving a variety to deliver power over the time period as:
of problems and can greatly reduce the computational effort in = ( ) (18)
finding optimal trajectories or control policies. The theoretical
Subject to: = =1 (19)
mathematical background, based on the calculus of variations,
And = +
is somewhat difficult. The applications are not, however, since
they depend on a willingness to express the particular With
optimization problem in terms This optimization problem can be solved with
appropriate for a dynamic-programming (DP) formulation [1]. dynamic programming and the control performance
In the scheduling of power generation systems, DP of the dispatch will be considerably better than that
techniques have been developed for the economic dispatch of using dynamic programming and no ramp limit
thermal systems, the solution of hydrothermal economic- constraints.
scheduling problems and the practical solution of the unit
commitment problem. If valve points are considered in the 6. Result And Discussions
input-output curve, the possibility of non-convex curves must Lambda iteration method is implemented on 3
be accounted for if extreme accuracy is desired. If nonconvex cases of without losses and generation limits, with
input-output curves are to be used, an equal incremental cost losses and with generation limits. Newtons method
methodology cannot be used since there are multiple values of was used and the dynamic programming method for 1
MW output for any given value of incremental cost. m/c, two m/c and 3 m/c was used. The programs were
Under such circumstances, there is a way to find an written on MATLAB 9 and implemented on Intel Core
optimum dispatch which uses dynamic programming (DP). 2 Duo having 2.4 GHZ 3GB RAM.
The dynamic programming solution to economic dispatch is
done as an allocation problem. Using this approach, not just a 6.1 Lambda Iteration method
single optimum set of generator MW outputs is calculated for MATLAB promming of three kind of economic
a specific total load supplied-rather a set of outputs are dispatch problem was done namely without losses and
generated, at discrete points, for an entire set of load generation limits, with generation limits and with
values[16]. One problem that is common to economic dispatch losses. Problem taken from Wood and Wollenberg
with dynamic programming is the poor control performance of [1], p 36 given as,
the generators. When the economic dispatch is to be done with The heat rates for three thermal plants in MBtu/h
dynamic programming and the cost curves are nonconvex, a are given as: H1 = 510.0 +7.2 P1 + 0.00142 P12, H2 =
difficult problem is encountered whenever a small increment 310.0 + 7.85P2 + 0.00194 P22, H3 = 78.0 + 7.97 P3 +
in load results in a new dispatch that calls for one or more 0.00482 P32. With the following fuel costs: Unit 1 =
generators to drop their output a great deal and others to 1.1 Rs/MBtu, Unit 2 = 1.0 Rs/MBtu, Unit 3 = 1.0
increase a large amount. The resulting dispatch may be at the Rs/MBtu and total load of 850MW.Results obtained
most economic values as determined by the DP, but the control are summarized in TABLEI.
action is not acceptable and will probably violate the ramp rates In TABLE II, lambda iteration method is applied
for several of the units. on economic load dispatch problem with generation
The only way to produce a dispatch that is acceptable to limits whose problem considered is given by-
the control system, as well as being the optimum economically, The fuel cost functions for 3 thermal plants in
is to add the ramp rate limits to the economic dispatch Rs/hr are, C1 = 500 + 5.3 P1 + 0.004 P12, C2 = 400 +
formulation itself. This requires a short-range load forecast to 5.5 P2 + 0.006 P22, C3 = 200 + 5.8 P3 + 0.009 P32, the
determine the most likely load and load-ramping requirements total load PD, is 975 MW with the following generator
of the units. This problem can be stated as follows [17], [18]. limits (in MW): 200 P1 450, 150 P2 350, 100
Given a load to be supplied at time increments t = P3 225.
1 . . . , with load levels of and N generators on-line to supply In TABLE III, lambda iteration method is applied
the load: on economic load dispatch problem with losses
= (17) considered whose problem considered is same as
Each unit must obey a rate limit such that: problem one with the loss expression given by P L =
= + 0.00003 P12+ 0.00009 P22+ 0.00012 P32.
And

205
ICARI
 ISBN 978-93-5156-328-0

International Conference of Advance Research and Innovation (ICARI-2014)

_ delP dellambda Total Cost

2 -915.2778 1890.3 7.1632 9.4000 450 325 200 8236.3

7 404.1667 570.8333 2.1632 9.4000 450 325 200 8236.3
8 668.0556 306.944 1.1632 9.400 450 325 200 8233
Table: 1. Economic Load Dispatch result for Lambda Iteration Method
Parameter Value
Ploss 15.600
Lambda 9.5279
1 435.0215
2 299.8997
3 130.6788
Ploss 15.8211
Table: 2. Economic Load Dispatch result for Lambda Iteration Method

6.2 Newtons Method

MATLAB priming of Newtons method was done. Problem was taken as, The fuel cost functions for 3 thermal
plants in Rs/hr are C1 = 561 + 7.92 P1 + 0.001562 P12, C2 = 310 + 7.85 P2 + 0.001940P22, C3 = 78 + 7.97 P3 + 0.00482
P32 with total load varied.

1 800 450 325 200 975 9.0749 369.68 315.69 114.61

0 800 300 200 300 800 9.0749 369.68 315.69 114.61

8 600 250 100 150 500 8.78 275.75 140.08 84.17
4 900 400 300 200 900 9.22 416.65 353.51 129.83
8 900 300 250 350 900 9.22 416.65 353.51 129.83
Table: 3. Economic Load Dispatch result for Newton Method

6.3 Dynamic Programming Method

MATLAB priming of dynamic programming method was done. Problem taken from Wood and Wollenberg [1], p
51 given as, There are three units in the system; all are on-line. Their input-output characteristics are neither smooth
nor convex. Data are as follows.
Power Levels (MW) Costs (Rs/hour) Costs (Rs/hour) Costs (Rs/hour)
P1 = P2 = P3 F1 F2 F3
0
50 810 750 806
75 1355 1155 1108.5
100 1460 1360 1411
125 1772.5 1655 11704.5
150 2085 1950 1998
175 2427.5 2358

206
ICARI
 ISBN 978-93-5156-328-0

International Conference of Advance Research and Innovation (ICARI-2014)

200 2760
225
In this problem, 10000 have been taken in place of infinity
Demand Cost P1 P2 P3
300 4168 150 100 50
250 3558 150 50 50
200 2971 100 50 50
250 3558 150 50 50
Table: 4. Economic Load Dispatch result for Dynamic Programming Method
7. Conclusion lambda method always requires that one be able to find
the power output of a generator, given an incremental
This paper gives overview of economic load
cost for that generator. In cases where cost function is
dispatch problems and solution methodologies.
much more complex, method like gradient and Newton
Implementation is done using MATLAB programming
can be used. If non convex input-output curves are to
and results are given in tabular form. Conventional
be used dynamic programming can be used to solve
method like lambda iteration method converges rapidly
economic
but complexities increases as system size increase also
dispatch problems. Hence different methods have different applications.

References

207
ICARI
 ISBN 978-93-5156-328-0

International Conference of Advance Research and Innovation (ICARI-2014)

[1] J. Wood and B. F., Wollenberg, Power Apparatus and Systems, PAS-93. pp 1292-1297,
generation, operation and control, New York: 1974
John Wiley Inc., 1984. [10] H. H. Happ. "Optimal Power Dispatch," IEEE
[2] R. Balamurugan and S. Subramanian, An Transactions on Power Apparatus and Systems,
Improved Dynamic Programming Approach to PAS-93, pp 820-830, No. 3, May I June 1974.
Economic Power Dispatch with Generator [11] D. I. Sun, B. Ashley, B. Brewer, A. Hughes and
Constraints and Transmission Losses, Journal of W.F. Tinney, "Optimal Power Flow by Newton
Electrical Engineering & Technology, 3 (3), pp. Approach," IEEE Transactions on Power
320~330, 2008 Apparatus and Systems, PAS-103, pp 28642880,
[3] G. Sreenivasan, C. H. Saibabu, S. Sivanagaraju, 1984.
Solution of Dynamic Economic Load Dispatch [12] K.S. Hindi, M. R. Ab Ghani, Multi-period
(DELD) Problemwith Valve Point Loading economic dispatch for large scale power systems,
Effects and Ramp Rate Limits Proc. IEE Pt. C 136 (3) (1989) 130 136.
Using [13] X. S. Han, H.B. Gooi, D. Kirschen, Dynamic
PSOInternational, Journal of Electrical and economic dispatch: feasible and optimal
Computer Engineering (IJECE), 1(1), September solutions, IEEE Trans. Power Syst. 16 (1) (2001)
2011 2228.
[4] H. H. Happ, Optimal power dispatch a [14] D. Attaviriyanupap, H. Kita, E. Tanaka, J.
comprehensive survey, IEEE Trans. Power Hasegawa, A hybrid EP and SQP for dynamic
Apparatus Syst. PAS-96 (3) (1977) 841854. economic dispatch with nonsmooth incremental
[5] B.H. Chowdhury, S. Rahman, A review of recent fuel cost function, IEEE Trans. Power Syst. 17
advances in economic dispatch, IEEE Trans. (2) (2002) 411416.
Power Syst. 5 (4) (1990) 12481259. [15] M. Basu, Particle swarm optimization based
[6] W.G. Wood, Spinning reserve constraints static goal-attainment method fordynamic economic
and dynamic economic dispatch, IEEE Trans. emission dispatch, Elect. Power Components
Power Apparatus Syst. PAS-101 (2) (1982) 338 Syst. 34 (2006) 10151025.
1338. [16] M. Basu, Dynamic economic emission dispatch
[7] C.B. Somuah, N. Khunaizi, Application of linear using evolutionary programming and fuzzy
programming redispatch technique to dynamic satisfied method, Int. J. Emerging Elect. Power
generation allocation, IEEE Trans. Power Syst. 5 Syst. 8 (4) (2007) 115.
(1) (1990) 2026. [17] M. Basu, Dynamic economic emission dispatch
[8] X.S. Han, H.B. Gooi, Effective economic using nondominted sorting genetic algprthim-II,
dispatch model and algorithm, Elect. Power Elect. Power Energy Syst. 30 (2008) 140149 [18]
Energy Syst. 29 (2007) 113120. P. Attaviriyanupap, H. Kita, E. Tanaka, J.
[9] T.E. Bechert, H.G. Kwatny, On the optimal Hasegawa, A fuzzy-optimization approach to
dynamic dispatch of real power, IEEE Trans. dynamic economic dispatch considering
Power Apparatus Syst. PAS-91 (1972) 889898. uncertainties, IEEE Trans. Power Syst. 19 (3)
Flow Solution Using Lagrangian Multipliers and (2004) 12991307.
the Hessian Matrix.'' =E Transactions on Power

208
ICARI