You are on page 1of 10

688

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 18, NO. 2, MAY 2003

Comparison and Application of Evolutionary Programming Techniques to Combined Economic Emission Dispatch With Line Flow Constraints
P. Venkatesh, R. Gnanadass, and Narayana Prasad Padhy

AbstractEconomic load dispatch (ELD) and economic emission dispatch (EED) have been applied to obtain optimal fuel cost and optimal emission of generating units, respectively. Combined economic emission dispatch (CEED) problem is obtained by considering both the economy and emission objectives. This biobjective CEED problem is converted into a single objective function using a price penalty factor approach. A novel modified price penalty factor is proposed to solve the CEED problem. In this paper, evolutionary computation (EC) methods such as genetic algorithm (GA), micro GA (MGA), and evolutionary programming (EP) are applied to obtain ELD solutions for three-, six-, and 13-unit systems. Investigations showed that EP was better among EC methods in solving the ELD problem. EP-based CEED problem has been tested on IEEE 14-, 30-, and 118-bus systems with and without line flow constraints. A nonlinear scaling factor is also included in EP algorithm to improve the convergence performance for the 13 units and IEEE test systems. The solutions obtained are quite encouraging and useful in the economic emission environment. Index TermsCombined cycle cogeneration plant, combined economic emission dispatch, economic load dispatch, evolutionary programming, price penalty factor.

I. INTRODUCTION HE main objective of Economic Load Dispatch (ELD) is to minimize the fuel cost while satisfying the load demand. The classical lambda iteration method has been used to solve ELD problem. This method has used equal increment cost criterion for systems without transmission losses and penalty facmatrix for considering the losses. Other methods tors using such as gradient, Newton, linear programming and interior point have also been applied to solve ELD problem [1]. In all these methods, the fuel cost function is chosen to be of quadratic form. However, the fuel cost function becomes more nonlinear when valve point loading effects are included [2]. Recently, combined cycle cogeneration (CCCP) plants have shown their importance in both developing and developed countries in order to improve the efficiency of generation.
Manuscript received October 9, 2002. P. Venkatesh is with the Department of Electrical and Electronics Engineering, Thiagarajar College of Engineering, Madurai, 625 015, India (e-mail: venkatesh_paramasivam@yahoo.com). R. Gnanadass is with the Department of Electrical and Electronics Engineering, Pondicherry Engineering College, Pondicherry, 605 014, India. (e-mail:gnanadass@yahoo.com). N. P. Padhy is with the Department of Electrical Engineering, Indian Institute of Technology, Roorkee, 247 667, India. (e-mail: nppeefee@iitr.ernet.in). Digital Object Identifier 10.1109/TPWRS.2003.811008

The Eastern region of Electricity Generating Authority of Thailand (EGAT) has a large amount of thermal units including CCCP. From the survey the fuel cost characteristic of the CCCP is of nonsmooth and nondifferentiable [3], [4]. Thus, the nonlinearity in the ELD problem is further increased. Gent and Lamont have started the early work on minimum emission dispatch [5]. Optimal power-dispatch problem, considering practical constraints such as transformer taps, voltage, and line-flow constraints has been solved by Fletchers quadratic programming method [6]. Nanda, Hari, and Kothari made an attempt to explore the feasibility of developing a classical technique based on coordination equations to solve economic emission load dispatch with line flow constraints [7]. Chen and Chen proposed a fast Newton-Raphson method for a multiobjective power dispatch problem with line flow constraints [8]. Hota et al. proposed a sequential quadratic programming technique to solve CEED problem by assigning weighting factors for generation and emission cost functions [9]. The line flows are computed by developing distribution factors using sensitivity information from the Jacobian elements of load flow [5][9]. Researchers proposed a price penalty factor for solving the CEED problem which blends the emission costs with the normal fuel costs [10]. Numerical optimization techniques and EC methods can play a key role in future for solving the ELD problem [11]. EC technique consists of three main areas such as genetic algorithms (GAs), evolution strategies (ES), and evolutionary programming (EP). Each of these methods represented their data structures in binary (or) real coded (or) in both forms. Operators such as crossover and mutation are also used in EC techniques [12], [13]. GA and simulated annealing method have been applied to solve the complex nonlinear ELD problem, as these methods are independent of the shape of fuel cost functions [14]. Different forms of GA methods have also applied to the ELD problem of a 13-unit test system with valve point loading. Researchers have applied GA and EP methods to the ELD problem having three units test system [2], [14], [15]. The fuel cost functions for the three units system were considered as quadratic form, and the losses were ignored. Research work was carried out for the solution of ELD problem using genetic algorithm for the same three units system by assuming one unit as a combined cycle cogeneration plant (CCCP) [16]. But the algorithm required large number of iterations/generations when the system has large number of

0885-8950/03$17.00 2003 IEEE

VENKATESH et al.: COMPARISON AND APPLICATION OF EVOLUTIONARY PROGRAMMING TECHNIQUES

689

units. In order to minimize the number of generations and avoid the loss of useful chromosome for further generation micro GA (MGA) was developed [4], [17]. Yang et al. have also applied EP to ELD problem containing nonsmooth fuel cost functions [15]. In general EP, mutation probability is fixed throughout the whole search processing. However, in practical application, fixed mutation results in premature condition. L. L. Lai and Ma proposed the EP based reactive power flow problem having mutation changing from higher value to lower value with a decrease in step size depending on the maximum generation number [18]. Wong proposed a cooling mutation technique in EP algorithm to solve CEED problem for nine-units system [19]. Yuryevich et al. validated EP algorithm to solve optimal power flow problem with quadratic and sine component cost functions [20]. K. Y. Lee et al. has attempted the applicability of the three EC methods and compared among them to the optimal reactive power planning problem [21]. This paper presents the application of GA, MGA and EP methods to ELD problem for three, six and thirteen units systems having nonsmooth and nondifferentiable fuel cost functions. The comparison among these methods and comparison with conventional methods are presented. A novel modified price penalty factor is introduced to find the exact economic emission fuel cost with respect to the load demand. In this paper, EP algorithm is proposed to solve the CEED problem with line flow constraints. The line flows in MVA are computed directly from the Newton Raphson method. The proposed EP algorithm is demonstrated to solve the CEED problem for IEEE 14, 30 and 118 bus systems. In the proposed EP algorithm, mutation is changing nonlinearly with respect to the number of generations to avoid premature condition. So the nonlinear scaling factor is introduced in EP algorithm and investigated on the thirteen units system with valve point loading and IEEE test systems. II. PROBLEM FORMULATION Optimization of cost of generation has been formulated based on classical ELD with emission and line flow constraints. The detailed problem is as follows. For a given power system network, the optimization cost of generation is given by the following equation: (1) where optimal cost of generation; and total fuel cost and total emission of generators, respectively; represents the number of generators connected in the network. The cost is optimized with the following power system constraint (2)

where real power generation of th generator; total load of the system; transmission losses of the system. The power flow equation of the power network (3) where

For each

bus

For each bus not including the ref.bus where and

are, respectively, calculated real and reactive bus ; power for and are, respectively, specified real and reactive bus ; power for and are, respectively, calculated and specified real bus ; power for and voltage magnitude and phase angles of different buses. of The inequality constraint on real power generation each generation (4) and are, respectively, minimum and where maximum value of real power allowed at generator . bus The inequality constraint on voltage of each (5) and are, respectively, minimum and where maximum voltage at bus . Power limit on transmission line (6) is the maximum rating of transmission where line connecting bus and . Total fuel cost of generation FC in terms of control variables generator powers can be expressed as hr where real power output of an th generator; , , fuel cost curve coefficients; Total emission of generation EC can be expressed as lb/hr where , and are emission coefficients. (8) (7)

690

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 18, NO. 2, MAY 2003

The biobjective combined economic emission dispatch problem is converted into single optimization problem by introducing price penalty factor [10] as follows: Minimize hr (9)

Subject to the power flow constraints of (2)(6). The price penalty factor blends the emission with fuel cost and is the total operating cost in U.S dollars per hour. The price penalty factor is the ratio between the maximum fuel cost and maximum emission of corresponding generator lb (10)

The following steps are used to find the price penalty factor for a particular load demand. 1) Find the ratio between maximum fuel cost and maximum emission of each generator. 2) Arrange the values of price penalty factor in ascending order. one at 3) Add the maximum capacity of each unit a time, starting from the smallest unit until . 4) At this stage, associated with the last unit in the process is the price penalty factor for the given load. The procedure just shown gives the approximate value of price penalty factor computation for the corresponding load deis introduced mand. Hence, a modified price penalty factor in this paper to give the exact value for the particular load demand. The first two steps of computation remain same for the calculation of modified price penalty factor. Then, it is calculated by interpolating the values of corresponding to their load demand values. III. EP-BASED COMBINED ECONOMIC EMISSION DISPATCH (EPCEED) Classical nonlinear programming methods suffer in achieving the global optimum solutions due to the presence of nonsmooth (valve point loading) and nondifferentiable (CCCP) functions. Hence, EC methods are exploited to obtain the global optimum solutions. Natural evolution is a population-based optimization process. Simulating this process on a computer becomes evolutionary computation. The application of GA methods is in need of encoding, decoding process, and fitness function formulation. Series binary coding is used in this paper. Exponential scaling is used in formulation of fitness function [16]. To improve the performance of simple GA, micro genetic algorithm (MGA) is proposed in this paper. It employs a deterministic sampling mechanism to create the sampling space for selection process. In this approach, truncation selection procedure is adopted where a threshold number of best strings (selection of subpopulation) are truncated according to their survival probabilities from the enlarged sample space (Np). The subpopulation size (SNp) is obtained by executing a GA using a population size (Np) for a fewer number generations [10% of maximum number of generations (Nm) by satisfying the (2)]. The truncated parent strings enter in to genetic operations such as crossover and mutation.

The resulting strings are compared with the parent strings and the best ones are retained for the next generation. This new technique is coined as Elitism. It guarantees that at least the best individual of the last generation is selected. The further information regarding MGA is also available in [17]. Crossover and mutation operations are performed in case of GA and MGA. Among EC methods, EP is better in obtaining global optimum, which relies on mutation rather than crossover. Due to inherent flexibilities in fitness function and ease in coding, EP method produces the best solution with less number of generations. Evolutionary programming is a probabilistic search technique, which generates the initial parent vectors distributed uniformly in intervals within the limits and obtains global optimum solution over number of iterations. The main stages of this technique are initialization, creation of offspring vectors by mutation and competition and selection of best vectors to evaluate best fitness solution. The implementation of EP algorithm is given below. A. Initialization The initial population ( number of parent vectors) is generated after satisfying the constraints given in (2) and (4).The are the real power outputs of elements of parent vectors generating units distributed uniformly between their minimum and maximum limits. Slack bus generator vector is calculated using Newton Raphson method for the generations shown before. B. Mutation is created from each parent vector by An offspring vector adding Gaussian random variable with zero mean and standard deviation , denoted as for where (12) where scaling factor fuel cost of the th generator and minimum value of the fuel cost in the generation The created offspring vector must satisfy the minimum and maximum generation limits of the units and line flow constraints. The slack bus generator vector for the above offspring vector generations are calculated using Newton Raphson method. The constrained optimization CEED problem is converted into unconstrained optimization problem using Penalty Factor (PF) as given in the (13). This becomes the fitness function in EP method Fitness function (13) (11)

The second term in (13) becomes zero during initialization and its gets a nonzero value after mutation only if generator

VENKATESH et al.: COMPARISON AND APPLICATION OF EVOLUTIONARY PROGRAMMING TECHNIQUES

691

vectors violate its limits. Therefore, only becomes the fitness function and is computed for the offspring vectors similar to the parent vectors. Infeasible EP solutions can be produced if the mutation is high. This also contributes to the long execution time and poor convergence of EP algorithm when the mutation is very large is too (above 1). Conversely, if the value of scaling factor small, the EP algorithm converges rapidly initially but fails to converge completely to the global optimum in later generations. The selection of value of scaling factor must therefore be such that it provides a suitably high initial mutation level and does not cool too rapidly so as to present complete convergence to the global optimum [19]. Normally, constant scaling factor is used in conventional EP. In the proposed EP method, step and nonlinear scaling factor are used to obtain the best solution. The for step size scaling factor is evaluated as decrement step follows: (14) is the maximum number of generations. where In this paper, nonlinear scaling factor is also proposed. The following steps are used to compute the decrements of nonlinear scaling factor. For the first 30% of total number of generations the decrement of the scaling factor is computed as follows: (15) is the midpoint of scaling factor range. where For the remaining 70% of the total number of generations , the decrement of the scaling factor is given by (16)

C. Competition and Selection and the corresponding offspring The parent trial vectors ( number of vectors) compete with each other for survival within the competing pool. The competition is based on with the corresponding cost the cost of parent vectors in this population. The best vector of offspring vectors (or) offspring having minimum cost, whether parent vector is selected for the new parent for the next generavector tion. Initialization and mutation are repeated until there is no appreciable improvement in the fitness value. The step-by-step computational flowchart is given in Fig. 1. IV. SIMULATION RESULTS AND DISCUSSION Solutions for ELD problem were obtained for three- and sixunits system with quadratic fuel cost function. Two units system with CCCP and 13 units system with valve point loading were solved by conventional lambda iteration, dynamic programming and EC methods. Computer programs were developed for GA and MGA to solve ELD problem [16]. To determine the effectiveness of EC methods, the results of the conventional methods were compared with EC methods such as GA, MGA, and EP.

Fig. 1. Flowchart for the proposed EP method.

Among the EC methods, EP was found to be better in solving the ELD problem. In the proposed EP-based CEED approach, Newton Raphson method and EP algorithm have been used for

692

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 18, NO. 2, MAY 2003

TABLE I SOLUTION OF THREE-UNITS SYSTEM

TABLE II SOLUTION OF SIX-UNITS SYSTEM

power flow and economic dispatch, respectively. The simulation studies were carried out on PIII 700-MHz system in Matlab environment. A. Three-Units System The cost coefficients and generation limits of three units system are taken from [2]. Transmission loss for this system is matrix. ELD solution for the three-units calculated using system is solved using conventional technique (lambda-iteration), evolutionary algorithms such as GA, MGA, and EP and the results are shown in Table I. The best solution of EC techniques and lambda-iteration method are found to be the same for various load demands. B. Six-Units System The cost coefficients, generation limits, and transmission loss matrix for this system are taken from [10]. ELD sousing lution for the six units system is solved using lambda-iteration

method, evolutionary algorithms, and the results are shown in Table II. The difference in best solution of GA, MGA, and EP are approximately U.S.$ 10/h, U.S.$ 1.5/h, and U.S.$ 1.0/h, respectively. So the performance of EP is found to be better than GA and MGA. The parameters used in EC methods for the three- and six-units systems are given in Table III. The population size (Np) and number of generations (Nm) are the parameters will decide the execution time for EC methods. From the results, it is inferred that, Np and Nm selected in EP are less when compared with GA, GA [15], and MGA. In comparison with EP [15], these values are less for obtaining best solution, because the selection of for three-units system. Hence, the minimum solution is obtained with less number of generations. Np and Nm used in EC methods are doubled (or) more for six-units system than the three-units system in obtaining the best solution. These two values are less in EP compared to GA. Thus, EP is giving better solution with less execution time than GA and MGA for both the three- and six-units systems.

VENKATESH et al.: COMPARISON AND APPLICATION OF EVOLUTIONARY PROGRAMMING TECHNIQUES

693

TABLE III PARAMETERS USED IN EC METHODS FOR THREE- AND SIX-UNITS SYSTEMS

THE

Fig. 2.

Fuel cost characteristic of CCCP. TABLE IV SOLUTION OF THREE UNITS SYSTEM INCLUDING CCCP

C. Three Units System Including CCCP In this case, the first two units are the same as three-units system and the third unit is replaced with combined cycle cogeneration plant (CCCP). In CCCP, gas and steam turbines are working in combination to generate electric power. CCCP has two 75-MW gas turbine units and one 50-MW steam turbine unit. As an example, the modeling of the plant in two units system with CCCP is carried out by the following descriptions: Third generator of this system is operated between 50 and 200 MW. This generator is assumed to be combined cycle type with two 75-MW gas turbines and one 50-MW steam turbine. Five points of 50, 82.75, 93.75, 176.625, and 200 MW from original quadratic curve values are fixed and used for creating CCCP cost characteristics. The first gas turbine is operating in an open cycle till it reaches 85% of its capacity (63.75 MW). During 50-to-63.75-MW power generation, this gas turbine is operated in the open cycle. Thus, linear region AB is formulated. The exhaust heat produced by the gas turbine is enough to operate the steam turbine. Hence, during 63.75 to 82.875 MW of power generation, steam turbine is allowed to operate without any additional fuel. It results in a constant region BC in the fuel cost characteristic. During 82.87593.75 MW of generation, both the gas and steam turbines are operated in a closed cycle (CD region). Similarly, the operation of steam turbine and the second gas turbine result in regions DE, EF, and FG. The fuel costs of CCCP are computed by substituting generator powers in its quadratic fuel cost functions. The cost characteristic of CCCP is modeled as shown in Fig. 2. The cost characteristics of CCCP is simplified in (17), it has a linear and constant (nondifferentiable) region. hr Linear region hr Constant region

Transmission loss is also calculated using matrix. EC and dynamic programming (DP) methods were applied as classical methods fail to give global minima due to the nondifferentiable term in the fuel cost characteristic of CCCP. The best roundoff solution of DP and EC techniques such as GA, MGA, and EP are given in Table IV. From the results, it is inferred that, the optimal solution of EP is better than any other methods and results in a saving of approximately U.S.$ 50/h against DP. D. Thirteen-Units System With Valve Point Loading The cost coefficients and generation limits for the 13-units system with valve point loading are taken from [14]. It is difficult to compare the solution obtained using EP with conventional methods as they fail to handle valve point loading. A comparison was made among evolutionary computation methods such as GA, EP, EP* (EP with nonlinear scaling factor) and the results are given in Table V. The load demand is 2520 MW and the losses are neglected. From the Table V, it is inferred that EP and EP* are giving better

(17)

694

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 18, NO. 2, MAY 2003

TABLE V SOLUTION OF13 UNITS SYSTEM WITH VALVE POINT LOADING

TABLE VI PARAMETERS USED IN EC METHODS FOR THE 13-UNITS SYSTEM WITH VALVE POINT LOADING

solutions when compared with various forms of GA [14]. In addition both EP and EP* result in a saving of U.S.$ 10/h. The parameters used in EC methods for 13-units system with valve point loading are given in Table VI. The population size for EP* was chosen as half of population size in EP. Due to the use of nonlinear scaling factor (0.20.005) population size in EP* was reduced to 250 instead of 500 in case of EP. However, the population size of EP* is higher than GA [14] for the same number of iterations. E. IEEE Test Systems In the proposed approach, the minimum generation and cost of the generating units were obtained using evolutionary programming with and without violating transmission constraints. The line flows were computed using Newton Raphson method. The cost and emission coefficients of IEEE 14- and 30-bus systems are same which are taken from and for IEEE 118-bus system is given in Appendix. Table VII provides a comparison of combined economic emission load dispatch results obtained by various optimization methods for two test systems with and without line flow constraints. The results of proposed EP approach are in close with classical and sequential quadratic programming methods. The line flows in MVA of best generation schedules for IEEE 14-bus system were shown in Table VIII. The star marked line was over loaded with economic generation schedule when the line flow constraints were not considered. Similarly, the line flow violations were incorporated for IEEE 30- and 118-bus systems. For implementing the EP algorithm, population size of 50 was taken and the maximum number of generations was taken as 100. The minimum solution was obtained for 100 trial runs. The developed EPELD algorithm provides nonlinear scaling factor (0.50.005) based mutation that prevents premature solution. Fig. 3 shows the convergence characteristics of EP algorithm having constant, step, and nonlinear scaling factor

for IEEE 30-bus system. EP with constant scaling factor fails to convergence to the local optimum value at the lower number of generations (less than 50). It was clearly shown that there is no rapid change in fuel cost function value after 50 generations. The combined economic emission dispatch was performed for IEEE 14-, 30-, and 118-bus systems having system load as 259, 283, and 3668 MW, respectively. Table IX summarized the minimum solution obtained by evolutionary based economic load dispatch (EPELD), economic emission dispatch (EPEED), combined economic emission dispatch (EPCEED), and modified combined economic emission dispatch (EPMCEED) with line flow constraints for IEEE 14- and 30-bus systems. The , total minimum solution includes optimum generations , total fuel cost (Fc), total emission output (Ec), and loss . Results of Table VIII reveal that the total operating cost for IEEE 14- and 30-bus systems price penalty factor remains the same even though their load demands are different. Hence, it gives only the approximate value of total operating cost for their corresponding load demand. In this paper, a approach is proposed to modified price penalty factor give exact total operating cost. The computational procedure of proposed modified price penalty factor for IEEE 14-bus system was explained as follows. The ratio between the maximum fuel cost and maximum emission of three generating units were found and arranged in ascending order

The corresponding maximum limits of generating units are given by

For a load of MW starting from the lowest value unit, the and maximum capacity of the units is added one by one associated with when this total equals or exceeds the load, the last unit in the process is price penalty factor [10]

For

MW,

MW

MW.

VENKATESH et al.: COMPARISON AND APPLICATION OF EVOLUTIONARY PROGRAMMING TECHNIQUES

695

TABLE VII CEED RESULTS (FUEL COST IN $/HR) OBTAINED BY VARIOUS METHODSWITH AND WITHOUT LINE FLOW CONSTRAINTS

TABLE VIII LINE FLOWS WITH AND WITHOUT LINE FLOW CONSTRAINTS OF IEEE-14 BUS SYSTEM

Fig. 3.

EPELD convergence characteristics with constant, step, and nonlinear scaling factorIEEE 30-bus system.

696

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 18, NO. 2, MAY 2003

TABLE IX MINIMUM SOLUTION OF DIFFERENT METHODSIEEE 14- AND 30-BUS SYSTEMS

TABLE X MINIMUM SOLUTION OF DIFFERENT METHODSIEEE 118-BUS SYSTEM

Hence, price penalty factor is determined as 2.6221 for IEEE 14-bus system. Even though the price penalty factor was computed for 259 MW but it gives the value up to 510-MW load demand. So the modified price penalty factor is computed by interpolating the values of for last two units by satisfying the corresponding load demand

V. CONCLUSION The EC algorithms were tested and the results were presented for various test systems. Results showed that EC methods are well suited for obtaining the best solution for fuel cost functions of differentiable, nonsmooth, and nondifferentiable of the test systems. Savings of approximately U.S.$ 50/h and U.S.$ 10/h were obtained by applying EP method for the test system containing CCCP and 13-units system with valve point loading. The combined economic emission dispatch results of two IEEE test systems were compared with those obtained from classical technique and sequential quadratic programming technique to validate the effectiveness of the proposed algorithm. The Mega Volt Ampere (MVA) line flow limits of the test systems were incorporated and the overloaded lines were observed. The validation of modified price penalty factor to solve CEED problem corresponding to the load demands was carried out to obtain exact best solution. The effectiveness of EP with nonlinear scaling factor to obtain best values was illustrated. The performance of the developed algorithm has been demonstrated with practical IEEE 14-, 30-, and 118-bus test systems.

Similarly, the modified price penalty factor for the IEEE 30- and 118-bus systems were computed as 2.1737 and 6.339, respectively, corresponding to their load demands. By following the above procedure, the minimum solution obtained by different methods for IEEE 118 bus system is given in Table X. By incorporating modified price penalty factor approach, the operating cost savings of approximately U.S.$ 426/h, U.S.$ 296/h, and U.S.$ 2512/h were obtained for the three test systems. Hence, it is justified that modified price penalty factor approach gives the exact minimum dispatch solution. The execution time for the proposed method of IEEE 14-, 30-, and 118-bus systems is obtained as 14.12, 52.01, and 224.1 s, respectively.

VENKATESH et al.: COMPARISON AND APPLICATION OF EVOLUTIONARY PROGRAMMING TECHNIQUES

697

TABLE XI COST AND EMISSION COEFFICIENT OF IEEE 118 SYSTEM

APPENDIX COST AND EMISSION COEFFICIENTS OF IEEE-118 SYSTEM See Table XI. REFERENCES
[1] A. I. Wood and B. F. Woolenburg, Power Generation Operation and Control. New York: Wiley, 1996. [2] G. B. Sheble and K. Brittig, Refined genetic algorithm-economic dispatch example, IEEE Trans. Power Syst., vol. 10, pp. 117124, Feb. 1995. [3] Y. H. Song and Q. Y. Xuan, Combined heat and power economic dispatch using genetic algorithm based penalty function method, Int. J. Electr. Mach. Power Syst., vol. 26, pp. 363372, 1998. [4] C. Achyuthakan, Genetic Algorithms Application to Economic Load Dispatch, Master, Asian Institute of Technology, Thailand, 1997. [5] M. R. Gent and J. W. Lamont, MinimumEmission dispatch, IEEE Trans. Power Apparat. Syst., vol. 90, pp. 26502660, June 1971. [6] J. Nanda, D. P. Kothari, and S. C. Srivastava, New optimal power-dispatch algorithm using fletchers quadratic programming method, Proc. Inst. Elect. Eng., pt. C, vol. 136, no. 3, pp. 153161, 1989. [7] J. Nanda, L. Hari, and M. L. Kothari, Economic emission load dispatch with line flow constraints using a classical technique, Proc. Inst. Elect. Eng., pt. C, vol. 141, no. 1, pp. 110, 1994. [8] J. F. Chen and S. D. Chen, Multiobjective power dispatch with line flow constraints using the fast Newton-Raphson method, IEEE Trans. Energy Conversion, vol. 12, pp. 8693, Mar. 1996. [9] P. K. Hota, R. Chakrabarti, and P. K. Chattopadhyay, Economic Emission Load Dispatch With Line Flow Constraints Using Sequential Quadratic Programming Technique, India: Institution of Engineers, 2000, vol. 81, pp. 2125. [10] P. S. Kulkarni, A. G. Kothari, and D. P. Kothari, Combined economic and emission dispatch using improved back propagation neural network, Int. J. Electr. Mach. Power Syst., vol. 28, pp. 3144, 2000.

[11] S. C. Srivastava, Optimal power flow: state-of the art and new challenges, in Proc. Int. Conf. Comput. Applicat. Electr. Eng.-Recent Advances, Roorkee, India, 1997, pp. 741750. [12] D. E. Goldberg, Genetic Algorithm in Search, Optimization, and Machine Learning, International Student ed. Reading, MA: Addision-Wesely, 1999. Data Structures Evolutionary [13] M. Zbigniew, Genetic Algorithm Programs. New York: Springer-Verlag, 1992. [14] K. P. Wong and Y. W. Wong, Genetic and genetic/simulated annealing approaches to economic dispatch, Proc. Inst. Elect. Eng.Gen., Transm., Dist., vol. 141, no. 5, pp. 507513, 1994. [15] H. T. Yang, P. C. Yang, and C. L. Huang, Evolutionary programming based economic dispatch for units with nonsmooth fuel cost functions, IEEE Trans. Power Syst., vol. 11, pp. 112118, Feb. 1996. [16] P. Venkatesh, P. S. Kannan, and M. Sudhakaran, Application of computational intelligence to economic load dispatch, J. Inst. Eng. (India), vol. 81, pp. 3943, 2000. [17] [Online]. Available: http://cuaerospace.com/carroll/whats_new_ga. html [18] L. L. Lai and J. T. Ma, Application of evolutionary programming to reactive power planningComparison with nonlinear programming approach, IEEE Trans. Power Syst., vol. 12, pp. 198207, Feb. 1997. [19] K. P. Wong and J. Yuryevich, Evolutionary-programming based algorithm for environmentally constrained economic dispatch, IEEE Trans. Power Syst., vol. 13, pp. 301306, May 1998. [20] J. Yuryevich and K. P. Wong, Evolutionary programming based optimal power flow, IEEE Trans. Power Syst., vol. 14, pp. 12451250, Nov. 1999. [21] K. Y. Lee and F. F. Yang, Optimal reactive power planning using evolutionary algorithms: a comparative study for evolutionary programming, evolutionary strategy, genetic algorithm, and linear programming, IEEE Trans. Power Syst., vol. 13, pp. 101108, Feb. 1998.

P. Venkatesh is currently pursuing the Ph.D. degree in the Department of Electrical and Electronics Engineering, Thiagarajar College of Engineering, Madurai, India.

R. Gnanadass is currently pursuing the Ph.D. degree in the Department of Electrical and Electronics Engineering, Pondicherry Engineering College, Pondicherry, India.

Narayana Prasad Padhy received the Degree in Electrical Engineering and the Masters degree in power systems engineering with distinction in 1990 and 1993, respectively. He received the Ph.D. degree in electrical engineering from Anna University, Chennai, India, in 1997. In 1997, he joined Birla Institute of Technology and Science (BITS) as an Assistant Professor, Electrical Engineering Department. Currently, he is Assistant Professor in the Department of Electrical Engineering, Indian Institute of Technology, Roorkee. He taught courses in basic electrical engineering, power systems, and artificial intelligence. His field of interest is power system privatization, restructuring and deregulation, artificial intelligence applications to power system operation and optimization problems, unit commitment, power system wheeling, and FACTS.

You might also like