Evolutionary algorithms based decentralized congestion management for multilateral transactions

Abstract

The restructuring of the electric sector has been stimulated by the economic benefits to society resulting from the deregulation of other industries such as telecommunication and airlines. The congestion management problem is generally described as a Centralized Optimal Power Flow (COPF) problem with the objective of maximizing social welfare and the constraints of load flow equations and operation limitations. A single objective decentralized model for congestion management in the forward markets using a resource allocation technique is proposed. In this model, each transaction maximizes its profit under the limits of transmission line capacities allocated by the Independent System Operator (ISO). The proposed approach is applied to smooth and non-smooth cost functions. The results obtained for the decentralized model using the CMAES algorithm, are compared with those of the Particle Swarm Optimization (PSO) algorithm and Sequential Quadratic Programming (SQP) technique. It considers the conflicting objectives of the maximization of social welfare and the minimization of emission impacts. An elitist evolutionary multiobjective optimization algorithm called the Modified Non-dominated Sorting Genetic Algorithm II (MNSGA-II) with controlled elitism and the dynamic crowding distance method is applied. The effectiveness of the proposed approach is demonstrated by comparing the obtained Pareto front with the reference Pareto front generated by multiple runs of the Covariance Matrix Adapted Evolution Strategy (CMAES) algorithm with respect to minimum spacing, diversity and convergence metric performance measures. Once the solutions lying on the Pareto front are determined, the best among the solutions is obtained by posterior evaluation of the Pareto front using the TOPSIS concept. In order to validate the consistency of MNSGA-II, statistical performance measures are obtained by comparing the reference Pareto front with the Pareto fronts obtained from 20 independent runs. newline