Comparison of emerging metaheuristic algorithms for optimal hydrothermal system operation

MARTHA P. CAMARGO, JOSÉ L. RUEDA, ISTVÁN ERLICH, OSVALDO AÑÓ

Swarm and Evolutionary Computation

ELSEVIER

Año: 2014 vol. 18 p. 83 - 96

2210-6502

Optimal hydrothermal system operation (OHSO) is one of the complex and hard-to-solve problems in power system field due to its nonlinear, dynamic, stochastic, non-separable and non-convex nature. Traditionally, this problem has been solved through classical optimization algorithms, which require some approximations to tackle a more tractable variant of the original problem formulation. Metaheuristic optimization has undergone a significant development in recent years, thus, there is a variety of tools with different conceptual differences, which offer a great potential for solving OHSO without extensive simplifications. This paper provides a comparative study on the application of six emerging metaheuristic algorithms to OHSO, namely, the Comprehensive Learning Particle Swarm Optimizer (CLPSO), Genetic algorithm with Multi-Parent Crossover (GA-MPC), Differential Evolution with Adaptive Crossover Operator (DE-ACO), Covariance Matrix Adaptation Evolution Strategy (CMA-ES), Linearized Biogeography-based Optimization (LBBO), and the Hybrid Median-Variance Mapping Optimization (MVMO-SH). Since these tools have been successfully applied to other hard-to-solve optimization problems, the goal is to ascertain their effectiveness when adapted to tackle the OHSO problem by evaluating their performance in terms of convergence speed, achieved optimum solutions, and computing effort. Numerical experiments, performed on a test system composed by four cascaded hydro plants and an equivalent thermal plant, highlight the relevance of the adopted global search mechanisms, especially for LBBO and MVMO-SH. A nonlinear programming (NLP) algorithm is used as reference to validate the results.