Evolutionary algorithms for solving power system optimization problems

The scope of optimization in power system is ample. In general, optimization helps efficient and economical operation of the electrical system. It is worthwhile to note that power system problems are mostly non-linear and non-convex, and they often require optimization of two or more conflicting objectives. The problem can also be a mix of discrete and continuous variables that are needed to be handled by the optimization algorithms. During earlier days, classical numerical optimization methods were in use. The classical methods are known to perform well for convex and continuous optimization problems as these methods usually employ a gradient based search that has the tendency to converge to local optima. Revolution in numerical optimization introduced several evolutionary algorithms (EAs) and techniques in the last three decades. Most of these methods can successfully overcome the problem of premature convergence and explore the search space in pursuit of the global optimum. The field of evolutionary algorithms being dynamic, the classical power system optimization problems can be revisited to check whether performance of the network can be enhanced with the application of new algorithms. Furthermore, renewable energy sources are becoming integral parts of the modern smart grid, a fact that opens up avenues for formulation of new optimization problems incorporating many more components. The primary scope of this research is in applying the state-of-the-art variants of differential evolution (DE) algorithm for single-objective optimization and selected evolutionary algorithms for multi-objective optimization problems in power system. This thesis includes a detailed survey of the algorithm success history based adaptive differential evolution (SHADE) and its variant the linear population size reduction technique of SHADE (L-SHADE) for single-objective optimization; decomposition based multi-objective evolutionary algorithm (MOEA/D) and summation based multi-objective differential evolution (SMODE) for multi-objective optimization. Furthermore, operational and security constraints are also common in the electrical network. Static penalty function method has been the easiest and simplest way to deal with the constraints in power system problems. In this thesis, we present a review of three constraint handling (CH) techniques for EAs - superiority of feasible solutions (SF), epsilon constraint (EC) handling and stochastic ranking (SR) methods. The constraint handling techniques have been successfully implemented in conjunction with the EAs and applied to the existing and newly proposed formulations of constrained optimization problems in power system such as optimal power flow (OPF), economic-environmental dispatch (EED) and optimal reactive power dispatch (ORPD). In addition, studies on optimal siting of wind turbines in a windfarm and loss minimization in a distribution network are also presented in relevant chapters. The final chapter of the thesis concludes with potential future works on optimization in smart grid.Doctor of Philosoph