Evolutionary algorithms based solution to reactive power planning and dispatch problems with security constraints

Abstract

Nowadays power system engineers show much attention on Reactive Power Optimization (RPO) problem in order to improve the economy and security of power system operation. The objectives of RPO are to improve the voltage profile, to minimize the system active power losses and to determine optimal VAr compensation placement under various operating conditions. These objectives are achieved by adjusting the transformer tap-ratios, generator voltages and reactive power output of VAr sources like shunt capacitors etc. To overcome the drawback of classical techniques the flexible Evolutionary Algorithms (EAs) such as Genetic Algorithms (GA), Evolutionary Strategies (ES), Evolutionary Programming (EP), Particle Swarm Optimization (PSO), Differential Evolution (DE) and Real coded Genetic Algorithm (RGA) have been applied for RPO problems. Recently RPO problem is formulated as Multiobjective Optimization (MOO) problem. In this thesis, the suitability of the EAs, namely Modified PSO (MPSO), RGA with Simulated Binary Crossover (RGA-SBX), Covariance Matrix Adapted Evolution Strategy (CMAES) and Self-adaptive Differential Evolution (SaDE) on three single objective optimization problem, namely RPD with Flexible AC Transmission Systems devices such as Thyristor Controlled Series Capacitor and Thyristor Controlled Phase Shifting Transformer, RPP with voltage stability enhancement and RPP in hybrid electricity markets are investigated. In this thesis, Modified Nondominated Sorting Genetic Algorithm-II (MNSGA-II) is applied on multiobjective RPD and RPP problems. MNSGAII incorporates the controlled elitism and the Dynamic Crowding Distance (DCD) concepts with the existing Nondominated Sorting Genetic Algorithm (NSGA-II) to improve the lateral and longitudinal diversity of the Paretofront. Further the results obtained by CMAES algorithm for single objective optimization problem and MNSGA-II algorithm for MOO problem are verified for optimality by using Karush-Kuhn-Tucker (KKT) conditions. The resulting KKT error value confirms the opt