Certain investigation on the performance of distribution generators using evolutionary algorithms

Abstract

Currently, Distributed Generations (DGs) are used in distribution system to satisfy the increasing demand. According to the demand, the dispatch of the generator should be modified for economical operation. The objective of Economic Dispatch (ED) is to share the demand for power among the online generators keeping the least cost of generation as an objective. It determines the optimal settings of generator units with predicted load demand on a particular time interval. ED of DGs are typically resolved by anyone of the following techniques: conventional Lambda iteration method, Dynamic Programming (DP), etc., or optimization method such as Genetic Algorithm (GA), Evolutionary Programming (EP), Differential Evolution (DE) Algorithm, etc. In this research, various kinds of ED problems have been solved such as i) Online ED of various non-renewable DGs for different demands using Artificial Neural Networks (ANN); ii) ED problem for minimizing the fuel/ emission penalty cost with the wind power penetration of the power system using Backtracking Search Optimization Algorithm (BSA); iii) Application of Self-adaptive Differential Evolution (SaDE) algorithm for optimal position and sizing of renewable DGs in distribution network together with different load models; iv) Application of Multi Objective Differential Evolution (MODE) to find out the minimum fuel cost, environmental pollution penalty cost and minimum system real power loss by solving ED problem. The off-line methods such as Lambda iteration method, DP, etc., or any optimization technique such as GA, EP, etc., for solving ED problem require comparatively significant computation time and are not appropriate for on-line applications. Therefore, it is significant to approximate real power dispatch values within a short period. This research presents an online ED of various non-renewable DGs for different problems using ANN namely, Back Propagation Neural Network (BPNN) and Radial Basis Function Neural Network (RBFNN). In Neural Networks (NN), the electricity demand has been given as input data and the optimal real power dispatch has been obtained as output. The input and output patterns for NN are attained using EP. In this work two diesel engines and two fuel cells are used as DGs. In this study four DGs has been considered. Moreover, this research also has proposed BSA based ED that considers wind power penetration. Simulation studies have been conducted on an IEEE 33-bus system with the objective of minimizing the fuel/ emission penalty cost with the wind power penetration of the power system newline