Genetic-Algorithm-Optimized Artificial Neural Network for Short-Term Load Forecasting: An Indian Scenario

Abstract

Power system planning (PSP) plays a path-decider role for power engineers, during power system operation/expansion or establishing an entirely new plant. Accurate load forecasting (LF) is a mandatory need for good PSP. Short-term LF is one of the vital tools, to support short-term planning issue. There is a rich literature exhibiting artificial-neural-network-(ANN)-based forecasting models. These models have certain limitations like slow and convergence in local optima, and low accuracy, especially when dealing with recent power system load profiles. This paper proposes the implementation of genetic algorithm and ant colony (AC) as weight optimizer for ANN-based LF model. It is hence established from the obtained results that GA-optimized ANN model outperforms, comparing to AC-optimized ANN model as well as Levenberg–Marquardt trained ANN model. While developing these models meteorological parameters, viz temperature, precipitation, and wind speed besides the electric load belonging to Shimla, Himachal Pradesh is considered.