Forecasting Electric Load Demand Using Hybrid Nonlinear Autoregressive Neural Network with Exogenous inputs and Genetic Algorithm

Abstract

Electricity demand and supply is a very crucial amenity in running the day to day activities of organizations, business enterprises, government, household, social functions, etc. in the twenty first century. Too much supply of electricity causes wastages and too little causes loss. Forecasting electric load demand is vital as it provides an avenue to reduce wastages and loss on the part of any system that depends upon it in her daily activities. In this paper, an attempt is made to forecast the monthly electric load demand of an automobile assembly plant by the application of hybrid Nonlinear Autoregressive Neural Network with Exogenous Inputs and Genetic Algorithm (NARX-GA). The data set consisted of monthly historical data of Electric Load Demand, Temperature, Relative Humidity and Production records for nineteen years from the year 2000 to 2018. The result from the simulation gave a mean average percentage error (MAPE) of 0.56%. The use of Non Linear Autoregressive Neural Network with Exogenous Inputs and Genetic Algorithm (NARX-GA) optimized models should be encouraged in Utility industries to enhance decision making and planning purposes especially in a deregulated economy.