Hybrid optimization approach to estimate random demand

Abstract

The main objective of this study is to develop a demand forecasting model that should reflect the characteristics of random demand patterns. To accomplish this goal, a hybrid algorithm combining a genetic algorithm and a local search algorithm method was developed to overcome premature convergence in local optima problems. The performance of the hybrid algorithm was compared with a single algorithm model in estimating parameter values that minimize objective function which was used to measure the goodness-of-fit between the observed data and simulated results. However, two problems had to be overcome in the forecasting random demand model. One was the fitness evaluation in the demand forecasting model in which more than one variable was included, and the other was accuracy of the demand forecasting model to predict the future projection of random energy demand. A local search was proposed to assist in overcoming the first problem. It was used to approximate the input-output response relationship underlying random energy demand forecasting models which was then incorporated into the hybrid algorithm to reduce the local optima problem. To assist in overcoming the second problem, scenario analyses were adopted to determine the future projection of random energy demand.