A Method of Predicting Energy Demands in Consideration of Minimizing Variations in Their Predicted Values

Abstract

It is necessary to predict energy demands appropriately for rational operation of energy supply systems. The conventional method of identifying parameters of a prediction model minimizes only the errors between measured and predicted energy demands. Since the variations in predicted ones are not taken into consideration, the optimal operation of energy supply systems based on predicted ones may repeat startup and shutdown of equipment. In this paper, a method of identifying parameters of a multiplicative ARIMA prediction model is proposed, which minimizes the errors between measured and predicted energy demands as well as the variations in predicted ones by a global optimization approach for nonlinear programming. Through a case study, it turns out that the proposed method can reduce the variations in predicted energy demands.