Forecasting of electricity price and demand using auto-regressive neural networks

Abstract

This paper proposes a forecasting technique of electricity demand and price with volatility based on neural networks. Recent deregulation and liberalization are worldwide currents in the electric industry. The price competition was introduced in a spot market, and the price volatility is concerned because the demand side is non-elastic, and electricity differs from other general commodities. The authors firstly predict an uncertain electric power demand by using the auto-regressive model of the neural networks. The neural network is a popular feed-forward three-layer model, and the input variables of the neural networks include the historical demand, temperature, weather-related discomfort index, and the day of the week. Secondly, by using the demand forecasted and the past prices, we apply the technique for forecasting the electricity price of the next day. The utility of the proposed technique was verified by using real data of the electric power wholesale spot market. Copyright © 2007 International Federation of Automatic Control All Rights Reserved.