Modelling and forecasting short-term load demand: A multivariate approach

A multivariable time series model is proposed for short-term load demand forecasting. Unlike other approaches, the order of the model is determined without first finding the coefficients of the model. The Hankel matrix used for determining the order is also utilized for estimating the parameters of the model. This is then compared with order determination using the AIC criterion. Actual data provided by the Ontario Hydro for four loading buses is used for five-minute and hourly forecasts. The results show that the proposed approach is very attractive.