Automatic fuzzy model identification for short-term load forecast

The conventional fuzzy modelling of short-term load forecasting has a drawback in that the fuzzy rules or the fuzzy membership functions are determined by trial and error. An automatic model identification procedure is proposed to construct the fuzzy model for short-term load forecast. An analysis of variance is used to identify the influential variables of the system load. To set up the fuzzy rules, a cluster estimation method is adopted to determine the number of rules and the membership functions of variables involved in the premises of the rules. A recursive least squares method is then used to determine the coefficients in the concluding parts of the rules. None of these steps involves nonlinear optimisation and all steps have well bounded computation time. This method was tested on the Taiwan Power Company's (Taipower) load data and the performance of the proposed method is compared to those of Box-Jenkins (B J) transfer function and artificial neural network (ANN) models.