A Set-Membership approach to short-term electric load forecasting

Abstract

This work presents a model for the short-term forecast of electric load, based on Set-Membership techniques. The model is formed by a periodic component and an adaptive non-linear autoregressive component. The identifications set of the non-linear model is increased at each estimation step. The model is evaluated in a case study with more than 13,000 samples of hourly sampled energy demand, registered during three years at a rural town in Colombia. The performance of the estimator is evaluated and confronted to a linear autoregressive model and a standard Set-Membership model with fixed identification set. Results shows that the proposed estimator is able to predict demand with an RMS error below 2.5 % for validation data, using just a 5 % of the available dataset for the model identification.