ReNFuzz-LF: A Recurrent Neurofuzzy System for Short-Term Load Forecasting

A neurofuzzy system is proposed for short-term electric load forecasting. The fuzzy rule base of ReNFuzz-LF consists rules with dynamic consequent parts that are small-scale recurrent neural networks one hidden layer, whose neurons have local output feedback. particular representation maintains the learning nature typical static model, since can be considered as subsystems operating at subspaces defined by premise parts, and they interconnected through defuzzification part. Greek power examined, hourly based predictions extracted whole year. forecaster leads to use a minimal set inputs, temporal relations time-series identified without any prior knowledge appropriate past values being necessary. An extensive simulation analysis conducted, forecaster’s performance evaluated using metrics (APE, RMSE, forecast error duration curve). performs efficiently, attaining an average percentage 1.35% yearly absolute 86.3 MW. Finally, compared series Computational Intelligence models, such characteristics highlighted.