A new fuzzy-based feature selection and hybrid TLA–ANN modelling for short-term load forecasting

Abstract

In this paper, a new hybrid method based on teacher learning algorithm (TLA) and artificial neural network (ANN) is proposed to develop an accurate model to investigate short-term load forecasting more precisely. In contrast to the other evolutionary-based training techniques, the proposed method utilises both the ability of ANNs to generate a non-linear mapping among different complex data as well as the powerful ability of TLA for global search and exploration. In addition, in an attempt to choose the most satisfying features from the set of input variables, a novel feature-selection approach based on fuzzy clustering and fuzzy set theory is proposed and utilised sufficiently. In order to improve the overall performance of TLA for optimisation applications, a new modification phase is proposed to increase the ability of the algorithm to explore the entire search space globally. The simulation results show the feasibility and the superiority of the proposed hybrid method over the other well-known methods in the area.

Keywords::

• fuzzy-based feature selection
• artificial neural network
• modified teacher-learning algorithm
• short-term load forecasting