Modeling the electrical load profile of residential consumers using a Takagi-Sugeno fuzzy inference system

Abstract

This paper examines the problem of forecasting electricity load profiles for residential consumers, which is critical for the effective implementation of demand response in the context of a functioning electricity market and high uncertainty in consumption. An approach to modeling is proposed based on a Takagi-Sugeno-type fuzzy inference system, which relies on processing time series of hourly electricity consumption using the particle swarm method for parametric identification of membership functions. During computational experiments using real commercial metering data, the influence of the time lag and the number of membership functions on the model’s accuracy, assessed by the root mean square error and the coefficient of determination, as well as on computational costs was investigated. It was found that increasing the time lag significantly improves the quality of forecasting, whereas an excessive increase in the number of membership functions leads to model overfitting and an unproductive increase in computation time. An optimal configuration of a fuzzy model with a time lag of three and four membership functions has been determined, which provides the best balance between forecasting accuracy and computational efficiency, confirming the feasibility of its use for practical tasks of managing electricity consumption in residential areas.