HFRBC-GA: A fuzzy classifier for energy systems applications

This paper presents a hierarchical approach for building fuzzy classifiers directly from data following a multilevel grid-like partition of the input domain. The fuzzy classifier is actually the union of several fuzzy systems built on input domain regions increasingly smaller. In order to guarantee high interpretability and to avoid the explosion of the number of rules, only the necessary partitions are built as the hierarchical level increases. Finally, a genetic algorithm is employed to optimize some free parameters of the proposed methodology. The method has been validated on 10 well-known benchmark datasets, by showing how the achieved results compare favorably with those obtained by other fuzzy classifiers in the literature. In addition, we apply our method to three case studies related to energy systems. In the first case study we linguistically describe how the solar irradiation and the temperature of the photovoltaic (PV) panel relate to the quantity of energy produced by a PV installation. The second and third case studies refer to the estimation of energy consumption in buildings. More precisely, we describe how the solar irradiation affects the use of electric lighting, and how the outdoor temperature impacts on hot water boiler usage.