Prediction of hourly solar radiation using fuzzy clustering and linguistic modifiers

Abstract

Solar energy is subject to large temporal and spatial fluctuations. These fluctuations require predicting the share of solar energy and its requirements in energy supply systems for optimal use. These predictions form the basis for various solar management options such as storage and control. However, before predicting the production of solar systems, it is necessary to focus on predicting solar radiation. The global prediction of solar radiation is divided into two broad categories of prediction methods: cloud images associated with physical models and automated learning models. In this paper we present a new leaning MAMDANI fuzzy rules based system FRLC (Fuzzy Rule Learning through Clustering) for solar radiation prediction with meteorological data. FRLC based on linguistic modifiers and fuzzy clustering is compared to the most accurate machine learning algorithms such as multilayer feed-forward neural network, radial basis function neural network, support vector regression, and adaptive neuro-fuzzy inference system. FRLC outperforms all algorithms at interpretability level by offering a linguistic knowledge base to the experts of the domain.