Modelling the Level of Sunlight by Using Fuzzy Sets

The problem of proper management of electricity consumption in the era of high energy prices is a very important issue. In the case of large enterprises that consume significant a mounts of energy, there is an additional problem of correctly determining the tariff to be contracted. Development of an efficient energy consumption management system based on the use of additional energy sources from photovoltaic panels requires taking into account an effective system that allows to predict the level of insolation in the following hours. The paper presents an innovative model for forecasting the level of insolation based on historical data. Due to the constantly changing climatic conditions, a two-level of fuzzyfication was applied. At the first level, the level of membership to various descriptors describing the length of a given day was determined. On the second level of fuzzyfication, based on historical data, the degree of membership to classes representing different levels of sunshine at a given hour was determined. As part of numerical experiments, selected machine learning tools were used to predict the degree of membership to particular classes based on appropriately selected historical data. The obtained results confirm the very high prediction accuracy of the innovative solution that allows to determine the degree of insolation.