Fuzzy logic for renewable energy recommendation and regional consumption forecast using SARIMA and LSTM

Abstract

The use of renewable energy, notably solar and wind energy, has grown exponentially in Brazil. Consumers can generate their energy using renewable sources, whether interconnected to the distribution system (on-grid) or not (off-grid). In this paper, a fuzzy method is developed for the recommendation of solar and wind sources, for any location in the Brazilian territory. In many aspects, it can be viewed as a representation of human decision-making using sets and inference rules and also can be with vagueness and uncertainty, being very useful to idealize recommendation systems. Georeferenced and historical data were obtained from 2003 to 2019 on solar irradiation and wind speed, and electricity consumption until 2021. With the energy generation data from photovoltaic panels and wind turbines, this method allows us to propose installed areas by each technology and obtain the membership of fuzzy recommendation between solar, wind, both solar and wind, unfeasible or hybrid. In addition, a long short-term memory neural network and the seasonal autoregressive integrated moving average model were used to predict consumption for more than 30 months ahead, allowing the recalculation of fuzzy memberships and updating the installation area by respective technologies. As a result, the recommendation is given as the installed area (m2) of each technology per km2 of consumer units, as a function of the regional consumption density (MWh/km2). It can be concluded that it is possible to plan the viability of the type of renewable energy used, according to regional characteristics for smaller consumer units (farms, cooperatives, industries, consortiums), given the diversity of these factors in the huge Brazilian territory. This methodology is in line with the Brazilian Normative Resolution that authorizes the generation of energy by landowners.