Analysis of Anfis-Based Approaches for the Prediction of Net Energy Consumption

Abstract

The energy sector is undoubtedly an integral part of any country's economy. Hence, the ability to forecast its future consumption trends would be instrumental in channeling resources to meet up with its demand. Many existing studies have proposed the use of statistical methods or an Artificial Neural network approach to predict and analyze future energy demands. The proposed study sets out to employ an Adaptive Neuro-fuzzy Inference System (ANFIS) and a hybrid ANFIS-PSO approach to determine the future level of energy consumption. The study considers indicators such as population size, Gross Domestic Product (GDP), percentage growth forecast, the expected Final Consumption Expenditure of Households (FCEH) as well as the relevant manufacturing and mining indexes. Three different scenarios were used for these forecasts. In order to illustrate the proposed approach, a dataset containing the required indicators was acquired from the Council for Scientific and Industrial Research (CSIR). This dataset ranges from the year 2014 to 2050 and was used to train the ANFIS-based models to predict electricity demands. It is envisaged that the use of the ANFIS-based approach would yield relatively better results. Therefore, this study will contribute to the formulation of relevant energy planning and management policies.

Keywords: Adaptive Neuro-fuzzy inference system, particle swarm optimization, energy consumption