Fuzzy Time Series for Data Analytics and Optimal Electricity Consumption in Residential Area

Abstract

The need for generation and management of electricity persists to grow every year. The evolution in the total number of electric appliances (normally connected objects) per person is due primarily to the fast growth in the world’s population and the increasing trend. In addition, residential sector is the third largest consumer of energy in all economic sectors. Load Forecasting is used mainly for predicting future loads of a particular system for a certain time period. Short-term loads are usually thought to be a variable element affected by various features such as historical load information, datasets of weather elements such as precipitation, wind speed, temperature, air pressure and moisture. A precise forecasting with an individual model is almost difficult. The main problem for utility companies in the world is to forecast energy consumption. Proposed approach presents a deployment of fuzzy time series for the purpose of short-term load forecasting in a residential area. In this regard, our work focuses on hourly, daily and weekly forecasting of electricity consumption for the historical data. The main contribution for our research is to control the issue of overfitting and to enhance the accuracy of our system. There are several characteristics of fuzzy time series that make it more desirable than traditional prediction systems. We shall show the improved prediction system performance by simulations.