Methods of day ahead load forecasting on the example of a residential area

Abstract

Abstract The work is aimed at identifying the most effective methodology for short-term forecasting of energy consumption concerning intervals of 1 hour to 1 week based on the employment of im-proved fuzzy recurrence and long short-term memory neural networks. The object of research is a residential area with an uneven consumption of electrical energy. The article discusses methods for forecasting electrical load on the example of a residential area. The existing methods of forecasting were analyzed and the day-ahead and intra-day power forecast were reviewed. The list of relevant sources was presented. On the basis of the reviewed literature in this area, the most effective and modern methods were identified that allow determining the consumption of electrical loads in residential and industrial buildings with an accuracy of 98 percent. Tables were developed reflecting the effectiveness of the considered techniques. The time horizons of the forecast of electric energy consumption are considered. Four categories of load forecasting were identified: long-term forecasting with a forecast interval of more than one year; medium-term forecasting with a forecast interval from one month to one year; short-term load forecasting with a forecast interval from 1 day to several weeks.; operational forecasting, with a forecast interval from 1-2 hours to the end of the day. A comparative analysis of methods for predicting electrical load for intelligent network applications is carried out and its results are presented.