The study of methods of grouping electricity users for the precise prediction of demands

On the electricity markets, in many cases, users decide how much power they need before the day when they actually use. Therefore, precise prediction of demand load is important. In this study, the method of demand prediction is proposed by multivariate analysis using weather data, and date is also considered. However, because of the high volatility of demand loads, it is not efficient to predict demand of one user. Therefore, we propose three methods of grouping electricity users, grouping randomly, grouping by Euclid Distance and grouping by Discrete Fourier Transform. In conclusion, demand prediction about the group made by last method is done 9.2% lower by MAPE than that about the group made randomly.