Prediction of electricity production and demand in Russia by time series approaches

Abstract

With the increasing competition between world countries in the energy sector in the last period, the electricity consumption plays a vital role in drawing up an energy development policy for each country. This thesis aims to generate prediction models with high accuracy for forecasting the electricity production and consumption with monthly historical data from Jan 2013 to Jan 2020. Data analysis for forecasting electricity consumption includes creating five forecasting models; the first model by Holt-Winters and the other four models by ARIMA. The best model is ARIMA (1,1,4) where the training set error measures are least (RMSE=17.94, MAE=13.98 and MAPE=1.744). Data analysis for forecasting electricity production includes creating six forecasting models. Two models are by Holt-Winters techniques; the additive and multiplicative models, whereas other models are by seasonal ARIMA. The best model is seasonal ARIMA (1,0,0) (1,1,0) where the training set error measures are less (RMSE=1290.344, MAE=889.811, and MAPE=0.966). Electricity consumption over the next ten years shows a gradual increase as consumption is expected to reach in 2029 about 968 TW/h. The results of the seasonal electricity production increase in Nov, Dec, Jan, and Mar occur about 100000 GW/h.