Forecasting electricity demand in Ukraine using machine learning methods

Abstract

The article studies the solution of the problem of forecasting electricity demand in Ukraine. The sequence of data processing stages in solving the forecasting problem using machine learning methods is presented. It consists of the following stages: data collection, data research and preparation, construction and training of forecasting models, selection of the best model and calculation of forecasts, evaluation and verification of quality indicators of forecasts. A general methodology for solving forecasting problems is proposed. The methodology for solving the forecasting problem on time series is considered. The forecasting process consists of five stages. The first stage includes the collection, analysis and interpretation of data. The next stage includes the procedures of data research and preparation. The third stage - the modeling stage consists of three parts: preparation of a data set for modeling, selection and training of models and evaluation of their quality. The fourth stage is the forecasting stage and calculation of quality indicators of forecasts. At the fifth stage, procedures for improving the efficiency of the selected forecasting model are performed. The following models were used at the modeling stage: ARIMA, GAM, ANN and BSTS. The analysis of the models was carried out and forecasts were built based on each model. For the constructed models with the best quality indicators, the predictive values were calculated. The forecasts were compared with the data of the validation sample. The following indicators were used to select the optimal model: MAPE, MAE, MSE, RMSE. The BSTS model showed the best results.