The statistical price forecasting models in the wholesale electricity markets: Russian and foreign experience

Importance. The article analyzes the problems of price forecasting in the wholesale electricity market. The task of future electricity price forecasting is the basis for activity planning of subjects in the electrical power industry (preparation of electricity sale or purchase proposals, conclusion of bilateral contracts, planning of electricity consumption and power plant operations). Objectives. The detailed analysis of statistical models for forecasting electricity prices taking into account Russian and foreign experience. Methods. Methodological foundation includes study and systematization of scientific papers of Russian and foreign authors. Results. The main stages of developing a statistical model of electricity price forecasting have been specified: analysis of time series characteristics, primary series adjustment and consideration of seasonality, selection of a statistical forecasting model, including time parameters (the depth of the data used, the forecast horizon) and exogenous variables, the model accuracy appraisal. The systematization of approaches to consideration of periodicities and statistical price forecasting models, as well as the analysis of included exogenous variables have been implemented. Conclusions and Relevance. The forecasting model is developed taking into account the price series characteristics of a specific wholesale market. The combination of autoregressive integrated moving average model and generalized autoregressive conditional heteroscadasticity model allows to capture many important features of the series. This approach is widely and successfully used. In a number of papers, it was noted that wavelet transform in combination with this approach improves the accuracy of the forecast. However, unconditional evidence of superiority of any one model over the other in terms of forecasting accuracy is not found out. Demand volume, less often weather variables and fuel price are used as exogenous variables in most of the papers. A few papers address the problem of market technological characteristics, of market power from producers, of special aspects of the market rules. These problems deserve further attention in the forecasting modelling.