Aggregated forecast model for prices on electricity market

Abstract

An algorithm for constructing a forecast of the electricity price with the introduction of an aggregated model that includes the accounting of correction components according to the forecast of influencing factors is proposed. The algorithm provides a preliminary determination of the dominant factors depending on the specifics of solving the problem, including a specific region, the depth of the forecast, the established regional conjuncture of the electricity market. Based on the selected factors, a forecast model in the form of time series is constructed. The proposed forecast formation mechanism is implemented by the use of artificial neural networks (ANN). The structure of the ANN allows the convolution of models of influencing factors and the model of the main variable in a single time series of the predicted variable. Such aggregated forecast model makes it possible to significantly increase the accuracy of the forecast in constantly changing behavior conditions of micro- and macroeconomics, climate, production structure and consumption of energy resources, which is confirmed by the example of the Belgorod region.