Short-Term Forecasting of Imbalances in the IPS of Ukraine

This chapter discusses the characteristics of short-term forecasting of electrical energy imbalances in the Integrated Power System (IPS) of Ukraine. Several forecasting models have been proposed, and the impact of renewable energy generation forecasts on imbalances forecast accuracy has been examined. Furthermore, the text examines aspects of forecasting generation from solar power plants and compares the accuracy of different time series forecasting models such as ARMA, ARIMA, AR series, and SARIMA for short-term electricity imbalance forecasting. The SARIMA model was found to be the most suitable for the given features. The chapter also proposes an architecture and mathematical model of an artificial neural network for deep training to forecast short-term electricity imbalances using hourly data. The LSTNet model achieved the smallest forecast errors. Additionally, the use of information about the hour of the day and previous renewable energy generation values improved the accuracy of short-term electricity imbalance forecasting. For photovoltaic power plant generation, the use of forecast values of meteorological factors, particularly total/direct solar radiation, was found to significantly reduce the forecasting error, especially for longer forecast horizons. Aggregating data to an hourly resolution with subsequent forecasting was also found to give a lower forecast error than aggregating forecasts with a 15-min resolution. The chapter concludes that the autoregressive model and artificial neural network model studied are promising for short-term electricity imbalance forecasting and should be improved and tested on data from other seasons.