System Imbalance Forecast

Abstract

This thesis deals with the optimization of the BRP imbalance in the opposite direction to the system imbalance by the change of the output of the power plant. BRP imbalance optimization is a part of the energy market together with a long-term trading future or forwards contracts, midd-term day-ahead market and short-term intra-day and balancing market as well. BRP imbalance optimization is special in the Czech Republic by the ability of the BRP to change its imbalance in order to gain profit from achieving the opposite direction to the system imbalance. Therefore, it is needed to forecast the average system imbalance value with the highest prediction for BRP optimization. The objective of the thesis is developing a forecast model, which recommends the optimization of the BRP imbalance in order to gain profit. Lack of the state-of-the-art papers is the reduced usage of the data inputs. There are a lot of factors that influence the system imbalance and often create sudden step changes in the system imbalance. Therefore, forecasted model includes multiple exogenous variables, which can explain and thus forecast these changes. Input exogenous variables are used both in their numeric values and in the differential values. Differential values can be obtained by deduction of the neighbouring values of the input variable or the difference between planned and actual value of the exogenous variable as well. The forecast of the system imbalance is not needed in the point forecast as the concrete value of the system imbalance is not necessary for the optimization of the BRP imbalance. Therefore, I define the intervals of the system imbalance, for which is be the forecast made. Thresholds of these intervals have to be optimized carefully to utilize all the information from the input variables. I calculate the profit and loss resulting from the optimization to evaluate the BRP imbalance optimization. Opportunity costs result from the keeping of the power reserve for the optimization. It has to be kept in mind as these costs can be higher than the profit from the optimization. Results of the forecasted model are compared with the state-of-art and widely spread used ARMA model, which is significantly overcome by our proposed model.