Ensemble Forecasting Applied to Power Systems

Abstract

Modern power systems are affected by many sources of uncertainty, driven by the spread of renewable generation, by the development of liberalized energy market systems and by the intrinsic random behavior of the final energy customers. Forecasting is, therefore, a crucial task in planning and managing modern power systems at any level, from transmission to distribution networks, and in the new context of smart grids. Many important operations are scheduled and performed on the basis of predictions of several energy-related variables, such as non-controllable generation, loads, energy prices, and power quality indicators. Forecasts of these variables at different lead times, ranging from several minutes to several days, are needed by the operators in order to suit different applications and scenarios. The application of forecasting techniques to power systems, in both deterministic and probabilistic frameworks, is a topic that is still far from being fully explored. Recent trends suggest the suitability of ensemble approaches in order to increase the versatility and robustness of forecasting systems. Stacking, boosting, and bagging techniques have successfully been applied within several frameworks, and have recently started to attract the interest of power system practitioners. The subject is therefore worthy of further investigation. This book addresses the development of new, advanced, ensemble forecasting methods applied to power systems, collecting recent contributions to the development of accurate forecasts of energyrelated variables by some of the most qualified experts in energy forecasting. Typical areas of research (renewable energy forecasting, load forecasting, energy price forecasting) are investigated, with relevant applications to the use of forecasts in energy management systems. This book is relevant for forecasters and energy practitioners involved in power system planning and management. It also represents a good starting point for young researchers who want to discover potentialities of ensemble forecasting systems. Eventually, the contents of the book may be used by system operators and market operators in order to boost the performance of their inhouse forecasting systems through ensemble refinement.