Forecasting error processing techniques and frequency domain decomposition for forecasting error compensation and renewable energy firming in hybrid systems

Abstract

Battery storage can provide a wide range of services in power systems. This paper focuses on battery storage co-locating with a hybrid wind and solar system to achieve forecasting error compensation and renewable energy firming. Multiple forecasting error processing techniques are presented, including the exclusion of seasonal frequency components, along with scaling and shifting methods to remove the effect from battery round-trip efficiency. The impacts of these error processing techniques on reducing the size of the battery system required have been extensively investigated. To include a variety of forecasting errors, both day-ahead and hour-ahead forecasting were performed utilising four different forecasting methods, including persistence, Elman neural network, wavelet neural network and autoregressive integrated moving average (ARIMA). Numerical simulations demonstrate that the exclusion of seasonal components combined with the scaling method can substantially reduce the size of battery systems required for forecasting error compensation, whilst the shifting method makes a significant contribution to reducing the required battery size for renewable energy firming.