Gaussian process prior models for electrical load forecasting

This paper examines models based on Gaussian process (GP) priors for electrical load forecasting. This methodology is seen to encompass a number of popular forecasting methods, such as basic structural models (BSMs) and seasonal auto-regressive intergrated (SARI) as special cases. The GP forecasting models are shown to have some desirable properties and their performance is examined on weekly and yearly Irish load data.