Gaussian Process Model-Based Short-Term Electric Load Forecasting Using Cuckoo Search

Abstract: This paper deals with a Gaussian process model-based short-term electric load forecasting using cuckoo search. The Gaussian process model is a nonparametric model and the output of the model has Gaussian distribution with mean and variance. The multiple Gaussian process models as every hour ahead predictors are used to forecast future electric load demands up to 24 hours ahead in accordance with the direct forecasting approach. The separable least-squares approach that combines the linear least-squares method and cuckoo search is applied to train these Gaussian process models. The results of electric load forecasting for Kyushu district in Japan are shown to demonstrate the effectiveness of the proposed forecasting method.