Spectral mixture kernel for pattern discovery and time series forecasting of electricity peak load

Abstract

In this paper, the author presents the joint of spectral mixture Gaussian and a single squared exponential kernel function which used in predictive solution of Gaussian process (GP) to find new pattern discovery and forecasting of electricity peak load demand of Thailand in next five years. Several analytical results have been evaluated in simulations such as pattern discovery performance, property of each kernel function, and mean absolute percentage error (MAPE) of the method.