Long term peak load forecasting in Thailand using multiple kernel Gaussian Process

This paper presents the forecast of the peak electricity demand (peak load) between 2014 to 2024 of the “Electricity Generating Authority of Thailand” (EGAT) by using Gaussian Process (QV), which used training data set since 2000 to 2013. The training data set composed of two important factors, including time on a monthly and the monthly of electricity peak load Moreover, it proposes a solution to model multiple kernel function by consider training data, how to compute the hyper-parameters (Θ) that is the important factor to optimize peak electricity demand Simulation results show the proposed forecasting method that gives a “Mean Absolute Percentage Error” (APE) 2.102 % in validation period when compare with the peak electricity demand from Jan.2013 to Sep.2013, proposed the trend of peak electricity demand until 2024.