Long and medium term power load forecasting based on a combination model of GMDH, PSO and LSSVM

Abstract

In this paper, a new method based on Group Method of Data Handling (GMDH), Improved Particle Swarm Optimization (PSO), and Least Squares Support Vector Machine (LSSVM) is proposed to solve the problem in power load forecasting, which is difficult to determine the input node and model parameters of minimum support vector machine (LSSVM) modeling. The specific method is as follows: firstly, the authors use the GMDH algorithm to obtain the input variables of the LSSVM modeling. Secondly, the adaptive mutation PSO algorithm is analyzed to optimize the parameters of the LSSVM model, and then the trained LSSVM model is utilized to predict the test samples. Furthermore, a real case about the actual load of a certain city from the year 2008 to 2013 is analyzed, and the power load in 2014, 2015 were forecast. The simulation results verify that the forecasting accuracy has been improved obviously.