Electricity price forecasting by clustering-least squares support vector machine

Abstract

In the electricity market, the price as the lever results in the dramatic variations, especially, the capacity or willingness of electricity consumers and then demand may be low, particularly over short time frames. Therefore demand-side management (DSM) has been put into practice, and the market supervisors become more and more focused on the price dynamics of the short-term, because of its effects on the modification of consumer demand for energy through various methods especially financial incentives. But due to the complexity of the price, the electricity price forecasting is along one of focused and unsolved problems in the researches of electricity market. This paper describes a novel model for price forecasting is proposed by the developed least squares support vector machine (LS-SVM), which integrates Clustering algorithm with LS-SVM. First, clustering of the data samples are performed, which aims at mining the latent patterns in the data. After that, LS-SVM is applied for the nonlinear regression modeling of electricity price and its influence factors signed with its class, which results in a more efficient training and forecasting. Finally, hourly prices and loads of different market are employed to test the proposed approach.