基于支持向量机的电力市场价格预测中的核函数比较

A Comparison of Kernel Functions of Support Vector Machine for Electricity Price Forecasting

电价预测是电力市场中的一个重要研究课题。支持向量机(SVM)已被广泛应用于这一领域。然而,电力市场电价的高波动性和随机性等特征给支持向量机核函数的选择带来了挑战。本文在选择不同核函数的基础上,分别建立两个电力价格预测模型,并用真实电力市场价格数据对两个模型进行验证。实验结果表明,与其他支持向量机预测研究相比,本文精心选择的SVM核函数对短期电价预测具有较好性能。

Accurate forecasting of spot price is an essential issue in electricity market. Support Vector machines （SVM） has been widely adopted to deal with this issue. However, the high fluctuation and randomness features of electricity market present a number of challenges for the choosing of kernel functions for SVM. In this paper, by using different kernel functions, two SVM models for electricity price forecast have been developed. Case studies, adopting data from an actual electricity market, have been performed and the results are presented. In addition, comparisons with results from other SVM forecasting studies have shown that the performance of SVM models could be improved by choosing a tailored kernel function.