短期负荷预测中SVM参数选取的混沌优化方法

Chaos Optimization Method of SVM Parameters Selection for Short-term Load Forecasting

支持向量机已成功地应用于短期负荷预测领域,但其学习和泛化能力取决于参数的有效选取。为进一步提高预测精度,针对目前支持向量机参数选取方法的人为盲目性等缺点,在分析各个参数对其预测性能的影响的基础上,将混沌优化技术应用于参数的选取过程。对组合优化问题建立目标函数,采用一种改进的变尺度混沌优化算法来搜索全局最优值,从而得到最优的参数组合。通过湖南某地区电网日负荷预测的仿真结果表明,该方法与常规方法相比,显著地降低了模型的建模误差和预测误差,具有更好的性能。

The support vector machine（SVM） has been successfully applied in the short-term load forecasting area, but its learning and generalization ability depends on a proper setting of its parameters. In order to improve forecasting accuracy, aiming at the disadvantages like man-made blindness in the parameters selection of SVM, chaos optimization method is introduced to select parameters by analyzing the performances of different parameters that also have different influence. The objective function of the combination optimization problem is set and an improved mutative scale chaos optimization algorithm is employed to search global optimal value. Thereby the optimal parameters combination are obtained. Through the analysis of the daily forecasting results for a certain district in Hunan, it is shown that the proposed method could reduce modeling error and forecas- ting error of SVM model effectively and has better performance than general methods.