基于杂草算法的支持向量机电力负荷预测

Support Vector Machine based on Invasive Weed Optimization to Power Load Forecasting

分析各参数对支持向量机学习能力的影响,提出一种基于杂草算法的支持向量机电力负荷预测方法。首先运用杂草算法对支持向量机的两个关键参数进行智能寻优,然后将最优参数运用到支持向量机预测模型。采用新的预测模型对EUNITE第一次竞赛提供的相关电力数据进行分析,并与基于回归树和基于神经网络的预测方法进行比较。结果表明本文方法智能化地解决了传统参数选择方法的缺陷,且对电力负荷预测具有较高的预测精度。

The impact of kinds of parameters of support vector machine（SVM）on learning ability was studied,and support vector machine based on invasive weed optimization（IWO）to power load forecasting was described in this paper.The two crucial parameters were ifrst optimized intel igently by invasive weed optimization,and then the optimal parameters were applied in support vector machine forecasting model.Relevant power load to the ifrst time of EUNITE competition was taken as the experiment sample to be analyzed,and then the improved forecasting model was compared with the forecasting methods based on regress tree and neural network.The results show that the proposed scheme solved intel igently the defects of the traditional parameter selection methods,and improved the forecasting accuracy.