基于支持向量机负荷预测应用的探究

理论研究中基于支持向量机的负荷预测的精度已得到了验证,但实际应用中还与其对当地负荷特性的适应性及工作人员对相应软件的应用密切相关,需要进一步验证并解决可能出现的问题。因此,为了将支持向量机预测法应用到新疆电网实际工作中并确保其精度,笔者通过实例将向量机预测法与BP神经网络预测法作了比较,其结果证明了该方法对新疆电网负荷特性具有更好的适应性,同时,重点探讨了在实际应用中出现训练集与预测集存在交集时预测精度与数据重合程度间的非线性关系,并指出预测时需对数据重合程度不同的训练集与预测集的组合进行选择,以确保预测精度。

Research on the application of load forecasting based on support vector machine

The accuracy of load forecasting based on support vector machine having been verified in theoretical stud- y, it relates closely, however, with the adaptability to local load characteristic and the application of relevant soft- ware by staff members in practice. Therefore further verification and solutions to possible problems are necessary. In order to adopt support vector machine in forecasting for Xinjiang Grid and guarantee its accuracy, this paper compares vector machine forecasting and BP neural network forecasting, its result verifies that this method adapts well to load characteristic in Xinjiang Grid, and emphasizes on discussing the nonlinear relation between forecasting accuracy and coincidence degree of data when there is intersection of training set and forecasting set in practice, and points out that the combination of training set and forecasting set with different coincidence degree of data should be selected to guarantee forecasting accuracy.