短期负荷预测的自适应加权支持向量机新方法

提出了一种基于自适应加权最小二乘支持向量机（AWLS-SVM）理论的电力系统短期负荷预测新方法。在对已知负荷数据及影响因素的分析学习基础上,先用自适应参数优化法整定最小二乘支持向量机的参数,确定最优参数对,然后针对各样本重要性的差异,赋予每个样本惩罚参数不同的加权系数,建立了具有良好推广性能的AWLS-SVM回归模型。本方法突出了不同样本在训练过程中贡献不同的特性,具有结构简单、泛化性能好、不易发生过拟合现象等优点。通过对真实数据的建模预测,证明了该法在短期负荷预测中的可行性和有效性。

An Adaptive Weighed Least Squares Support Vector Machine Approach for Short-term Load Forecasting

A new method based on adaptive weighed least squares support vector machine（AWLS-SVM） for the power system short-term load forecasting is presented.On the basis of analyzing known load data and effect factors,this paper firstly decides parameter of least squares support vector machine employing adaptive parameter optimization method and ascertained optimal parameter pair,then endows each training sample′s penalty with different weighed coefficient according to importance of each training sample,and establishes adaptive weighed LS-SVM model for regression with excellent generalization performance.This method emphasizes characteristic of different training samples having different contributions,and with prominent advantages of simple structure,excellent generalization performance,and over fitting being unlikely to occur.Analysis of the forecasting results of real data proves that this approach is feasible and effective in short-term load forecasting.