基于最小二乘支持向量机的风电功率超短期预测

Ultra-short-term Forecast for Wind Power Based on Least Squares-Support Vector Machine

针对当前风电场发电功率预测时间较长、预测误差较大,易影响风力微电网根据用电负荷变化适时调度及有效电力资源配置的问题,提出了一种基于最小二乘支持向量机(least squares-support vector machine,LS-SVM)的微电网风电功率超短期预测方法。该方法根据风电场数据采集与监视控制(supervisory control and data acquisition,SCADA)系统获取原始功率数据样本,经归一化法预处理,运用网格搜索法确定模型参数,并依据LS-SVM法建立预测系统模型,利用MATLAB工具箱LS-SVMLab进行仿真实验,跟踪及预测风电功率变化曲线,实现时间跨度小至5min的超短期预测。实验验证结果表明,该方法比传统预测方法具有较高的精确度和较大的适用性,为风力微电网优化调度控制工程提供一种新思路。

The current power forecasting of wind farm is too long and the error is large, which causes it clifficult to timely dispatching of wind microgrid and efficienfly distributing of electrical recourses in accordance with load variation. A new method of ultra-short-term forecasting for wind power was introduced based on least squares support vector machine （LS-SVM）. By unitary preprocessing sample data of wind farm with supervisory control and data acquisition system and adopting grid search method to determine model parameters, the forecasting system model was built according to the LS-SVM method. The simulation of forecasting experiment was realized to follow the curve of wind power and ultra-short-term forecast in 5 minutes later by MATLAB toolbox LS-SVM Lab. The results show that LS-SVM method has a higher accuracy, stronger robust than traditional manner, which provide a new train of thought for optimal dispatch of wind microgrid.