基于CS-SVR模型的短期风电功率预测

Short-term Wind Power Prediction Based On CS-SVR ModelS

摘要:

为了提高短期风电功率预测精度,提出一种布谷鸟搜索算法(Cuckoo Search Algorithm, CS)优化支持向量回归(Support Vector Regression, SVR)机的预测方法,该方法首先根据上截断点和下截断点对输入数据进行预处理,剔除异常数据,之后以输入数据中的风速、平均风速、风机状态等属性数据作为SVR算法模型的输入,以风电功率数据作为SVR算法模型的输出,建立短期风电功率的SVR预测模型,针对SVR算法存在难以选择最优参数的缺点,提出采用布谷鸟算法优化SVR参数的方法,建立短期风电功率的CS-SVR预测模型。通过与SVR、PSO-SVR预测模型进行了对比仿真实验,实验结果表明,CS-SVR预测模型具有较高的预测精度。

Abstract:

In order to improve the prediction accuracy of short-term wind power, a prediction method of cuckoo search algorithm (Cuckoo Search Algorithm, CS) to optimize support vector regression (Support Vector Regression, SVR) machine is proposed. The method preprocesses the input data according to the upper cutoff point and the lower cutoff point to eliminate the abnormal data. Then, the wind speed, average wind speed, fan state and other attribute data in the input data were used as the input of the SVR algorithm model, and the wind power data was used as the output of the SVR algorithm model to establish the SVR prediction model of short-term wind power. Considering that the SVR algorithm is difficult to select the optimal parameters, a CS-SVR prediction model for short-term wind power is established by using the cuckoo algorithm optimize the SVR parameters. Compared with SVR and PSO-SVR prediction models, the simulation experiment was conducted. Experimental results show that the CS-SVR prediction model has higher prediction accuracy.