PCA-PSO/GS-SVM组合方法在风电齿轮箱故障预测中的应用研究

APPLICATION OF PCA-PSO/GS-SVM COMBINATION METHOD IN FAULT PREDICTION OF WIND TURBINE GEARBOX

在标准支持向量机(SVM)的基础上,引入主成分分析法(PCA)、粒子群算法(PSO)以及网格算法(GS),构建针对风力机故障的PCA-PSO/GS-SVM组合预测模型。相对于标准SVM,该模型采用PSO以及GS算法寻优参数,能够更准确地建立各变量间的相关关系以提高模型的预测准确性。以中国北方某风场2 MW风电齿轮箱在2017年上半年某2个月的SCADA监测数据为例进行分析。结果表明,对于以齿轮箱输出功率为例的骤变信号的预测,采用PSO算法寻优后的绝对误差均值是采用GS算法的3.0647倍,而对于以高速侧轴端温度为例的缓变信号的预测,则采用PSO算法更加合理;同时发现剔除训练样本数据中的奇异点能够有效提高模型的预测精度及其泛化能力。

On account of the SCADA data has been deficiently studied via the existing fault prediction,this paper proposed a PCA-PSO/GS-SVM integrated model of fault prediction based on the standard support vector machine(SVM)method,principal components analysis(PCA),particle swarm optimization(PSO)and grid search(GS). Compared with the standard SVM,the proposed method establishes the correlation between variables accurately with the PSO and GS methods. Meanwhile,SCADA monitoring data of the 2 MW wind turbine gearboxes in north China is as an example for analysis. The results show that the mean prediction absolute error of the gearbox output power after using PSO algorithm is 3.0647 times that of the GS algorithm. It is more reasonable to use PSO algorithm to optimize the parameters for the prediction of the temperature. Moreover,eliminating singular points in training sample data can improve the prediction accuracy and generalization ability of the model.