风力发电机温升故障预警方法研究

Study on the Method of Wind Turbine Temperature Fault Warning

针对传统PCA在解决变量重要性差异很大问题时的局限性,提出一种结合relief F算法和加权主元算法的方法。该方法首先采集张家口某风场5月某段时间内风力发电机组运行数据并建立数据集,通过relief F算法确定权值向量W,进而使用加权主成分分析法建立正常工作状态下发电机温度模型。利用该模型对同一机组另一段时间运行数据进行故障检测,通过对统计量HotellingT^2(简称T^2)和平方预测误差(简称SPE)的趋势分析最终确定机组运行状态。实验结果表明,提出的方法准确预测出了发电机的温升故障,并减少了误报情况。

Aimed at the limitations of the traditional PCA in solving the problem of variable big difference importance,this paper proposes a method comprising of relief F Algorithm and Weighted Principal Component Analysis（ WPCA）. This method first collects the data of wind turbine operation in a certain wind farm in Zhangjiakou in May and sets up the data set,through the relief F algorithm to determine the weight vector W,and uses the weighted principal component analysis method to establish the normal working state of the generator temperature model.The model is used to detect the faults of the same set of another period of time,and trend analysis of the model statistics HotellingT^2（ T^2） and squared prediction error（ SPE） to finalize the running status of the unit. The experimental results show that the proposed method accurately predicts the failure of the generator temperature rise,and reduces the false alarm situation.