基于组合预测模型的风电机组关键部位故障检测

FAULT DETECTION OF KEY COMPONENTS OF WIND TURBINE BASED ON COMBINATION PREDICTION MODEL

提出一种基于组合模型的风电机组关键部件状态建模和故障识别方法。首先对机组采集与监控数据进行参数辨识,提取与故障检测相关的参数;然后利用残差最优化问题建立非线性状态估计和神经网络组合的预测模型,将前轴承温度作为参数分别输入组合模型和单一模型中,通过评估指标反映模型的精确度;最后采用风电场SCADA实际运行数据对风电机组发电机和齿轮箱的温度进行状态监测,分别建立组合预测模型并根据预测残差是否超过设定阈值判断故障状态,对比故障前后记录的数据并进行分析,实验结果证明了该文所建立的预测模型对机组部件故障检测的可行性。

This paper proposes a method for state modeling and fault identification of key components of wind turbines based on a combined model.Firstly,the SCADA data is identified,and the parameters related to the fault detection are selected.Then,the residual optimization problem is used to establish a nonlinear state estimation and neural network combination prediction model.The front bearing temperature is input into the combined model and the single model as parameters,and the accuracy of the model is reflected by the evaluation index.Finally,the SCADA data of the wind farm is used to analyze the wind turbine generator and gearbox temperature,and the combination prediction model is established to detect the fault.According to the residual temperature of the corresponding part will exceed the set threshold in the fault state,the data recorded before and after the fault are compared.The simulation results show that the proposed method is effective for fault detection of key components by establishing combined model.