考虑样本优化的风电机组齿轮箱轴承故障预测

Wind Turbine Gearbox Bearing Fault Prediction with Sample Optimization

由于风电机组SCADA(supervisory control and data acquisition)系统包含较多冗余信息并且数据之间具有较大耦合性,因此在数据挖掘过程中进行样本优化意义重大。文章采用非线性状态估计(NEST)方法建立齿轮箱轴承温度模型并用其进行轴承故障预测,首先采用灰色关联度分析法为选择观测向量提供理论依据,再采用相似度分析法构造简约过程记忆矩阵,使其在不冗余的情况下尽可能覆盖齿轮箱全部正常工作状态以实现样本优化,进而当齿轮箱发生故障时,通过简约矩阵训练的模型残差将较早超出阈值并提前进行预警。最后结合某风电机组实际运行数据进行仿真分析,验证了模型的时效性与优越性。

Because the wind turbine SCADA system contains too much information which is redundant and has large coupling,it＇ s important to carry out the sample optimization in the process of data mining. This paper uses the non- linear state estimation （NEST） to establish temperature model of gearbox bearing by which the bearing fault can be predicted. Firstly,it uses the gray correlation analysis method to provide the theoretical basis, and which caters for se- lecting observation vector. Secondly, it uses the similarity analysis to construct the simple process memory matrix and achieve sample optimization, which can make the data cover the whole normal working condition of the gearbox as far as possible without redundancy. Thirdly,When the gearbox fault occurs,the improved model＇ s residual will exceed the threshold and alert. Besides, the warnings can be launched in advance. Finally, it deals with the wind turbine op- eration data by simulation analysis to verify the model＇ s timeliness and superiority.