基于贝叶斯方法的发电机生命周期预测方法

Lifecycle prediction method for generators based on Bayesian method

为了对关键元件进行生命周期预测,提出了一种两阶段数据驱动的生命周期预测方法。在离线阶段,使用无监督变量选择方法寻找包含退化行为信息的变量,在此基础上,从所选变量中构建不同的健康指标,这些变量将退化状态表示为时间的函数,并作为参考模型保存在离线数据库中。在在线阶段,该方法将k近邻(k-NN)分类器作为剩余使用寿命预测器,找到与在线健康指标最相似的离线健康指标,使用离散贝叶斯滤波器来估计退化状态。利用发电厂发电机退化仿真数据对该方法进行了验证,结果表明,该方法能有效地预测发电机的剩余使用寿命,进而得到其生命周期。

In order to predict the lifecycle of key components,a two⁃stage data⁃driven lifecycle prediction method is proposed.In the offline stage,an unsupervised variable selection method is used to search for variables containing degradation behavior information.Based on this,different health indicators are constructed from the selected variables,which represent the degradation state as a function of time and are saved as reference models in the offline data base.In the online stage,this method uses a k⁃Nearest Neighbor(k⁃NN)classifier as a residual lifespan predictor to find the offline health indicator that is most similar to the online health indicator.A discrete Bayesian filter is used to estimate the degradation state.The method was validated using simulation data of generator degradation in power plants,and the results showed that it can effectively predict the remaining service life of the generator,thereby obtaining its lifecycle.