基于自学习概率模型的风电机组故障诊断方法

Fault Diagnosis of Wind Turbine Generator Based on Self-learning Probability Model

为了使故障诊断系统能够提供日常有效的维护建议和有针对性的维修计划,降低风电场的维护成本,提升其产量,在对风电机组的监控数据进行分析、分类的基础上,提出一种基于条件概率分布的数据异常状态自学习评估方法。定义了数据异常度的评价指标,实现了基于数据的风电机组低成本故障诊断,并利用风电场历史数据进行了算例训练与验证。结果表明,基于自学习概率模型的故障诊断方法可有效反映风电机组的异常状态信息。

The remote location of wind farms brings high failure incidencets and difficulties to maintenance work. Hence an effective fault diagnosis system can help to provide daily maintenance suggestions and specific repairing plan in order to decrease the maintenance cost and increase energy production of wind farms. This study analyzed and classified the data of wind turbines in SCADA system, proposed a type of self-learning method for evaluating abnormal state of data based on conditional probability distribution. The abnormal degree of data is defined to represent the direction and degree deviating from the normal state of monitoring data in the same scope. The study case makes use of wind farm historical data to train and verify the fault diagnosis system. The results indicate that the fault diagnosis method based on self-learning probability model can effectively reflects the abnormal state information of wind turbines.