周致富 罗浩然 牟令

【摘  要】发电机轴承温升高是直驱风力发电机频发的故障，易导致风机长时间停机，维修难度大，造成设备重大损失。通过建立基于BP神经网络和Pearson相关系数的发电机轴承温升故障诊断模型，并以湖北荆门某风电场为例，试验结果表明：该方法能够较准确预测风机发电机轴承温升变化，通过将温度预测值与正常工况下的温度值进行误差对比，判断发电机是否处于正常工状，实现发电机潜在故障的预警，为风力发电机的故障诊断提供了一种切实可行的方法。

【Abstract】The bearing temperature rise of generator is a frequent fault of direct drive wind turbine， which is easy to lead to long-term shutdown of the wind turbine， which is difficult to maintain and cause heavy losses of equipment. By establishing a fault diagnosis model of generator bearing temperature rise based on BP neural network and Pearson correlation coefficient， and taking a wind farm in Jingmen， Hubei Province as an example， the test results show that the method can give a fairly accurate forecast for the temperature rise change of wind turbine generator bearing. By comparing the temperature prediction value with the temperature value under normal conditions， whether the generator is in normal working condition can be judged， and the early warning of potential fault of generator can be achieved， which provides a feasible method for fault diagnosis of wind turbine.

【關键词】BP神经网络;Pearson相关系数;故障诊断;发电机轴承温升;直驱风力发电机

【Keywords】BP neural network; Pearson correlation coefficient; fault diagnosis; temperature rise of generator bearing; direct drive wind turbine

【中图分类号】TM315                                             【文献标志码】A                                                 【文章编号】1673-1069（2021）01-0190-03

1 引言

经过多年的快速发展，我国风电装机规模已居世界前列，如何提高风力发电机的运行小时数，降低风机故障率和故障停机小时数己经成为研究热点。直驱风力发电机是一种无增速系统，叶轮直接带动永磁发电机旋转发电的设备，其中永磁发电机是其核心部件，发电机轴承温升高是其频发的故障，易导致风机长时间停机，维修难度大，极坏情况下易引起风机寿命缩短，甚至造成绝缘损坏，造成巨大经济损失。目前，风电机组SCADA的温度自动预警值多为对应部件材料的耐受国标温度，往往较高，故障早期的轻微温升变化难以预警。而且风力发电机轴承温升故障受风机多个关联变量的影响，温升变化与其受影响的多个关联变量构成非线性公式。……