基于机器学习的风电机组齿轮箱故障预警

Fault Early Warning of Wind Turbine Gearbox Based on Machine Learning

风电机组长期工作在恶劣的环境中,导致故障频发,运用合理高效的方法对风电机组部件进行故障预警,具有十分重要的现实意义。首先利用3σ-中位数准则对数据进行预处理,然后利用非线性状态估计(nonlinear state estimate technology, NSET)方法对风电机组的齿轮箱温度进行预测。当齿轮箱工作出现异常时,预测值与实际值的残差增大,超出预先设定的阈值发出报警信息。实验结果表明,提出的3σ-中位数准则组合法能够有效识别数据中的异常值,效率高,清洗效果好。对处理后的数据进行NSET建模,利用NSET模型实现了齿轮箱的故障预警。

Wind turbines work in harsh environments for a long time, resulting in frequent failures. It is of great practical significance to use reasonable and efficient methods for early-warning wind turbine components. In this paper, the 3σ-median combination method is used to preprocess the data, and then the nonlinear state estimate technology(NSET) method is used to predict the gearbox temperature of wind turbine. When the gearbox works abnormally, the residual error between the predicted value and the actual value increases, and an alarm message is sent out when the gearbox exceeds the preset threshold value. The experimental results show that the 3σ-median criterion combination method proposed in this paper can effectively identify the outliers in the data, which has high efficiency and good cleaning effect. Then, the processed data are modeled by NSET,and the gearbox fault warning is realized by using the NSET model.