基于LSTM-WPHM模型的风机轴承故障报警与寿命预测方法

A Hybrid LSTM-WPHM Method for Fault Alarming and Life Estimation of Wind Turbine Bearings

受极端恶劣运行环境和复杂多变运行工况的影响,风电机组的轴承存在很高的故障率,进而导致风电机组容易发生非计划停机,减少风机的发电量,增加风场的运维成本。考虑风机运行的非稳态性和不确定性,提出一种基于SCADA数据的多变量LSTM时序预测模型和WPHM可靠性分析模型相结合的混合方法,实现风电机组轴承故障预警和寿命预测。该方法通过LSTM时序预测模型集成多变量对轴承故障的影响,提高了轴承可靠性分析精度,并简化了可靠性模型的参数估计。同时,以模型预测残差为状态指标,建立WPHM可靠性分析模型,以累积失效概率为失效指标,实现故障报警和寿命预测的有机结合。利用某海上风场实际机组的故障信息和监测数据,验证了该方法的准确性和有效性。

Owing to extremely harsh ambient environment and complicated operating conditions,the bearings of wind turbines have a high failure rate,which leads to unplanned shutdown of a wind turbine,thus reducing its generating capacity while increasing its operation and maintenance cost.This paper presents a hybrid intelligent probabilistic method for simultaneous fault warning and remaining life estimation for a wind turbine bearings by adeptly combining multivariate LSTM prediction model and WPHM reliability model based on SCADA data.This method integrates the impact of multiple critical variables on bearing faults through the LSTM model to improve the accuracy of reliability analysis.The prediction residual is employed as the state indicator to establish a WPHM model.The cumulative failure probability is obtained as the failure index for fault alarming and remaining life estimation of a turbine bearings.The accuracy and effectiveness of the proposed method are verified by using the fault information and monitoring data for turbines in a real-world offshore wind farm.