基于传感器技术和I-LSTM算法的风电机设备运行故障检测及诊断研究

Equipment operation fault detection and diagnosis research based on sensor technology and machine learningSUN Ye1 ?? ZHAO Hua2??? GUO Lin3

摘要:

有效的故障检测与诊断将极大地提高风电机设备运行效率和可靠性，降低维修成本，保障生产过程的顺利进行。为实现高效率的设备故障预警与维护，研究基于传感器技术和机器学习的设备运行故障检测及诊断方法。首先对箱型图法和小波包降噪法等对传感器传输的数据信号进行预处理。然后利用双向长短时记忆网络构建时间序列预测模型。最后，基于预测残差和贝叶斯概率理论，设计了信号异常识别策略，以实现实时监测与故障预警。对提出的风电机设备故障监测模型进行性能分析，结果表明，研究所构建模型的诊断准确率为98.88%，无漏诊情况，误诊率在1.5%以下，在提前14小时以上进行预警。研究模型能够及时对风电机设备故障进行预警，同时能够在较高的准确率下对故障进行诊断。

Abstract:

Effective fault detection and diagnosis will greatly improve the operational efficiency and reliability of wind turbine equipment, reduce maintenance costs, and ensure the smooth progress of the production process. To achieve efficient equipment fault warning and maintenance, research on equipment operation fault detection and diagnosis methods based on sensor technology and machine learning. Firstly, preprocess the data signals transmitted by sensors using methods such as box plots and wavelet packet denoising. Then, a time series prediction model is constructed using a bidirectional long short-term memory network. Finally, based on prediction residuals and Bayesian probability theory, a signal anomaly recognition strategy was designed to achieve real-time monitoring and fault warning. Performance analysis was conducted on the proposed wind turbine equipment fault monitoring model, and the results showed that the diagnostic accuracy of the model constructed by the research institute was 98.88%, with no missed diagnosis and a misdiagnosis rate below 1.5%. Early warning was given at least 14 hours in advance. The research model can provide timely warning for wind turbine equipment faults and diagnose faults with high accuracy.