| 基于自编码网络模型的风机故障检测研究 |
| |
| 引用本文： | 章浩伟,周琪馨,任筱倩.基于自编码网络模型的风机故障检测研究[J].教育技术导刊,2019,18(9):158-162. |
| |
| 作者姓名： | 章浩伟 周琪馨 任筱倩 |
| |
| 作者单位： | 上海理工大学 医疗器械与食品学院，上海 200093 |
| |
| 摘 要： | 为了保障风机的正常运行，提出基于自编码（Autoencoder，AE）的风机故障检测方法。依据法国风机ENGIE公开的风速传感器数据，建立欠完备自编码模型（UAE）、去噪自编码模型（DAE）与收缩编码器（CAE）模型，对风机风速传感器数据进行编码和解码，计算重构误差并设定阈值进行故障检测。用多风机风速传感器数据建立PCA模型并与去噪编码器模型对比。根据ROC曲线与AUC值对比，得出欠完备自编码模型、去噪自编码模型、收缩编码器模型均可用于风机异常检测，且收缩编码器效果最好的结论。PCA模型也可用于故障检测，同时多风机故障检测效果高于单一风机。
|
| 关 键 词： | 风机故障 自编码 去噪自编码 收缩编码器 |
| 收稿时间： | 2018-12-12 |
Anomaly Detection for Wind Turbines Based on Autoencoder Model |
| |
| ZHANG Hao-wei,ZHOU Qi-xin,REN Xiao-qian.Anomaly Detection for Wind Turbines Based on Autoencoder Model[J].Introduction of Educational Technology,2019,18(9):158-162. |
| |
| Authors: | ZHANG Hao-wei ZHOU Qi-xin REN Xiao-qian |
| |
| Institution: | College of Biomedical Engineering， University of Shanghai for Science and?Technology， Shanghai 200093， China |
| |
| Abstract: | In order to achieve anomaly detection for wind turbines and keep the running of wind turbines，an anomaly detection method based on autoencoder is proposed. We obtained the ENGIE open data of the wind speed sensor on the French wind turbines and built the under-complete autoencoder model， the denoising autoencoder network model and the contractive autoencoder model. The power reconstruction error by encoding and decoding the power of wind turbines was calculated and the appropriate threshold value was selected as the decision criteria for anomaly detection. The PCA was built to be compared with denoising autoencoder model with four wind speed sensor data. According to the ROC curve and the AUC value， the under-complete autoencoder model， the denoising autoencoder model and the contractive autoencoder model can all be used for anomaly detection for wind turbines， and the contractive autoencoder model is better than the other two models. The PCA model can also be used for fault detection and the wind speed sensor data of four wind turbines is better than one single sensor data. |
| |
| Keywords: | anomalies of wind turbines autoencoder denoising autoencoder contractive autoencoder |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
