化工学报 ›› 2018, Vol. 69 ›› Issue (S2): 252-259.DOI: 10.11949/j.issn.0438-1157.20181084

• 过程系统工程 • 上一篇    下一篇

基于长短期记忆神经网络的数据中心空调系统传感器故障诊断

王路瑶, 吴斌, 杜志敏, 晋欣桥   

  1. 上海交通大学制冷与低温工程研究所, 上海 200240
  • 收稿日期:2018-09-26 修回日期:2018-10-08 出版日期:2018-12-31 发布日期:2018-12-31
  • 通讯作者: 杜志敏
  • 基金资助:

    国家自然科学基金项目(51776118,51776119)。

Sensor fault detection and diagnosis for data center air conditioning system based on LSTM neural network

WANG Luyao, WU Bin, DU Zhimin, JIN Xinqiao   

  1. Institute of Refrigeration and Cryogenics, Shanghai Jiao Tong University, Shanghai 200240, China
  • Received:2018-09-26 Revised:2018-10-08 Online:2018-12-31 Published:2018-12-31
  • Supported by:

    supported by the National Natural Science Foundation of China (51776118, 51776119).

摘要:

数据中心空调系统的故障直接影响其能耗及运行可靠性。结合深度学习技术,利用长短时间记忆神经网络提出了一种空调系统传感器故障检测与诊断的方法。经实验验证,该方法可通过对液管温度传感器、排气温度传感器分别建立故障诊断模型,成功检测出传感器固定偏差故障和漂移偏差故障。对于无故障数据,该方法的检测正确率在90%左右;对于偏差程度大于该方法的最小检测偏差的传感器故障数据,其检测正确率在94%以上。

Abstract:

The faults of data center air-conditioning system directly results in increasing energy consumption and reducing operation reliability. Combined with deep learning technology, a fault detection and diagnosis method of sensors for air conditioning system is proposed by using LSTM (long short term memory) neural network. Verified by the experiment, the method can identify the fixed biases and drifting biases of both liquid line temperature sensor and discharge temperature sensor through building the fault detection and diagnosis models for them respectively. For the fault-free condition, the detection accuracy of this method is about 90%. For the sensor fault whose deviation degree is greater than the minimum detection deviation of the method, the detection accuracy is above 94%.

中图分类号: