Research On Fault Detection Of Air Conditioning Sensors In Data Center Computer Room

A variety of devices are stored in the data center computer room,running with a large amount of heat load.The device hardware itself has high requirements on the accuracy of the temperature and humidity control of the environment.Unsuitable environment may cause hardware damage.In order to maintain the normal working environment of the equipment hardware,the data-control rooms must be equipped with 24 hours of uninterrupted and stable operation of precision air conditioning.However,the air conditioner in the data-control rooms will inevitably fail during long-term operation,and the sensor fault is difficult to find in time.For high-dimensional,large-volume,high-value air-conditioning data,data-driven fault detection methods have been widely concerned by scholars in recent years.The research object of this article is an In-Row Air Cooled air conditioner with 33 sensors and high precision.This article explores the performance and usage conditions of the principal component analysis model and the deep neural network model in air conditioning sensor fault detection.Then,based on the real-time and accuracy of the air conditioning fault detection requirements in the data-control rooms,a combination of principal component analysis and deep neural network is proposed.According to the characteristics of high correlation between sensor measurement data of the data-control rooms,the detection capability and applicability of the fault analysis model of the principal component analysis are tested.Through data inspection,it is found that the principal component analysis model has different degrees and different degrees in a single working condition.All types of faults show excellent ability to detect abnormalities(87.86%-100%),and the speed of fault detection is extremely fast.However,the principal component analysis model has a certain adaptability to the working conditions,it performs well under the same working conditions,but performs poorly under the comprehensive working conditions.In order to solve the problem of adaptability of the principal component analysis model,the deep neural network model is introduced into the sensor fault detection research of the air conditioning system.In this article,the original data has been input into the deep neural network model for training firstly,and the fault detection result is between 72.64%-93.60%.When the principal variables obtained by principal component analysis as the input of the deep neural network model,the average fault detection efficiency increases from 85.73% to 92.67%.