Research On Abnormal Detection Of Natural Gas Dehydration Device And Prediction Of Process Parameter Trend

In the process of natural gas collection,transportation and storage,natural gas containing water vapor will cause blockage and corrosion of pipeline equipment,so a dehydration device is required for dehydration process treatment after natural gas collection to ensure the quality of natural gas production.This paper studies the methods of abnormal detection of dehydration devices and the prediction of process parameter trends for the problems of equipment abnormality,lack of monitoring data,and difficulty of dynamic prediction of process parameters during the operation of natural gas dehydration devices.The main works of this article are as follows:(1)In view of the difficulty in detecting abnormal equipment in the dehydration device,two methods of abnormal identification were studied.The multiple regression sequence method separately establishes regression models for the groups generated by hierarchical clustering,and then combines model prediction values and true value residuals to identify anomalies.The principal component analysis method extracts the main change pattern from multiple parameters,so that the abnormality can be reflected on the principal component for easy detection.These two methods confirm each other,and can simultaneously find the abnormal state in the device.(2)A large number of missing values may appear during the monitoring process of the dehydration device,and directly deleting the entire missing data may mistakenly delete other valuable parameter data.In this paper,using the characteristics of functional data analysis,a functional principal component analysis method based on expectations is studied to estimate the original monitoring curve.Experiments show that this method can better estimate missing data.Aiming at the problem that the water dew point detection instrument of the dehydration device is easily damaged and the water dew point cannot be obtained in real time,combined with the hysteresis of the water dew point change and the characteristics of functional data analysis,a model function type generalized addition between the dehydration device monitoring data and the water dew point is established Sexual model.Compared with the functional model FLM and the discrete model random forest respectively,the FGAM model can not only predict the change trend of water dew point,but also capture the small characteristics of data changes,and can also predict the extreme points with large changes effect.(3)Combined with the characteristics of functional data,a method for dynamically predicting the process parameters of the dehydration unit based on the FPCA-FGAM model is studied.This method uses the FPCA to dynamically predict the changes in the monitoring data of the dehydration unit,and combined with the FGAM model to realize the dew point of the water Dynamic prediction.The dynamic prediction effects of 30%,60%,and 80% of the monitoring data are compared respectively.As the monitoring data increases,the dynamic prediction effect is closer to the true value.Finally,the dynamic prediction method is applied to the dynamic prediction of water dew point.The results show that the FPCA-FGAM model can achieve dynamic prediction of process parameters such as water dew point.(4)Aiming at the problems of abnormality detection and trend prediction in the monitoring process of natural gas dehydration equipment,Py Qt was used to develop an abnormality detection system for dehydration equipment,using multiple regression sequence and principal component analysis method to identify abnormalities,and FPCA-FGAM model for process Dynamic prediction.And introduces the system's multi-thread design,optimizes the system interactivity.