Research On Remote Intelligent Ultra-Early Warning Method Of Continuous Chemical Process

The continuous chemical production process has the characteristics of short sampling time,strong dynamics,and non-linearity.Monitoring and early warning of process conditions are particularly important for ensuring safe and stable production activities.The existing warning methods are based on short-term predicting.However,on-site workers cannot handle complex operations for a short period of time.Therefore,it is necessary to advance the warning time to as much as a few minutes or more to ensure adequate operation time.While early warning time,it will cause a large number of false alarms and missed alarms.It will cause certain problems to the on-site personnel.How to guarantee the accuracy of the early warning as possible is a major problem.In response to the above issues,this thesis develops the following four areas of research:(1)For the chemical industry,the sampling interval is short and the dynamics are strong.Short-term parameter prediction can not provide enough time for the on-site personnel,then combined with the improved PSO and least squares support vector machine monitoring and early warning method for abnormal working conditions is proposed.The least squares support vector machine is used to train the normalized process data.The improved particle swarm optimization algorithm is used to quickly and accurately optimize the kernel parameter and penalty factor in the least squares support vector machine.Compared to the least squares support vector machine prediction model and the normal particle swarm optimization least squares support vector machine prediction model,the prediction error is reduced,and the warning time is advanced.(2)In view of the existing monitoring and early warning methods for the overall operating conditions,most of methods adopt statistical monitoring and result to many false alarms and omissions.A combination warning mechanism based on multi-source information fusion for abnormal conditions is proposed.First,the improved PSO-LSSVM algorithm is used to perform super-early prediction of individual parameters.Then,ADF test for individual parameters is used to determine whether the trend is stable.Last,the cluster center is calculated using the K-means clustering method,and the Mahalanobis distances are calculated using outlier analysis.According to these three kinds of situations,if there are two or more of them is abnormal,the overall condition is determined to be abnormal and an early warning is issued.Compared with the KPCA method,the warning is advanced 31.7s and the false alarm rate and false negative rate are reduced.(3)In view of the fact that the existing early warning system does not analyze and deal with the cause of the early warning,the existing diagnostic methods have the problem of slow on-line analysis and large amount of work in the early period,an intelligent fault diagnosis method for abnormal conditions is proposed,and the correlation between parameters is analyzed in real time.The Spearman correlation coefficient method is used to analyze the correlation coefficient matrix between the calculated parameters.After converting into a Boolean matrix,the association rule algorithm is used to analyze the matrix,and the strong association rules between parameters are obtained,and then the cause of the parameter early warning is determined.After three groups of abnormal alarm events are verified,the proposed method can effectively analyze the cause of the alarm.(4)For the problem of the existing remote monitoring system slow data transmission speed,easily losing data,and not enabling experts to accurately understand the on-site operating conditions,a set of 3G network remote intelligent monitoring and early warning system to achieve remote expert decision is developed.According to the principles of 3G network,network protocols and wireless transmitter principles,the composition of remote monitoring systems is researched.A scheme for interconnecting a 3G network with an Internet wide area network is designed,based on serial communication theory,to realize serial communication and data remote transmission.Combined with the production process of the gas separation device,establish a remote real-time monitoring and ultra-early warning module is developed.