Research On Missing Data Completion Leakage Fault Diagnosis Method Of Distributed Pipeline Network Based On Data Drive

With the rapid development of China’s industrial modernization,the consumption of petroleum and other fossil energy is increasing day by day.In order to carry out oil transportation safely and efficiently,pipeline transportation has become the best choice at present.Because of the special industrial attributes of petroleum,the safety and stability of its transportation equipment and system are very strict.In the process of oil transportation,if the pipeline breaks down and causes leakage failure,it will lead to a large number of casualties and economic losses.From the above analysis,it can be seen that it is of great significance to implement the necessary leakage fault diagnosis for oil pipelines.Therefore,according to the current situation and characteristics of oil pipeline and pipeline network,this thesis carries out further research on pipeline leakage fault diagnosis.The main innovations are as follows:Firstly,In order to better find the characteristics of pipeline network and integrate more pipeline network information into the model of feature extraction,this paper first summarizes and designs the physical information system framework of oil pipeline network.According to the framework proposed in this paper,the operation and physical state of the station yard and the whole pipe network have been digitized in practical applications,making it easier for researchers and staff to obtain the data to be analyzed and laying a solid foundation for decisionmaking.Secondly,In view of the fact that the work conditions of oil pipelines are becoming more and more diversified,this paper focuses on the problem of missing pressure data of a station in the field pipe network caused by the damage of pressure gauge caused by hardware failure or communication failure.A data-driven algorithm to complete the missing pressure data of oil pipeline network is proposed by analyzing the data characteristics of stress information,this paper puts forward a structure combining the generated antagonistic network and self-attention network,and puts forward corresponding solutions to the problems existing in the original network in completing the task of pressure information generation.Finally,the above model is trained with actual pipeline data,and the rationality of the model is verified by evaluating multiple evaluation indicators,including Pearson correlation coefficient and mean square error.Finally,the influence of different stress factors on the learning effect of the model is explored.Thirdly,on the basis of fault diagnosis of single pipeline leakage,considering the huge topology structure of oil pipeline network and the characteristics of interconnection between pipelines,this thesis proposes a fault diagnosis method of pipeline network based on Convolution Neural Network from the perspective of mining more effective features and reflecting local correlation of data.The network model of pipeline network fault diagnosis based on Convolution Neural Network is designed,and the forward and backward propagation of pipeline network data samples in Convolution Neural Network is analyzed.On this basis,the real data of pipeline network are analyzed and organized,and the fault diagnosis network model of pipeline network is trained.Then,the effectiveness of the fault diagnosis method based on Convolution Neural Network is verified by analyzing its fault diagnosis effect through experiments.