Research On Pipeline Leakage Monitoring And Early Warning In Unattended Natural Gas Station

With the increase of the proportion of natural gas in the energy consumption structure,the construction of natural gas pipeline network has developed rapidly.Fire and explosion accidents caused by natural gas station pipeline leakage and other factors often cause serious casualties,economic losses and environmental pollution.Therefore,it is of great significance to the study of the gas pipeline leakage monitoring and early warning technology.This paper takes the pipeline leakage of natural gas station as the research object,use the image data of pipeline leakage of natural gas station and the internal pressure data of pipeline of natural gas station,proposes the pipeline leakage monitoring technology of natural gas station based on deep learning,as well as a natural gas station pipeline leakage warning technology based on unbalanced data processing and gate recurrent unit(GRU).Finally,we develop the unattended pipeline leakage monitoring and early warning system of natural gas station,realize accurate identification and early warning of natural gas pipeline leakage in the actual project,and improve the safety level of natural gas stations.Aiming at the problems of low accuracy and slow response in the process of natural gas pipeline leakage monitoring process,a deep learning-based pipeline leakage monitoring method of natural gas station is studied.By using the existing data set of natural gas pipeline leakage,this article firstly uses the deep learning segmentation network U~2-Net to process the image of pipeline leaking gas.From the segmentation effect,it can completely extract the region of leaking gas without being affected by the similarity of objects.Then,we use the improved deep neural network VGG16 to identify pipeline leaking.The method proposed in this paper is compared with the existing natural gas pipeline leak monitoring methods.It has higher accuracy,F1-score,and average accuracy,and is an efficient natural gas pipeline leak monitoring method.Because most of the data in natural gas pipelines are unbalanced time series data,a natural gas pipeline leakage early warning method based on unbalanced data processing and GRU is studied to address the problem that unbalanced characteristics can cause a decline in the early warning rate of natural gas pipeline leakage.The NE-AHC-MAHAK oversampling method is used to process the pressure data in the natural gas pipeline,to achieve the effect of data balance.By calculating the average ranking of each oversampling method,the NE-AHC-MAHAK oversampling method has advantages in terms of evaluation indicators.Input normal working condition pressure data into a GRU model for training to predict the pressure value at the output end of the pipeline,and then obtain the threshold of the difference between the predicted value and the actual value of the model.When data from other working conditions appear,the pipeline data at this time will not adapt to the GRU model.Therefore,the working condition of the natural gas pipeline can be judged based on the fluctuation of the difference,and a timely warning should be given when leakage occurs.The method in this paper has advantages in terms of error warning rate,false alarm rate,and missed alarm rate,indicating that the method in this paper can more quickly and effectively early warning the leakage of natural gas station pipelines.The above two methods are systematically integrated to design the unattended gas station pipeline leakage monitoring and early warning system.This article uses the image data of natural gas pipeline leakage and pipeline pressure to provide data support for the early warning system.By training and loading the above two methods into the early warning system,an early warning system with strong real-time performance and high accuracy is developed.In this way,the safety management level and intelligence level of natural gas station and yard pipeline can be improved,and the intelligent pipe network can be built.