Spatio-temporal Fusion Model of Natural Gas Pipeline Condition Monitoring Based on Convolutional Neural Network and Long Short-term Memory Neural Network

Abstract

In order to make full use of the spatio-temporal scale change information of Supervisory Control and Data Acquisition (SCADA) data to accurately identify the health status of natural gas pipeline, reduce the economic losses caused by abnormal working conditions and improve the working efficiency of operators, based on the real-time operation data of natural gas pipeline, this paper adopts the combination of one-dimensional convolutional neural network (CNN) and long short-term memory (LSTM) network, and integrates the spatio-temporal characteristics to establish a pipeline working condition data model to predict the flow and pressure at the output end of natural gas pipeline. We use the residual error between the predicted value of the model and the real data value at the output end of the pipeline to represent the current operation state of the pipeline. Finally, through case study and comparison with the traditional LSTM neural network model, it is verified that the proposed spatio-temporal fusion model can more accurately predict the flow and pressure data at the output end of natural gas pipeline, and realize the monitoring of the running state of natural gas pipeline.