Leakage diagnosis and localization of the gas extraction pipeline based on SA-PSO BP neural network

Leakage is a potential accident affecting gas pipeline reliability. To diagnose leakage of the main gas extraction pipeline in coal mines, a pipeline leakage diagnosis method based on Simulated Annealing (SA) and Particle Swarm Optimization (PSO) collaborative optimization Back Propagation Neural Network (BPNN) was proposed. On the basis of the mapping relationship between the location of the leakage point and monitoring value, the SA-PSO BPNN leakage diagnosis model was established and its reliability was verified. The results shown that the larger the leakage diameter at the same location, the greater the rate of change in flow and pressure at each monitoring point. The flow and pressure close to the leakage point decreased greatly when the leakage occurred, but these parameters changed slightly far from the leakage point. Compared leakage identification accuracy of BP, SABP, PSOBP and SA-PSO BPNN in different leakage points, the SA-PSO BPNN model had higher accuracy. The Area Under Curve (AUC) value under different leakage condition was 0.614-0.940, and the test accuracy was 79.61%, 87.21% and 92.25% respectively when input 2, 3 and 5 sets of monitoring point parameter. The diagnostic accuracy of the SA-PSO BPNN model was 93.33% through the verification samples. The SA-PSO BPNN diagnosis model provided theoretical guidance for realizing timely and accurate leakage detection.

Automated summary

This study proposed a pipeline leakage diagnosis method for the main gas extraction pipeline in coal mines based on Simulated Annealing (SA) and Particle Swarm Optimization (PSO) collaborative optimization Back Propagation Neural Network (BPNN). The SA-PSO BPNN leakage diagnosis model was established and its reliability was verified by mapping the location of the leakage point and monitoring value. The results showed that the SA-PSO BPNN model had higher accuracy in identifying leakage compared to other models. The diagnostic accuracy of the SA-PSO BPNN model was 93.33% through the verification samples.