Data-Driven In-Situ Diagnosis Method for Gas Metering Systems in Industrial and Commercial Users

Accurate on-site diagnosis of abnormal states in industrial and commercial gas metering systems is of great significance for reducing gas supply-demand discrepancies and ensuring the safe operation of gas pipelines. This study addresses the challenges of on-site diagnosis in existing gas metering systems and proposes a data-driven method for monitoring abnormal states of gas metering systems based on users' SCADA data. The method utilizes the correlations between pipeline flow, pressure, and temperature to extract features such as pipeline temperature-pressure correlation coefficient, pipeline temperature difference coefficient, and pipeline pressure fluctuation coefficient based on wavelet decomposition. These feature data are then fed into a particle swarm optimization-based support vector machine (PSO-SVM) model to achieve fault diagnosis for the gas metering system. The method was validated using data from over 700 industrial and commercial users in a certain region. The results show that the method achieves an accuracy of 96.7%, precision of 96.9%, recall of 97.1%, and an F1 score of 0.97 for diagnosing both gas-using and non-gas-using users, with particularly high accuracy in diagnosing suspected gas-using users. The experimental results demonstrate that the proposed method enables effective and accurate on-site diagnosis of abnormal states in gas metering systems.