Anomaly detectionof natural gas consumption for residential customers based on deep learning
GONG Yu
LI Qian
CAO Xin
Abstract：In recent years,smart gas has emerged as a significant trend in the natural gas industry,powered by information technology and artificial intelligence,accelerating industry's rapid development and transformation and representing enhanced productivity.Smart gas meters,playing a pivotal role in the smart gas,directly serve residential customers and monitor gas consumption data,enhancing the industry's safety and operational efficiency and boosting customer satisfaction.Aiming at the challenge of anomaly detection of natural gas consumption for residential customers,a novel anomaly detection model is proposed,integrating the Long Short-Term Memory(LSTM)algorithm from deep learning with the traditional Local Weighted Regression Smoothing(LOWESS)method.This model addresses the limitations of traditional anomaly detection methods,such as inefficiencies in modeling and an inability to pinpoint the exact time of outliers.The model's effectiveness was validated using production data from a natural gas company,demonstrating its superior efficiency and capability in identifying anomalies of gas consumption and the precise moments of outliers.This innovative solution offers a new approach to anomaly detection of gas consumption,promoting the industry's advancement in enhanced productivity.
Keywords：anomaly detection of national gas comsumptiondeep learningsmart gasnatural gas residential customersmodel effect
Publication Date：2024-07-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：8( 27-34 )