基于多变量时序数据的窨井燃气泄漏预警系统设计  

Design of an Early Warning System for Gas Leakage in Manholes Based on Multivariate Time-Series Data

作　　者：杨光[1,2] 周雨聪 陈凯源 马俊杰 史娜 YANG Guang;ZHOU Yücong;CHEN Kaiyuan;MA Junjie;SHI Na(North University of China,College of Information and Communication Engineering,Taiyuan 030051,China;Shanxi Key Laboratory of Intelligent Detection Technology&Equipment,Taiyuan 030051,China;College of Mathematics,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学信息与通信工程学院,太原030051 [2]智能探测技术与装备山西省重点实验室,太原030051 [3]中北大学数学学院,太原030051

出　　处：《计算机测量与控制》2025年第4期277-283,共7页Computer Measurement &Control

基　　金：山西省科技成果转化引导专项(202304021301028);山西省科技成果转化引导专项(202204021301044);中央引导地方科技发展资金(YDZJSX20231A025);山西重点研发计划项目(202202010101007)。

摘　　要：为解决单变量阈值法在窨井燃气泄漏监测中存在误判率高、响应延迟及环境参数缺失的问题,设计了基于多变量时序数据的智能监测与预警系统;通过构建多传感器数据采集终端,实时获取甲烷浓度、温度、气压等多维度环境参数,建立区域窨井燃气时序数据库;基于Bi-LSTM算法设计了可并行处理多变量时序数据的预测模型,利用模型的双向特征提取能力捕捉燃气泄漏动态关联特征;经实验测试,在独立测试集的模型预测准确率为98.3%,较传统单变量阈值法平均预警时间提前48.5 min;经实际工程应用,满足了复杂城市窨井场景下燃气泄漏的多节点协同监测需求,有效克服环境干扰并实现高鲁棒性预警,同时支持分钟级实时响应与规模化部署。To address the issues of high misjudgment rates,response delays,and environmental parameter deficiencies in traditional single-variable threshold methods for manhole gas leakage monitoring,this study designs an intelligent monitoring and early warning system based on multivariate time-series data.By constructing a multi-sensor data acquisition terminal,the system dynamically captures multi-dimensional environmental parameters including methane concentration,temperature,and atmospheric pressure,establishing a regional manhole gas time-series database.A prediction model based on the Bi-LSTM network is developed to parallel process multivariate time-series data,leveraging the model's bidirectional feature extraction capability to capture dynamic correlation characteristics of gas leakage.Experimental results show that the model achieves a prediction accuracy of 98.3%on independent test datasets,with the average early warning time advanced by 48.5 minutes compared to traditional single-variable threshold methods.Practical engineering applications demonstrate that the system meets the requirements of multi-node collaborative monitoring for gas leakage in complex urban manhole scenarios,effectively overcoming environmental interference while achieving high-robustness early warnings.It also supports minute-level real-time response and large-scale deployment.

关 键 词：燃气泄漏 异常检测 Bi-LSTM 多变量时间序列 预警 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]