基于无线传感与XGBoost模型的供水管网监测与分析

Integrated Monitoring and Analysis of Water Supply Network Based on Wireless Sensing and XGBoost Model

摘要:

基于自主研发的无线监测系统，在上海中心城区布设25个监测点位，对管道结构与运行环境等多源数据进行实时采集。采用XGBoost（eXtreme Gradient Boosting）算法对数据进行量化分析，系统分析各环境因素对管道结构安全影响。结果表明，管网运行中土压力、土体沉降、管顶与管底温度及孔隙水压力等监测参数变化，反映了回填土体、地基沉降及地下水的动态特征。利用监测数据训练的XGBoost模型预测精度优秀，预测值与实际值分布重叠率达98%，验证了模型的高精度与可靠性。此外，特征重要性分析显示，土体沉降变化占特征权重77.8%，是管道结构转角的主导驱动因素，为此，供水管网监测点应优先布设于“三交区域”等易发生变化的区域。

Abstract:

In this study， a self-developed wireless monitoring system was deployed at 25 monitoring sites in the central urban area of Shanghai to collect real-time multi-source data on pipeline structural conditions and surrounding environmental factors. The XGBoost （eXtreme Gradient Boosting）algorithm was applied for quantitative analysis to reveal the influence mechanisms of various environmental factors on pipeline structural changes. The results indicate that soil pressure， soil displacement， pipe crown and invert temperatures， and pore water pressure exhibit significant variations during pipeline operation， reflecting characteristics such as soil backfilling processes， foundation settlement， and groundwater dynamics. The XGBoost model trained on the monitoring data demonstrated excellent predictive performance， with a 98% overlap between predicted and observed values， confirming the model’s high accuracy and robustness. Furthermore， feature importance analysis showed that soil displacement changes accounted for 77.8% of the total feature weight， making it the dominant driving factor influencing changes in pipeline structural angles. Therefore， monitoring points for water supply networks should be preferentially installed in areas where the tri-cross junction is prone to change.