数据驱动的城市供水管网异常事件侦测方法

针对供水管网爆管等监控问题，提出数据驱动的供水管网异常事件侦测方法.在某市供水管网独立计量区域（DMA）内，采用打开消防栓方式进行模拟爆管实验，在对实测数据进行小波降噪等预处理的基础上，分别采用基于VARX模型的差异分析方法和统计过程控制（SPC）方法进行异常侦测，再使用贝叶斯网络综合推理.通过引入信噪比分析，降低了综合推理后的误报次数.实验结果表明，该异常侦测方法可以达到较高的准确性.
Abstract: A data driven method for abnormal event detection of water supply network was proposed in order to solve the pipe burst monitoring problem. Pipe burst experiments were conducted by opening fire hydrant in a district metering area (DMA) of an urban water supply network. Based on the measured data of wavelet de-noising preprocessing, the difference analysis method of VARX model and statistical process control (SPC) methods for abnormal event detection were used. Then Bayesian network was used for comprehensive inference, and false alarm after comprehensive inference was reduced with signal-noise ratio analysis. The experimental results show that the data driven method can achieve higher accuracy.