基于人工神经网络的供水管网爆管定位方法

Pipe Burst Localization in Water Distribution Systems Based on Artificial Neural Network

供水管网发生爆管事故后,快速确定爆管位置,可以实现迅速抢修,有效降低事故的损失。针对爆管定位问题,本文基于人工神经网络（ANN）,建立爆管位置与事故时压力监测点的压力变化率之间的非线性映射关系,构建了ANN爆管定位模型,并选取了一个供水管网案例,引入相关系数（R-2）指标评估模型的精度,验证了方法的可行性。此外,分析了不同监测点组合对模型定位精度的影响,发现监测点组合均匀分布在管网内部时,模型定位精度越高。

Once pipe burst events occur in water distribution systems, the prompt pipe burst localization can achieve the rapid repair of pipes and the reduction of incident loss. Concerning this issue, the paper presents the mapping of nonlinear relation between pipe burst location and the pressure variation rate at monitoring points during the events based on artificial neural network（ANN）. The ANN model for the pipe burst localization is set up. A water distribution system case is used to validate the feasibility of the method using correlation analysis（R-2）. Moreover, the impact of different combinations of monitoring points on localization accuracy is analyzed. The results show that the accuracy of model is higher when the monitoring points are installed uniformly within the water distribution systems.