基于神经网络的管网漏失定位实例研究

为了在复杂且不直观的供水管网中快速实时定位漏点,以某城市区域管网为对象,在其上游、中游和下游各选取3条没有经过训练的管道,进行不同损坏程度的漏损模拟,利用BP神经网络可以逼近任意的非线性映射特点,将模拟数据作为训练样本训练神经网络,建立漏损点位置与测压点压力之间的非线性关系,构建基于BP神经网络的实时漏点定位模型。结果表明,模型预测的漏点位置横纵坐标的平均相对误差分别为4.16%和1.40%,预测的漏点偏移距离最小为6 m。当漏损面积比为0.01时,泄漏流量只有1 L/s,完全可以达到快速定位。此研究成果对实际管网的应用提供了理论基础和技术支持。

Study on leakage location of pipe network based on neural network

In order to quickly locate leaks in a complex and unintuitive water supply network，we take the pipeline network of a city area as the object， select 3 untrained pipelines in the upstream，midstream and downstream，and perform leakage simulations with different degrees of damage. We use the simulation data as training samples to train the neural network by using BP neural network which can approximate arbitrary nonlinear mapping，and establish a nonlinear relationship between the leakage point and the pressure point to construct a realtime leak location model based on BP neural network. The results show that the average relative error of the horizontal and vertical coordinates of the leak point position predicted by the model is 4. 16% and 1. 40%，respectively，and the predicted leak point offset distance is a minimum of 6 m. When the leakage area ratio is 0. 01，the leakage flow rate is only 1 L/s，which helps to achieve rapid positioning. The research provides theoretical basis and technical support for the application of the actual pipe network.