基于人工神经网络的供水管网压力管理探索

Exploration of ANN-Based Water Pressure Management for Water Supply Distribution Network

供水管网的压力管理对实现按需供水、减少漏损和降低能耗具有重要的意义,实现压力管理的重要手段是对供水管网进行数学建模和数值仿真。但是,供水管网是非常复杂的大型非线性系统,按照传统的微观建模方法建立的供水管网模型往往精度不够,其数值求解效率也低,不适合基于这样的模型进行压力管理。该文提出了一种基于人工神经网络的供水管网压力管理的方法,即利用人工神经网络在供水管网的压力和流量之间建立非线性模型,并利用该非线性模型进行供水管网压力管理,而供水管网的压力管理则通过最优化问题的数值求解来实现。试验表明在不降低流量的情况下,供水的水压可降低1%。这对于减少供水管网的漏损、降低产销差率以及减少能耗具有重要的意义。

Pressure management on a water-supply network is significant to realize on-demand supply, to restrain the leakage, and to decrease energy consumption. Since the water-supply network is a complicate large scale nonlinear system, its numerical modeling, which is necessitated in the pressure management, cannot be accurately and efficiently fulfilled through the conventional microscopic methodology. A new ANN-based approach was proposed to model a water-supply network with 52 pressure probes and 32 flowrate probes. An acceptable nonlinear relationship between pressure measurements and flowrate measurements was established through a BP artificial neural network, which was utilized as the numerical model in the pressure management manifesting itself as a numerical optimization problem. The numerical test observes about 1% reduction in the pressure without flowrate violation. The novel approach and the numerical results presented in this paper are vital to the pressure management of water-supply networks.