用聚类算法及Hopfield神经网络布置环状管网

Clustering Algorithm and Hopfield Neural Network Used to Design Looped Network

为了降低供水管网的工程投资,提出了一种环状给水管网优化设计的方法,把环状管网布置问题用K-Means算法分解成许多小规模的环状问题,对于每一个小规模的环状管网用Hopfiled连续型神经网络求解环状的最短距离,然后把各个基环当成一个点,再用Hopfield神经网络以较大的概率求出最优路径。对某小区58个供水点进行了供水管网实例计算,结果表明:该供水管网的质心间最短距离为2.893 7 km,能量函数为递减状态且最终的能量函数值为1.447,再连接相邻聚类中的最接近的两个点就形成供水管网的环状布置。

This paper proposed a method of optimum looped network designing,which aiming at reducing the investment of water supply network.It disintegrated looped network designing into many small scale looped problems by the method of K-Means algorithm.The looped minimum distance of every small scale looped network was solved by using Hopfield neural network.And then each basic loop was viewed as a point and Hopfield neural network were used to obtain the optimum path with high probability.The method was proved feasible by examples with 58 water supply sources in a certain district and the results showed that the minimum distance of that water supply network centroid is 2.893 7 km,the ultimate results of the decreasing energy function is 1.447.At last,the arrangement of the circle water supply network can be obtained by connecting the nearest two points of adjacent clusters.