基于压力估计的供水管网水力模型校核

Calibration of Hydraulic Model of Water Distribution System Based on Pressure Estimation

供水管网水力模型校核过程中,由于监测点个数较少而校核的参数过多,校核结果往往与实际不符。为使校核结果较为符合实际,提出一种基于压力估计的水力模型校核方法,根据实测监测点压力和入口流量数据估计非监测点的压力,并将估计的节点压力作为校核问题中的已知条件。以管道海曾威廉系数为校核参数,使用多层感知器(Multilayer perceptron,MLP)估计节点压力,使用遗传算法(Genetic algorithm,GA)优化目标函数。案例分析表明,提出的MLP-GA比GA校核效果更好,3个监测点压力绝对误差之和从37.81 m减小到33.65 m,降幅达11.0%;管道海曾威廉系数的绝对误差之和从933.5降低到847.0,降幅达9.27%。

In the calibration process of hydraulic model of a water distribution system,it is challenging to make calibration results consistent with the actual values due to the lack of sufficient monitoring locations but too many calibration parameters.In order to make the calibration results match the actual values more consistently,a hydraulic model calibration method based on pressure estimation was proposed,in which the pressures at non-monitoring locations were estimated based on the measured pressures at the monitoring locations and the inlet flow data,and then the estimated nodal pressures were used as known conditions in the later calibration process.The nodal pressures were estimated by Multi-layer Perceptron(MLP)with the Hazen-William coefficients of pipelines as the calibration parameters,and the objective function was optimized using Genetic Algorithm(GA).The case study showed that the proposed MLP-GA was more effective than the GA method,which reduced the sum of absolute errors of pressures at monitoring locations by 11.00%,from 37.81 m to 33.65 m,and the sum of absolute errors of Hazen-Williams coefficients of pipelines by 9.27%,from 933.5 to 847.0.