Comparative Study On Hydraulic Model Calibration Of Water Supply Network Based On Genetic Algorithm And Particle Swarm Optimization Algorithm

An accurate hydraulic model can fully reflect the operational conditions of a water distribution network(WDN).However,it is impossible to measure the Hazen-Williams coefficient of a pipeline directly.Since the accuracy of this kind of parameter has a significant impact on simulation results of a hydraulic model,the hydraulic model calibration technique has become a research hotspot for scholars across the world.This thesis established a mathematical optimization model for the implicit calibration of the Hazen-Williams coefficients in the hydraulic model of a water distribution network,located in the HX development zone of a city in Central China.This work compared two prevalent evolutionary algorithms,namely the Genetic Algorithms(GAs)and the Particle Swarm Optimization(PSO),on their efficacy of solving the hydraulic model calibration problems of the case study.The main conclusions were obtained as follows.(1)Both GAs and PSO can effectively solve the calibration problem of the Hazen-Williams coefficients of the hydraulic model of the WDN in the HX development zone.The overall calibration accuracy is higher than(or very close to)the threshold specified in the domestic hydraulic model calibration standards.The results of calibrating only top 20 important pipes show that the absolute pressure errors are less than or equal to ±1m at 56.0% of the monitoring positions;less than or equal to ±2m at88.4% of the monitoring positions;and less than or equal to ±4m at 99.6% of the monitoring positions.(2)This thesis proposed an orthogonal experiments based technique to investigate the parameterization issues of evolutionary algorithms.It helps identify the best and the worst parameter combinations of GAs and PSO.It is found that PSO is more sensitive to parameter tuning compared with GAs.The comparison experiment results show that when only important pipes were calibrated,the results obtained by the worst parameter combination of PSO were slightly lower than the threshold specified in the domestic hydraulic model calibration standards.Only 49.8% of the monitoring positions achieved the absolute pressure errors less than or equal to ±1m in the extended period simulation.There are 85.8% of the monitoring positions had the absolute pressure errors within±2m;and 98.7% of the monitoring positions were within ±4m.Compared with the results obtained by the best parameter combination of PSO,the performance of PSO with the worst parameter combination shows an obvious deterioration.In contrast,the calibration results obtained by the worst parameter combination of GAs was generally the same as the ones obtained by the best parameter combination.(3)For the hydraulic model of the WDN in the HX development zone,the calibration performance of GAs is better than that of PSO,so GAs are recommended as a priority when calibrating the actual WDN models.When all pipes were considered during model calibration,the percentages of the absolute pressure errors within ±1m across all the monitoring positions by GAs and PSO were 63.6% and 60.4%respectively;and the absolute pressure errors within ±2m were 91.6% and 91.1%respectively.Both algorithms were able to ensure the absolute pressure errors within±4m.(4)When GAs and PSO were used to calibrate the Hazen-Williams coefficients of the WDN,taking the coefficients of all pipes as decision variables in the optimization model can obtain better simulation accuracy than that only the important pipes were involved.The results show that the application of GAs can increase the percentage of the absolute pressure errors within ±1m,±2m,and ±4m by 7.6%,5.8%,and 0.4% at monitoring positions,respectively.In contrast,the application of PSO can increase the percentage of the absolute pressure errors within ±1m,±2m,and ±4m by 4.4%,5.3%,and 0.4% at monitoring positions,respectively.The contributions of this thesis provide a certain theoretical basis for selecting a reasonable method for calibrating hydraulic models.It also provides a new idea for quickly finding effective and efficient combinations of parameters when evolutionary algorithms are considered for model calibration.