YUAN Shao-chun,LI Di,CHEN Yao,et al.Automatic Calibration Procedure of Storm Water Management Model Parameters Based on Back Propagation Neural Network Algorithm[J].China Water & Wastewater,2021,37(21 21):125-130.
Automatic Calibration Procedure of Storm Water Management Model Parameters Based on Back Propagation Neural Network Algorithm
China Water & Wastewater[ISSN:1000-4062/CN:12-1073/TU]
volume:
第37卷
Number:
21 21
Page:
125-130
Column:
Date of publication:
2021-11-01
- Keywords:
- sponge city; storm water management model (SWMM); back propagation (BP) neural network; MATLAB; parameter calibration
- Abstract:
- The simulation accuracy of SWMM model is greatly affected by its calibrated parameter settings, while mathematical programming techniques such as linear and nonlinear programming are difficult to arrive at the overall optimum. In order to improve the accuracy of parameter calibration, the call functions of the SWMM calculation engine were coupled with the MATLAB software. The Latin hypercube sampling method was used to sample calibrated parameters to form the parameter groups, and then imported to the SWMM model. Finally, back propagation (BP) neural network training was conducted to complete the automatic calibration process of model parameters. The results of case study showed that the BP neural network algorithm could effectively complete the automatic parameter calibration process, and the mapping relationship between variables could be obtained after only 5 iterations. The Nash-Sutcliffe efficiency coefficient (NS) of the SWMM were all greater than 0.85 under four rainfalls with different intensity. Results indicated that this method could obtain the parameter settings that closed to the observed, with good simulation accuracy and stability, and could be used in the parameter calibration process of actual engineering project.
Last Update:
2021-11-01