Optimization Of Stormwater Storage Tank Volume And Control Rules Based On Agent Model

Low Impact Development(LID)based sponge city construction has emerged as a result of frequent urban flooding in China.Rainwater storage tank is a kind of LID facility with significant economic benefits,which can reduce the risk of urban flooding and control the pollution of initial stormwater runoff.In this paper,based on the agent model,the volume and control rules of rainwater storage ponds are optimized respectively.The main research results are as follows:(1)Randomly generate 10,000 sets of values considering the actual construction conditions of the storage ponds;replace the bottom area of the storage ponds in the Storm Water Management Model(SWMM)input files(.inp)separately and save them as new INP files;run each INP file with SWMM to obtain the corresponding time series of outflow and SS concentration;The full life cycle cost of the storage pond,Total Suspended Solids(TSS)and Catchment Peak Outflow(CPO)values at the outfall are calculated and then combined with the bottom area of the storage pond to form a sample set of size 10,000;it is divided into a training set and a test set in the ratio of 7:3.The backpropagation neural network(BPNN)model was used for training and evaluation.The results show that the BPNN model predicts TSS and CPO better,and the MAE,RMSE,MAPE,DC,NSE and PBIAS errors of both the test and training sets are smaller.(2)The trained BPNN is used as a proxy to optimize the storage pool volume by using nondominated sorting genetic algorithm-Ⅱ(NSGA-Ⅱ)with the bottom area of the storage pool as the decision variable and minimizing the total life cycle cost,TSS and CPO as the objective.The results show that the method can obtain the optimal solution efficiently and reliably;if the training time of the BPNN model is neglected,the BPNN agent optimization model can save about 95.7%of the optimization time compared with the violent optimization based on SWMM.(3)Analyzing the basin outflow process corresponding to the Pareto solution set,it is found that:for different volumes of storage ponds,the TSS reduction rate is smaller than the CPO reduction rate;the larger the volume of storage ponds,the later the outflow flood moment,the larger the reduction rate of TSS and CPO;as the marginal cost increases,the TSS reduction per unit cost shows a decreasing trend,and the CPO reduction per unit cost has a small fluctuation,but The overall trend is decreasing.(4)After determining the optimal volume,the control depth of the storage pond is used as the decision variable to optimize the storage pond control rule with the objective of minimizing TSS and CPO,and the rationality of the optimal rule is analyzed.The results show that after optimizing the control rule of water depth of the storage pool,the peak moment of outflow flood is later and the reduction rate of TSS is significantly smaller than that of CPO.(5)For the Chicago design storm with return periods of 2,5,and 10 years,for the initial LID facility(no storage pool)scenario,the TSS reduction rate is the largest,the total outflow reduction rate is the second largest,and the CPO reduction rate is the lowest for the same rainfall;all three reduction rates decrease as the return period increases.The reduction of total load was the second most effective.