Multisurrogate Assisted Evolutionary Algorithm–Based Optimal Operation of Drainage Facilities in Urban Storm Drainage Systems for Flood Mitigation

Abstract

Drainage facilities are key infrastructure elements of urban stormwater drainage system (USDSs) for flood mitigation during short-duration extreme rainstorms. To obtain the optimal operation rules of drainage facilities within the shortest response time, a novel framework and approach was developed in this study. In the proposed approach, a complex USDS based on Storm Water Management Model (SWMM) was developed to simulate rainfall runoff and provide hydraulic characteristics, and was linked to the particle swarm optimization (PSO) algorithm to design short-term operation rules. Also, to overcome computational inefficiencies of the PSO algorithm to meet the requirements of response time, multisurrogate models in combination with radial basis function network (RBFN), kriging model, and artificial neural network (ANN) were coupled in the optimization search process to substantially reduce the number of evaluations. A practical case in Yinchuan city of China was then taken for the application. The proposed approach demonstrated high performance in flood mitigation compared with the current operation rules for pumping stations and water gates, with the core area inundations reduced to almost 0 (10 years), and the reduction rate being 71.22% (30 years) and 55.16% (50 years), respectively. In addition, the number of pump switches and the water level of the storage tanks were limited within the allowable range to ensure the reasonable and safe operation of the pumping stations. The new method required only 120 evaluations to obtain the optimal solution, which could provide sufficient response time for operators and thus serve as a practical and highly effective way to reduce urban inundation.