Optimization of low-impact development facility area based on surrogate model

Abstract Low impact development (LID) facilities are an important element of sponge cities. In this paper, a system of siting suitability indicators was established for three types of LID facility: green roofs, permeable pavement and rain gardens. Based on the superposition analysis tool of ArcGIS, the suitable area and its upper limit of the LID facilities in each sub-catchment were determined respectively. A non-linear function of the LID deployment area and rainfall characteristics was constructed separately using a multi-layer perceptron (MLP) as a surrogate model. The results showed that: (1) The upper limit of the suitable area for the deployment of green roofs in each sub-catchment ranged from 5.46% to 22.87%. The upper limit of the suitable layout area for permeable pavement was between 8.31% and 23.65%. The maximum percentage of suitable area for rain gardens was between 18.12% and 36.95%. (2) The final layout scheme of the LID facilities was selected when the total outflow from the sub-catchment was 10,700,000 m3, the peak discharge was 3.49 m3/s and the total investment in construction and management was 0.76 billion Yuan. (3) The total outflow reduction rate of the sub-catchment under the LID model ranged from 32.15% to 40.29%, and the peak discharge reduction rate ranged from 17.93% to 36.20%. (4) The computation time of the surrogate optimization model could be compressed to 0.37% of that of SWMM violent optimization. The surrogate optimization model had both good convergence measure and diversity measure. This paper provides new ideas for the optimal deployment of LID facilities.