ABSTRACT:

Optimization algorithms are powerful tools to help identify cost-effective designs of urban green infrastructures such as low impact developments (LIDs). An urban inundation model named PCSWMM could integrate a one-dimensional (1D) drainage model and a two-dimensional (2D) overland flow model, and offer rich information of flood extent and depth for robust damage assessments. This work proposes a Surrogate Optimization method for LID design with an integrated 1D-2D inundation model (SOLID-1D-2D). Various LID configurations are simulated by the inundation model. For each LID configuration, design rainfalls with different return periods are generated to drive the model and produce a series of probable flood events. A probabilistic approach is adopted to calculate the expected total flood damage over the life cycle of LIDs. In the long run, the objective of optimization is to minimize the sum of life cycle cost of LIDs and the corresponding total flood damage. For solving this expensive and high-dimensional (103-dimensional) simulation-optimization problem, SOLID-1D-2D adopts a novel surrogate global optimization algorithm to save the budget for model executions by improving the efficiency of global searching. Through a case study, the performance of SOLID-1D-2D is compared with a traditional particle swarm optimizer-based method in terms of achieved best solutions, algorithm reliability, and computational efficiency. SOLID-1D-2D is a promising alternative to the traditional method for achieving a better trade-off between life cycle cost of LIDs and the benefits of flood reduction.