matlab latin hypercube sampling gumbel distribution

Thus, the developed SM can be reliably integrated into a simulation-optimisation framework for BRC optimal design.Īhiablame LM, Engel BA, Chaubey I (2012) Effectiveness of low impact development practices: literature review and suggestions for future research.

Additionally, a comparison of performance for SMs revealed that using 50 samples yields superior performance compared to larger samples. Results indicate that the SM can very accurately replicate SWMM results with Nash–Sutcliffe Efficiency ranging from 0.97 to 0.99. Latin Hypercube Sampling (LHS) was used to generate training samples ranging from 50 to 200 samples for 18 rainfall events. Following a case study for BRC design in the City of Toronto, an Artificial Neural Network (ANN) was used as an SM to SWMM. Therefore, SM predictive performance is highly dependent on the training dataset, yet questions remain on how to select the best training set for optimal performance. The SM is constructed by generating outputs from the original model (SWMM) on a set of known values and subsequently training the SM. SMs are approximate representations of more complex models. In this research, a Surrogate Model (SM) is proposed as an alternative tool for designing and modelling BRCs. Modelling BRC designs is typically performed using the Stormwater Management Model (SWMM), which can be time consuming given the iterative design process and nature of physical-based models. Existing design guidelines for BRCs offer a range of design parameters values, varying by region and climate conditions, meaning manual BRC design can be iterative and time consuming, with an optimal solution being difficult to arrive at. Bioretention cells (BRCs) are an emerging technology used for urban stormwater management.

Matlab latin hypercube sampling gumbel distribution