| The main goal of this research was to demonstrate the use of an advanced optimization technique that is suitable for watershed-level best management practice (BMP) optimization. This kind of simulation requires finding the optimal solution from many numbers of feasible alternatives. In this study genetic algorithms (GAs) were selected for optimization in part because they are known to search the solution space globally. Most previous work in developing an optimization tool for this problem has used GAs for optimization by individually considering two objectives: minimizing cost and pollutant reduction. The disadvantage of this approach is that some good solutions might be lost because the two objectives are considered separately.In this study a BMP placement tool was developed that searched the best solutions from two objective functions simultaneously. For each GA population, the BMP placement tool calculated pollution reduction by combinations of five BMPs for the seventeen watersheds. At the same time, BMPs total cost (Construction and Operation, Maintenance and Repair cost) was computed. Final results were selected from the best combination of both objectives.The input data including watershed area and pollutant loading for the optimization tool were adapted from the City of Los Angeles' Proposition O bond results. First generation of GA population was set with 100 chromosomes. Every chromosome was initialized by properties of seventeen watersheds. Each watershed DNA randomly contained properties of BMP type including pollutant removal rates, and total cost functions among five available set of BMPs.A sensitivity analysis of GAs parameters was performed by comparing fitness values to determine better parameters for the best solution result. The tested GA operators were the population size, the number of generations, the crossover rates, the mutation rates and the overlapping rates. In this study, population size of 100, crossover rate of 60%, mutation rate of 5%, overlapping rate of 60% and a number of generations of 300 gave the best results in terms of fitness values.Overall, the BMP placement optimization model performed well in reducing the each pollutant load and minimizing BMP total cost from the watershed. |
