Chance-Constrained Optimization Using Genetic Algorithms: An Application in Air Quality Management
Daniel H. Loughlin, S. Ranji Ranjithan · 2001
We explore the use of chance-constrained genetic algorithms (CCGAs) for stochastic optimization in air quality management. CCGAs allow uncertainties in optimization model parameters to be considered explicitly in the design of least cost strategies. To incorporate uncertainties into the CCGA, the original fitness function is replaced with a Monte Carlo (MC) simulation using Latin Hypercube Sampling (LHS). Strategies are penalized according to how frequently they fail to meet an air quality target. We demonstrate this approach for an air quality problem involving the control of tropospheric ozone. For this problem, we use the CCGA in a multiobjective context to generate a tradeoff curve between control costs and reliability of meeting an air quality target. This analysis is not tractable using classical stochastic mathematical programming techniques such as chance-constrained programming.