Incorporating a-priori expert knowledge in genetic algorithms

Conventional applications of genetic algorithm (GA) suggest using a random initial population. However, it is intuitively clear that any search routine could converge faster if starting points are good solutions. In this paper, a novel method is illustrated which incorporates a-priori knowledge in creating a fitter initial population while allowing for randomness among members of the population for diversity. Furthermore, the methodology is applied to optimization of a fuzzy controller´s membership parameters in a water desalination control process, in particular a brine heater temperature control problem. It is shown that the GA-improved PID fuzzy controller is able to reduce overshoot by 80 percent when compared to non-GA PID fuzzy controller.