抄録
Without considering various uncertainties involved in system model, it happens likely that the optimum solution is useful only in specific situation or insignificant at worst. Concerning with parameter uncertainties, in this study, we propose a new genetic optimization method to obtain an insensitive solution. We first formulate the problem as a statistical optimization problem that will maximize the expectation of the objective function. Then we propose a new genetic algorithm where fitness is calculated by both expectation and variance of the objective function. Through numerical experiment, we have shown the effectiveness of the proposed method compared with the conventional methods. Moreover, we have shown application of the PVM (Parallel Virtual Machine) can improve its solution efficiency.
