Abstract:

The algorithm of non-dominated sorting genetic algorithm II (NSGA2)and its parameter problems were discussed and applied for multiobjective optimization of a reservoir. The Pareto front of the optimization problem was obtained and the effects of the corresponding parameters on the optimal result were discussed. The study showed that the parameters in genetic algorithms in NSGA2 (include size of tournament selection, distribution parameter in simulated binary crossover, and distribution parameter in polynomial mutation) have little effects on the optimal result, and this means that NSGA2 is robust and a set of proposed values of the parameters can be used for most multiobjective optimization problems of a  reservoir.  A big enough and even enough distributed Pareto front can be obtained when the two parameters,population size and generation in NSGA2 are big enough, and this means that NSGA2 is simple for its parameter regulation. The Pareto front obtained gradually improves  with the regulation of population size and generation, and this means that NSGA2 is stable for multi-objective optimization problems of a reservoir.

Key words:  reservoir, multiobjective programming, optimal operation, non-dominated sorting gentic algorithm Ⅱ, Pareto front