求解非线性规划问题的遗传算法设计与实现.doc

摘 要 非线性规划在工程、管理、经济、科研、军事等方面都有广泛的应用。传统的解决非线性规划问题的方法，如梯度法、罚函数法、拉格朗日乘子法 遗传算法是模拟达尔文的遗传选择和自然淘汰的生物进化过程的计算模型对目标函数既不要求连续，也不要求可，适用于并行分布处理，应用范围广。 关键词：非线性规划；遗传算法；罚函数法 ABSTRACT Non-linear programming has a wide range of applications in engineering, management, economic, scientific, and military aspects. Traditional methods to solve the non-linear programming problem, such as the gradient method, penalty method, Lagrange multiplier method, have poor stability. They are sensitive to the function initial value and request the objective function to be continuous and differential. The results are also easily trapped into local optimal solution. Genetic algorithm is a kind of calculate model which simulates Darwins genetic selection and biological evolution of natural selection. Genetic algorithm is a global search algorithm. It has simple, universal, robust features, and does not request the objective function to be continuous and differential, and is suitable in parallel distribution processing. Genetic algorithm is widely applied in many areas. Based on the analysis of the disadvantage of traditional non-linear programming algorithm and the advantage of genetic algorithm, genetic algorithm is applied to non-linear programming in this paper. The introduction of the concept of penalty function is used to construct the fitness function with punishment. By using real-coded, Roulette Wheel selection method, two-point crossover, uniform mutation, we formed a genetic algorithm to solve the non-linear programming problem. Compared with the most classical and widely used traditional non-linear programming problem algorithm –SUMT algorithm, the results show that the new algorithm could effectively overcome the defect of the traditional algorithm in a certain extent. The new algorithm is more stable, less sensitive to the function initial value and conditions, and always could receive the optimal solution or approximate optimal solution. Its convergence results are more reasonable, the per