【摘要】 为保证复杂化学相平衡问题准确求解,本文提出了一种基于Gibbs自由能最小原理的遗传与梯度混合算法(GAG)。为加快混合算法的计算速度和确保有效收敛,在计算过程中引入了变化参数,即算法中的遗传变异算子不再为常数而是根据平均适应度的大小进行动态调整。在进化的过程中,记录所有个体的近代适应度,适应度高的个体进入下一代。一旦发现满足设定条件的最优个体,本算法将沿着该个体的负梯度方向搜索快速逼近最优解。多个复杂气—固和气—液体系计算实例表明,该算法可以不断拓展求解空间,能以完全概率、高速并行搜索到全局最优解,并有效避免假收敛。该算法的最佳计算初始值区域为[-3,30],该混合算法即使初值点非常不好的情况下也可以确保收敛到最优值。为保证结果的准确性,所选计算实例的运行次数均不低于30次。更多还原

【Abstract】 To obtain the accurate solution of complex nonlinear problems,a hybrid genetic algorithm and gradient algorithm(GAG) was proposed according to the minimization of the Gibbs free energy for solving simultaneous chemical and phase equilibrium.The converted variable was introduced in order to improve the compute speed by the GAG algorithm to assure convergence.The mutation rate(Pm) is variable,so it will be adjusted dynamically by the size of average fitness value.When it produces the next generation,all the individuals in the current generation are reordered on fitness value.Then the individual of high fitness value go into the next generation.Meanwhile,when finding the best individual that satisfies the converted condition,it will converge quickly by the minus gradient direction.The computed examples of the vapor-solid and vapor-liquid systems showed that the GAG algorithm was a very efficient searching method on large initial value region and more rapid than typical algorithm. There is a minimum computation time point when the initial value lies in[-3,30].This approach can guarantee finding a global solution,even starting from a poor initial value.To get a convergent result,the running times of the procedure should not be less than 30 times in our examples.更多还原