自适应变异差分进化算法估计软测量参数

提出一种自适应变异差分进化算法(ADE),能根据搜索进展情况自适应地确定变异率,使算法在初期保持个体的多样性,避免早熟:在后期逐步降低变异率,保留优良信息,避免最优解遭到破坏,增加搜索到全局最优值的概率．与传统的差分进化算法(DE)相比较,ADE算法的离线性能和在线性能都有较大的改进,搜索到全局最优解的概率获得较大提高,对算法参数的敏感性低．本文将ADE算法应用于对苯二甲酸中对羧基苯甲醛含量软测量模型的参数估计,获得了满意的结果．

Adaptive mutation differential evolution algorithm and its application to estimate soft sensor parameters

A novel adaptive mutation differential evolution (ADE) algorithm containing the adaptive mutation operator, in which the mutation probability is determined according to the evolved generations, is proposed in this paper. The adaptive mutation operator makes the individuals diversity in the population at the initial generations to overcome the premature, and reduces the mutation probability gradually during the evolutionary process to preserve the excellent individuals and enhance the probability of obtained the global optimal solution. To compare the performances of ADE with those of the traditional differential evolution (DE), ADE and DE are applied to search the global optimal solution of analytical function. The results demonstrate that both ADE's on-line and off-line performances are superior to those of DE, the probability of obtained the global optimal solution is larger than that of DE, and that the parameter sensitivity degree of ADE is lower than that of DE. Furthermore, ADE is applied to estimate the model parameters of 4-carboxybenzaldehyde (4-CBA) soft sensor, and satisfactory results are obtained.