改进的粒子碰撞算法及其在化工优化中的应用  

Modified particle collision algorithm and its application to optimization in chemical engineering

作　　者：郑启富[1,2] 苏国栋[1] 陶晓红[1] 

机构地区：[1]浙江工业大学浙西分校化工系,浙江衢州324000 [2]浙江工业大学催化研究所,浙江杭州310014

出　　处：《计算机与应用化学》2008年第10期1229-1232,共4页Computers and Applied Chemistry

基　　金：浙江省自然科学基金(Y407266);浙江省工业催化重中之重学科开放基金(200602)

摘　　要：化工优化问题往往较为复杂,传统的确定性优化方法容易陷入局部最优。粒子碰撞算法(PCA)是新近提出的一种随机全局优化算法,是模拟核反应时粒子与原子核碰撞发生的吸收和散射现象,设计成以扰动、探测、散射三种操作算子实现算法寻优,但全局寻优效率不高。通过分析PCA寻优机制,提出改进策略,包括设计多位交叉算子增加算法的交叉操作,以克服PCA缺乏协同进化机制的弱点;运用单纯形搜索改进探测算子,以增强局部寻优能力;采用交叉率自适应调整等,由此设计一种改进的粒子碰撞算法(MPCA)。Shaffer's F6函数和八维Alpine函数测试表明,MPCA的全局优化性能明显优于PCA和常规遗传算法(SGA)。将MPCA应用于L-异亮氨酸分批发酵动力学模型参数优化,结果满意。Optimization problems in chemical engineering are usually complicated, conventional deterministic optimization methods often result in local optimization. Particle collision algorithm （PCA） is a novel stochastic global optimization method proposed newly. It is loosely inspired by the physics of nuclear collision reactions, particularly scattering and absorption. Three important operators of PCA, namely perturbation, exploration and scattering were designed through simulating these phenomena of nuclear collision. However, the global optimization efficiency of PCA isn＇t satisfied because of its some shortcoming. In order to improving the optimization efficiency, in this paper, the mechanism of searching global optimization of PCA was analyzed, then the strategies of modifying PCA were proposed, and a modified particle collision algorithms （MPCA） was designed subsequently. The technologies for improving optimization performance of MPCA were composed of multipoint crossover arithmetic operator being designed to crossover operation, simplex method being as explora- tion arithmetic operator, crossover probability being adjusted felicitously, and so on. Shaffer＇s F6 function and eight-dimensional Alpine function was applied to testing MPCA, the results demonstrated that its global optimization performance was superior to those of PCA and simple genetic algorithms （SGA）. Further, MPCA was applied to optimize the kinetic models parameters of L-isoleucine batch fermentation, and satisfactory results were obtained. These results by MPCA were better than those by SGA in the reference.

关 键 词：随机优化 粒子碰撞算法 改进策略 参数优化 动力学模型 L-异亮氨酸发酵 

分 类 号：TQ021.8[化学工程]