论文部分内容阅读
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。
Chemical optimization problems are often more complex, the traditional deterministic optimization method easily fall into the local optimum. Particle collision algorithm (PCA) is a newly proposed stochastic global optimization algorithm, which simulates the absorption and scattering phenomena of collisions between particles and nuclei in nuclear reactions. It is designed to optimize the algorithm with three operation operators: disturbance, detection and scattering. However, the global optimization is not efficient. By analyzing the mechanism of PCA optimization, some improvement strategies are proposed, including the design of cross-operation of multi-bit crossover operator increase algorithm to overcome the weakness of PCA lack of co-evolution mechanism; the use of simplex search to improve the search operator to enhance the local optimization ability; Using adaptive crossover rate adjustment, an improved Particle Collision Algorithm (MPCA) is designed. Shaffer’s F6 function and eight-dimensional Alpine function tests show that the global optimization performance of MPCA is obviously better than that of PCA and conventional genetic algorithm (SGA). MPCA was applied to L-isoleucine batch fermentation kinetic model parameters optimization with satisfactory results.