论文部分内容阅读
粒子群算法(PSO)是一种基于迭代的智能算法,具有较好的全局搜索能力,但局部搜索能力较弱。针对粒子群算法容易陷入局部最优不足这一问题,本文提出了一种精英克隆选择的方法。该算法在基本粒子群算法的基础上保留了种群中的若干个精英粒子,然后将这些精英粒子进行克隆复制,并将复制之后的粒子进行变异操作,再将变异之后的粒子与变异前的粒子进行亲和度的比较,保留下来亲和度提高的粒子并替换之前适应值较差的粒子,通过这种方式增强了种群的多样性,从而避免了粒子陷入局部最优的问题。此外,本文引入了新的改进惯性权重的机制,根据粒子位置和速度的情况自适应地改变惯性权重,这样避免了粒子盲目运动,更有针对性的寻找最优解。对4个高维复杂函数寻优测试,分别从平均精度和标准差这两方面进行分析,结果表明改进之后的算法在寻优精度和稳定性方面都超过了基本PSO,从仿真图像中可以看出改进之后的算法在迭代末期跳出了局部最优更接近全局最优值。最后将这种改进的算法应用到优化乙烯、丙烯的收率模型中,应用结果表明当裂解原料属性发生改变时,本文提出的算法可以很快完成对操作变量的寻优,显著提高了“双烯”收率。
Particle swarm optimization (PSO) is an iterative intelligence algorithm, which has good global search ability but weak local search ability. Aiming at the problem of particle swarm optimization being trapped in local optimum, this paper presents a method of elitist cloning selection. The algorithm preserves a number of elite particles in the population based on the basic particle swarm optimization algorithm. Then, these elite particles are cloned and replicated, and then the replicated particles are mutated. Then, the mutated particles and the pre-mutation particles The comparison of affinity, retention of particles with increased affinity, and replacement of previously poorly-adapted particles enhances population diversity in this way, avoiding the problem of particles falling into local optima. In addition, this paper introduces a new mechanism to improve the inertia weight, which adaptively changes the inertia weight according to the particle position and velocity, thus avoiding the blind movement of particles and finding the optimal solution more specifically. The optimization test of four high-dimensional complex functions is performed respectively from the average precision and the standard deviation. The results show that the improved algorithm outperforms the basic PSO both in accuracy and stability, and can be seen from the simulation images After the improvement, the algorithm jump out of the local optimum at the end of the iteration closer to the global optimal value. Finally, the improved algorithm is applied to optimize the yield models of ethylene and propylene. The application results show that the algorithm proposed in this paper can quickly optimize the manipulated variables and significantly improve the quality of the raw materials when the properties of the cracked raw materials change. Dienes "Yield.