基于多策略扰动机制的飞蛾火焰算法及在乙烯聚合过程动力学参数估计中的应用

Moth-Flame Optimization Algorithm Based on Multi-Strategy Disturbance and its Application in Kinetic Parameter Estimation of Ethylene Polymerization

针对飞蛾火焰算法(MFO)早熟和易陷入局部最优的缺点,研究提出了一种基于多策略扰动机制的飞蛾火焰算法(MSMFO)。该算法通过莱维飞行和布朗运动两种机制的随机选择对飞蛾粒子进行重新定位,以克服种群后期多样性严重下降的缺陷。通过对8个测试函数的仿真对比,MSMFO具有更高的收敛精度和更好的全局收敛能力。将MSMFO算法用于工业乙烯聚合过程动力学参数估计。以聚合反应动力学参数为决策变量,模型输出的数均分子量和重均分子量与实际值之间的相对误差平方为目标。结果证明,MSMFO算法具有更好的收敛性。

A moth flame algorithm method based on multi-strategy disturbance（MSMFO） was proposed to overcome shortcomings of maturation and local convergence of typical moth flame algorithm（MFO）. This algorithm selects new positions of the moth particles by random selection of Levy flight and Brownian movement, which can overcome the decline of population diversity in late periods. As proved by simulation of eight test functions, MSMFO has higher convergence accuracy and better global convergence. Moreover, MSMFO was used in polyethylene production to minimize the square sum of relative error between the number-average molecular weight and the weight-average molecular weight with kinetic parameters as decision variables. The simulation results show that MSMFO has better convergence when comparing with ELPSO and DE algorithms.