基于随机森林-粒子群复合算法的精馏过程辨识

Identification of distillation processes based on combined random forest-particle swarm optimal algorithm

为了应对化工行业日益增长的性能需求,基于模型的控制策略已经被广泛应用于化工过程的控制与优化。精馏过程的高度非线性特点,使得其模型辨识一直是过程工业中的一个难点。文中基于精馏过程的数据特性,提出了具有良好识别度的随机森林-粒子群算法,通过将粒子群算法中的适应度函数定义为基于随机森林算法的均方误差以确定其最佳参数。最后通过Simulink-Aspen互访平台收集的数据样本进行仿真实验,验证了所提出识别方法的有效性。

In response to the increasing performance demands of the chemical industry,model-based control strategies have been widely used in the control and optimization of chemical processes.The highly non-linear nature of the distillation process causes model identification being a difficult task in the process industry.Based on the data characteristics of the distillation process,random forest-particle swarm algorithm(RF-PSO)with good identification was proposed by defining the fitness function in the particle swarm algorithm as the mean square error based on the random forest algorithm to determine its optimal parameters,and the effectiveness of the proposed identification method is verified through simulation experiments with data samples collected by the Simulink-Aspen inter-access platform.