基于改进动态神经网络的精馏塔温度控制

Temperature Control of Distillation Tower Based on Improved Dynamic Neural Network

精馏过程是一种具有多变量、强耦合、时变非线性特性的复杂工业生产过程。为实现对其温度的精确控制,首先通过对高阶耦合系统分析提出一种改进型动态神经网络;然后利用改进的动态神经网络建立精馏塔的模型并设计解耦器对精馏塔进行解耦;最后针对解耦后的每个SISO系统分别采用参数可在线调整的PID控制器进行控制。由于其时变特性,同时为避免反传算法和其他智能优化算法收敛慢、易陷入局部极值的缺点,对粒子群算法(PSO)进行改进,用改进的粒子群算法(PSO)训练神经网络并优化PID控制器的参数。经过仿真实验验证了所提出控制策略的有效性和可行性。

Distillation is a multivariable, strong coupling, time-varying and nonlinear industrial production process. In order to achieve precise control of its temperature, firstly, through the analysis an improved dynamic neural network is provided. Secondly, the improved dynamic neural network is used to identify the modol of the distillation tower and the decoupler is designed. Finally, a PID controller is adopted which can be adjusted online to control eatch SISO channel. Due to its timevarying characteristics, and it avoids the disadvantage that converges slowly and easy to fall into local minimum of the basic PSO algorithm, an improved PSO algorithms is proposed and uses for training the neural network and optimizing the parameters of the PID controller. The simulation shows the effectiveness and feasibility of the proposed control strategy.