基于神经网络的碱回收蒸发工段黑液液位及浓度控制系统设计

Design of Black Liquor Level and Concentration Control System in Alkali Recovery Evaporation Section Based on Neural Network

关键词:  碱回收  黑液液位  黑液浓度  神经网络

Key Words:alkali recovery  black liquor level  black liquor concentration  neural network

摘要:针对造纸黑液碱回收蒸发工段多变量、大时滞、强耦合、难控制的特点，设计了以PIDNN为控制器的黑液液位控制，将传统PID和神经网络的优点巧妙结合，不仅结构简单，而且具有自适应学习能力；黑液浓度设计成基于径向基函数（RBF）的神经网络控制，对黑液浓度和流量进行解耦在线辨识，对控制器参数在线实时调节。实践表明，黑液液位控制与传统PID控制相比，PIDNN的调节时间减少约18 s，超调量降低约20%；与BP-PID相比，PIDNN的调节时间降低约14 s。黑液浓度的实际值可以快速跟随给定信号，有效黑液流量的变化对其干扰很小。

Abstract:Aiming at the characteristics of multi-variable， large time lag， strong coupling， and difficult control in the evaporation section of papermaking black liquor alkali recovery， the black liquor level control with PIDNN as the controller was designed. It could combine the advantages of traditional PID and neural network， not only had a simple structure， but also had adaptive learning capabilities. The black liquor concentration was designed as a neural network control based on the radial basis function （RBF）， and the concentration and flow of the black liquor were decoupled online identification， and the controller parameters were adjusted online and in real time. Practice showed that compared with traditional PID control， the adjustment time of PIDNN was reduced by about 18 s， and overshoot was reduced by about 20%； compared with BP-PID， the adjustment time of PIDNN was reduced by about 14 s. The actual value of the black liquor concentration could quickly follow the given signal， and the change of the effective black liquor flow rate had little interference to it.