Reactor Temperature Control Based On Parameter Optimization MPC Algorithm

With the improvement of productivity,the requirements for control in the chemical industry are also increasing.Continuous stirred tank reactors are extremely important devices in the chemical industry,and the temperature control of reactors is a key and difficult point in the industry.The temperature control of reactors determines production efficiency and safety assurance,and ordinary controllers are difficult to meet the temperature control of complex working conditions of reactors.This article proposes a GLMPC-PID(Grey Wolf Optimizer-Long Short Term Memory-Model Predictive Control-Proportional Integral Differential)cascade intelligent control system to address the aforementioned shortcomings.The GLMPC-PID cascade intelligent control system is an intelligent cascade control system formed by GWOLSTM-MPC(Grey Wolf Optimizer-Long Short Term Memory-Model Predictive Control)as the outer loop controller,intelligent interference controller as the inner loop controller,three-point temperature detection module as the temperature acquisition module,and reaction kettle grey box model module.GWO-LSTM-MPC uses the gray wolf optimization algorithm to optimize the number of neurons,learning rate and L2 regularization parameters of the long-term and short-term memory neural network,and proposes to use the GWO-LSTM(Grey Wolf Optimizer-Long Short Term Memory)model to predict the prediction time domain,control time domain and weight factors of the model predictive controller online,forming the GWO-LSTM-MPC controller;The reaction kettle grey box model module is composed of a mechanism model of the reaction kettle and a BP(Back-propagation)neural network compensation module,thus proposing a reaction kettle grey box model based on BP neural network compensation;The intelligent interference controller module proposed in the article is a combination of semi closed-loop feedforward compensation,Proportional Integrated Differential(PID)controller in the secondary circuit,and fuzzy rule inference.Through fuzzy rule inference,the parameters in the secondary circuit are adjusted online to suppress interference;The three-point temperature detection module proposed in the article uses three temperature detectors to monitor the feed inlet,middle and bottom of the reactor with significant temperature changes in real-time,improving the accuracy of temperature collection in the reactor.Through simulation analysis using Simulink and DCS(Distributed Control System)software,the results show that the designed GLMPC-PID cascade intelligent control system has strong anti-interference ability and the ability to quickly reach the set value.