Prediction Controller Design For Coke Oven Blowing Cooler System Based On Multi-kernel SVR

Coke as a basic industry of the country,its quality and production capacity has an important impact on the development of the country.The technological process of this industry is complex,with many processing links and mutual interference among each link.In the coking process,the cold drum system plays an important role in the coking process.Its stability is one of the most important indexes in the coking process.Therefore,the research of scientific control algorithm is of great significance to the development of coke industry and even industry.As is known to all,the coke oven blowing cooler system is a nonlinear system and it is difficult to build accurate models.After reading a large number of articles and observing the on-site production process,the Support Vector Regression(SVR)machine was used to model it.It can predict the key indexes of the system.In addition,considering the variety of kernel functions,data processing capacity has its advantages and disadvantages.Therefore,different kernel functions are combined in this paper.This paper adopts multi-kernel support vector regression model.Under the combination of different weights,this paper selects the optimal combination.Meanwhile,there are many parameters in the multi-kernel model.To further optimize the multi-kernel model,Adaptive Weight Particle Swarm Optimization(APSO)was added to perform iterative optimization for each parameter of the model.Finally,the prediction model of cold drum system based on APSO-Mk-SVR is obtained.The optimized multi-kernel SVR model is applied to the prediction and control scheme of the system.According to the different production conditions of the coking plant,the data collected from the industrial site,of which 70% is the training data and30% is the test data,is used to establish the prediction controller of the system.MATLAB software was used to carry out relevant simulation experiments.The results show that compared with the traditional PID control,the predictive control model proposed in this paper can better maintain the stability of the suction before the primary cooler under different working conditions.When the system faces the interference,the prediction value will have a small amplitude fluctuation,but after a short time of adjustment,it can be stabilized quickly.The system has strong anti-interference,and it can better meet the actual coking process requirements.