| Drying is a process with the largest energy consumption and complex mechanism in papermaking process.Reducing its energy consumption is an urgent problem for enterprises to promote energy conservation and emission reduction.The existing research on drying process simulation and energy efficiency optimization research is mainly based on the mechanism model,and the modeling process needs an accurate theoretical basis,which is suitable for simple and specific process objects.For complex process objects,the simulation accuracy cannot meet the requirements of the production process,and the effect of the optimization model established on this basis is naturally poor.In this paper,a hybrid modeling method based on mechanism and data-driven is proposed to simulate the key parameters of paper drying process.Based on the process simulation model,a paper drying process energy efficiency optimization model is established.The data acquisition scheme is determined by analyzing the basic structure of drying section.Because industrial data is greatly affected by manual operation,equipment status and communication environment,data mismatch,abnormal and noise problems often occur.Therefore,this paper carries out data matching,outlier elimination,data filling and data filtering operations on the data,and the processed data is used for subsequent modeling.Based on the kinetic theory of paper drying,this paper divides the mechanism model into three parts: air hood model,heat and mass transfer model and paper drying model.Three machine learning methods,ridge regression,back propagation neural network and support vector regression,are used to build data-driven models to predict the simulation error of the mechanism model,and then the hybrid model is formed by combining the model with the mechanism model.This paper also uses the above three machine learning methods to establish the corresponding pure data-driven model,and uses two cases with large difference in working conditions to verify the performance of the three models.The results show that the simulation accuracy of both data-driven models and the hybrid models is better than that of the mechanism model in Case 1,and the maximum mean relative error is only 2.36%,and there is little difference in the simulation accuracy between the two models.In case 2,the performance of data-driven models decreased significantly,while the hybrid models still maintained good performance with the maximum average relative error of 4.84%.On the basis of process simulation model,this paper established energy efficiency optimization models of paper drying process based on genetic algorithm and sequential least square programming,and the performance of the two models is verified by using the industrial data.The results show that the energy efficiency optimization model based on the two different optimization algorithms can bring the cost reduction effect of 9.03 CNY/t and 9.24 CNY/t,respectively.In terms of time cost,the average time of genetic algorithm to optimize a single sample is 5237.42 s,while the sequential least square programming method is only 62.38 s.Compared with the paper quality indexes before and after optimization,it is found that the energy efficiency optimization model has no significant effect on the paper quality.Based on the data operation platform established by the enterprise,the real-time working condition data acquisition module,logic judgment module,crawler module and data preprocessing module are successively developed in this paper.After each module is encapsulated,a human-computer interface is designed to realize the online deployment and industrial application of the process simulation model and energy efficiency optimization model. |
