Research On The Data-driven Modeling And Optimization Of A Catalytic Cracking Reaction Regeneration Process

The establishment of a model for a catalytic cracking reaction regeneration system is a key issue in achieving optimization of the operating conditions of the production process and increasing the product yield.At present,various production data in the petrochemical industry production process can be collected from the device's database platform in real time,making the establishment of a intelligent reaction regeneration system model based on data-driven become one of the current research hot issues.To solve this problem,the BP neural network(BPNN)and Support Vector Regression(SVR)models for gasoline production rate reaction and regeneration process are established based on historical production data of the catalytic cracking unit.The experimental verification and comparative analysis of the model show that the proposed method is effective and feasible.The main research work of this article is as follows:Firstly,the catalytic cracking reaction regeneration process was studied in this paper.The input and output variables of the catalytic cracking reaction regeneration model were selected,and the gasoline yield was used as the output variable and the first reaction region temperature of the rise?the second reaction region temperature of the riser?catalyst oil ratio?weight space velocity and reaction pressure,these operating conditions as input variables.The principle of artificial bee colony algorithm(ABC),particle swarm optimization(PSO),ant colony optimization(ACO),and genetic algorithm(GA)were analyzed.These algorithms were used to optimize the model parameter.Secondly,a model for the catalytic cracking reaction regeneration system of BPNN is established,with the gasoline yield as the output variable and the operating conditions as input variables.Using ABC,ACO,PSO and GA to optimize the initial weights and thresholds of BP-NN,the results show that the accuracy of the optimized BP-NN model is greatly improved.Finally,the model of catalytic cracking reaction regeneration system was established using SVR,and SVR parameters are optimized using ABC,ACO,PSO,and GA in sequence.Comparing with BP-NN,the prediction error of SVR model is obviously smaller than that of BP-NN model,and the accuracy is higher.On the model of establishing a catalytic cracking reaction regeneration system,SVR is more advantageous for small sample data modeling.The support vector regression model(ABC-SVR)based on artificial bee colony algorithm optimization has the highest accuracy and best performance.After hypothesis testing,the reliability of the ABCSVR model was verified.ABC-SVR was chosen to predict the maximum gasoline yield.In the prediction process,an artificial intervention based on the actual production process is added,the corresponding process constraints are converted into model constraints,and the maximum gasoline yield is predicted under the constraint conditions.This ensures that the model prediction results are consistent with actual production conditions.The predicted maximum gasoline yield is greater than the actual production of the maximum gasoline yield,illustrating the effectiveness and practicality of the model,the operating conditions improve the gasoline yield can also be increased,and provide guidance for the actual production.