Research On Data Driven Optimization And Intelligent System Of Gasoline Blending Formula

As an indispensable production process of various gasoline products,gasoline blending plays an important role in petroleum refining and chemical production,and the quality of oil products has an important impact on the economic benefits of enterprises.In the production process of refined gasoline,the establishment of an effective formula model to strictly control the addition of component oil is the basis to ensure the quality of refined oil and the improvement of enterprise efficiency.Therefore,driven by the demand of trimming production in the blending process of refined gasoline,this thesis carried out a systematic research on the modeling of blending formula.The main work and achievements are as following:1)In order to solve the problems of excessive quality and low success rate of primary blending in oil production,a hybrid modeling scheme is proposed through indepth analysis of process mechanism and production data to ensure that the product oil is close to the edge sticking production.In other words,the historical production process data are used to establish the edge and conservative formula models,and then the mixed formula model is established according to the weighted fusion strategy.The establishment of conservative formula model aims to ensure the normal production of oil products,while the establishment of edge formula model aims to promote the edge production of oil products.In view of the fact that the quality of the model may not meet the standard due to the uncertainty of the production process,the optimization and integration of the two models are carried out,so as to improve the success rate of primary blending and approach the edge production of oil products as much as possible.2)In order to solve the problem of data completeness in formula modeling,a serial hybrid particle swarm optimization genetic algorithm is proposed to solve the problems of low accuracy and easy to fall into local minimum in the process of octane number and antiknock value prediction of BP algorithm,which combines the advantages of PSO and GA,and is used to optimize the weights and thresholds of BP neural network to improve the accuracy of octane number and antiknock value prediction model.Secondly,in order to fill in the missing recipe data,a recipe data filling method based on integrated decision tree is proposed,the original data set is divided into multiple sample sets,and then multiple sub models are established by using decision tree algorithm.Finally,multiple sub models are integrated based on bagging strategy to improve the accuracy of the model.3)In order to improve the accuracy of the conservative formula model,considering the advantages of deep belief network(DBN)in feature extraction and nonlinear processing,a conservative formula modeling method based on PSO-DBN is proposed,which is applied to the prediction modeling of gasoline blending formula.Meanwhile,aiming at the difficulty of parameter selection in the process of DBN training,particle swarm optimization(PSO)algorithm is used for the related parameters were optimized.Secondly,considering the good fitting ability and rapidity of the extreme learning machine,based on the above model,the DBN-ELM prediction model is constructed by improving the DBN network,which realizes the prediction modeling of refined gasoline blending and conservative formula.The simulation results show that the algorithm has greatly improved the training time and accuracy.4)In view of the small sample characteristics of the edge data,LSSVM is used to establish the edge formula,and PSO algorithm is used to optimize the parameters of RBF kernel function to improve the performance of the model.On this basis,the weighted fusion strategy is used to integrate and optimize the results of the edge and conservative formula models.The results show that the hybrid formula model has better performance for high-efficiency production.5)In order to apply the formula model to engineering,combined with the actual engineering needs,according to the above mixed modeling scheme,with the help of My SQL,MATLAB and Lab VIEW technology and software platform,the system database was designed,and the modeling methods were realized based on MATLAB.Finally,a formula intelligent management system with system login,octane number,antiknock index prediction,conservative formula prediction,edge formula prediction,mixed formula model and model evaluation,online optimization,data viewing,input,alarm and event recording function modules was developed.