Application Of Improved Sparrow Search Algorithm In Gasoline Octane Number Loss

At present,our society is progressing,economy is developing,people’s living standards have been improved significantly,family car has become a popular product,its production and sales are also increasing year by year.We are now at a stage of effective progress in epidemic control and gradual liberalization.People’s desire to travel will lead to a greater spike in car production and sales.The main source of automobile power is gasoline,and the important index of gasoline quality is octane number(RON),determines the price of gasoline,which directly affects the economic benefits of chemical enterprises,each unit of gasoline octane loss,about 150 yuan per ton of gasoline loss,so it is very important to control the octane loss in the process of gasoline catalytic cracking.In this paper,data mining technology is used to explore the operational factors affecting the octane number of gasoline,and a reliable prediction model of octane number loss is established to provide reference and ideas for the formulation of gasoline refining process strategy.Firstly,according to the defects of the traditional sparrow search algorithm,some strategies are adopted to improve the algorithm,including the initial position of the population adjusted by Tent chaotic mapping.Adding an adaptive weight factor to change the convergence step size dynamically makes the step size longer in the early iteration and shorter in the late iteration.And the integration of differential evolution algorithm and other strategies,so as to overcome the slow convergence speed,easy to fall into local extreme value and low stability.Through the standard test function of 10 different scenarios,the algorithm is evaluated,compared with genetic algorithm(GA),Grey Wolf algorithm(GWO)and Sparrow algorithm(SSA).The final experiment proves that the improved sparrow algorithm has faster convergence speed,higher accuracy and stronger global search ability.Secondly,the LSTM parameter value is analyzed,and the octane loss prediction model is built on the basis of the improved Sparrow algorithm.Firstly,a series of data preprocessing is carried out for a large number of redundant operating variables and the incomplete,difference and bad value of the data set.After processing the data,the data is normalized,and the feature extraction is carried out by low variance filtering method,distance correlation coefficient,etc.,to obtain multiple variables with the highest correlation coefficient,which greatly reduces the dimension of the data and avoids the problem that data redundancy is difficult to train.Finally,GA,GWO,SSA and multi-strategy improved SSA were used to construct LSTM gasoline octane loss prediction models,and the experimental comparison was made.Finally,according to the analysis of the experimental results of the above prediction model,the mean square error,mean absolute error,root mean square deviation,mean absolute proportion and other evaluation indexes of the model are significantly reduced compared with other models,which verifies the improvement of the model quality,indicating that the improved ISSA has good stability and optimization ability.