Research On Output Forecasting Method In Ethylene Production Process And Its Application

In 2022,China put forward a policy to vigorously promote economic development to enhance the well-being of the people.As a fundamental industry for economic construction and people’s livelihood,the petrochemical industry plays a particularly crucial role in promoting economic development.Ethylene,as an indispensable part of petrochemical products,is one of the important indicators for measuring a country’s level of development in the petrochemical industry.Improving ethylene production efficiency is one of the significant ways for China to achieve rapid economic growth and occupies an essential position in the national economy.Therefore,to accelerate national economic development,how to accurately predict ethylene production and improve ethylene production efficiency is the main problem facing my country’s petrochemical industry.Based on the input-output data in the complex ethylene production process of the domestic petrochemical industry,this paper analyzes the basic theory of ethylene forecasting and studies the output forecasting method of ethylene production process,and is committed to providing a certain theoretical basis and reality for petrochemical enterprises to improve ethylene production efficiency significance.The main work of this paper is as follows:(1)Aiming at the problems of high dimensionality of production data and low precision of production prediction model caused by factors such as raw materials,water and electricity consumption,and parameters of ethylene production equipment in the ethylene production process,a method based on Spearman correlation analysis combined with recursive feature elimination(RFE)and LightGBM feature selection method(SRL)was proposed.The Spearman correlation analysis method performs correlation analysis on ethylene data to obtain the correlation of feature vectors,and then uses RFE and LightGBM methods to calculate the importance of ethylene data to obtain the importance of feature vectors,count the feature vectors frequency of correlation and importance and sort them,and obtain the key features in high-dimensional data after comprehensive calculation as the result of data feature selection.The SRL method effectively avoids the problem that the single feature selection method may miss important features or fail to screen out redundant features due to the influence of noise data,realizes the dimensionality reduction operation of ethylene data in complex industrial production,and provides data support for subsequent research on ethylene production prediction.(2)In order to accurately predict the ethylene production in the complex ethylene production process,an ethylene production prediction method based on the Snake Optimizer Algorithm(SOA)combined with the Long Short-Term Memory(LSTM)was proposed to guide ethylene production to reduce costs,increase efficiency,and reduce carbon emissions.In order to reduce the impact of LSTM prediction model hyperparameter selection on prediction accuracy,SOA is used to find the optimal hyperparameters such as the number of hidden layer nodes,batch training data block size and training times in LSTM,and the root mean square error(RMSE)is used as the optimization goal.SOA adaptively adjusts the moving step size according to the current search state and the target state of the solution vector to obtain the global optimal solution of LSTM hyperparameters;then combined with the ethylene data processed by the SRL feature selection method for dimensionality reduction,the output of ethylene forecasting model is established based on the proposed improved LSTM method.Compared with the back propagation neural network,extreme learning machine and LSTM methods,the prediction accuracy of the proposed method has increased by36.09%,49.26% and 33.41%,respectively,it effectively guides the improvement of ethylene production efficiency and the reduction of carbon dioxide emissions.(3)Based on the research on ethylene production process data dimension reduction processing and SOA-based optimization LSTM prediction method,using Java and Python as programming languages,using Vue,Spring Boot and Flask frameworks,using MySQL database and Redis cache database,designed and developed a B/S framework ethylene production forecasting prototype system.The system has designed and developed four functional modules of user information management,experimental data management,predictive model management and predictive analysis to realize multi-dimensional display of data,ethylene production forecast,comparison and analysis of forecast results,etc.,providing a petrochemical enterprise staff A production management platform with convenient operation and complete functions.The system has designed and developed four functional modules of user information management,experimental data management,predictive model management and predictive analysis to realize multi-dimensional display of data,ethylene production forecast,comparison and analysis of forecast results,etc.It provides a production management platform with convenient operation and complete functions for the staff of petrochemical enterprises.