Modeling And Optimal Control Strategy For Yields Of Ethylene Cracking Furnace

The ethylene industry is the pillar of petroleum chemical industry, the production level of ethylene industry represents the level of the petroleum chemical development in a district or a country. Ethylene cracker is one of the core devices which have high production capacity and maximum consumption in petrochemical industry and is the major source of raw material, like ethylene and propylene and so on. Accurately predicting and optimal controlling the yields of ethylene cracking furnace has great significance in saving energy and improving the economic efficiency. Ethylene cracking furnace is the complex nonlinear system with lots of related variables, therefore it is difficult to build mechanism model. Industrial chromatograph has been used for measuring the yields of ethylene cracking furnace on line in the ethylene industry, but the problems of huge instrument investment, high operating costs and the measurement lag are neglected, therefore it is urgent to build model predicting the yields of ethylene cracking with the off-line or on-line data. However, the model usually contains one single operating condition, making the predicting precision insufficient. Besides, how to optimize the operation and realize the maximizing production is the key of the enterprise benefits. Based on the analysis of working principle for ethylene cracking furnace and combined the real operating condition of ethylene industry and advanced modeling and optimizing control theory, this article has researched on soft-sensing modeling and optimizing control strategy for ethylene cracking furnace. The main research work is shown as follows:(1) Review the subject research background, the overseas and domestic research status, the modeling and optimizing control of ethylene cracking furnace research status. Illuminate the research significance of this article. Briefly introduce the types of cracking furnace, the technological process and the operating principle of cracking furnace and the influence factors for the yields of ethylene cracking furnace.(2) Aiming at the inaccurate problem of single operating condition and single model, the model based on operating classification is proposed. The ethylene production process can be divided into three operating conditions based on the raw material characteristics, namely whether the raw material contains the first/top vacuum side stream, light hydrocarbon and hydrogenated C5. The feed rate, dilution steam flow, fuel gas flow, steam/hydrocarbon are chosen as operating condition classification variable. In the process of data processing, considering the workloads and the adequacy of data, transfer learning is adopted to improve the operating classification precision. After the classification, a typical operating condition is selected to research. Firstly, this article adopted the principal component analysis (PCA) to filtrate the variables. Then the data is used to modeling with the input variables like the feed rate, dilution steam flow, fuel gas flow, coil outlet pressure, coil outlet temperature and the output variable ethylene yields. The least squares support vector machine (LS-SVM) optimized by particle swarm optimization (PSO) is adopted to build the model. The data adopted in the industry scene is to verify the precision of the model, and the results show that the model is so efficient to predict the ethylene yields.(3) The optimization target is the sum of ethylene and propylene, and genetic algorithm (GA) is adopted to solve the optimal crack depth. The simulations show that the production of ethylene and propylene is improved and validate the effectiveness of optimal strategy. The models of modeling and optimal control strategy are applied to the energy efficiency evaluating system, which has great significance in engineering application.