Research On Modeling And Optimization Of Soft Sensing Based On Diene Yield

Ethylene is an important organic chemical raw materials,ethylene industrial products and derivatives accounted for more than 75%of petrochemical products.At present,the ethylene industry has been regarded as one of the most important indicators to measure the development level of a country’s petrochemical industry.In recent years,China’s ethylene industry has made great progress,but there is a gap between ethylene technology level and developed countries.In order to improve the competitiveness of China’s ethylene industry and achieve product quality control and stable operation,it is necessary to strengthen the real-time detection and optimization control for the important production quality index parameters of ethylene plant,such as the yield of diene.At present,due to the limitation of online analytical instrument technology,the yield of diene is difficult to measure online and in real time.Research on soft sensor technology and optimization of ethylene yield based on ethylene production process is developed.The contents of the thesis including the following aspects:First,according to the production process of ethylene production by naphtha as raw material,the cracking reaction process and the mechanism of naphtha cracking reaction were studied.The main factors affecting the yield of diene products were found out,and the auxiliary variables needed for the soft measurement modeling of diene yield were determined.Secondly,based on the basic principles of soft sensing technology and common modeling methods,the RBF(Radial Basis Function)neural network algorithm is deeply studied.In order to improve the accuracy of the initial center point selection,an improved RBF neural network modeling method of K mean clustering is proposed.In order to make the established soft measurement model adapt to the variety of working conditions,an improved RBF neural network modeling method for K mean clustering double model structure is proposed,and a soft measurement model of diene yield in ethylene plant is established by using two modeling methods.Finally,based on the modified K mean clustering double model structure RBF neural network double ene yield soft measurement model,with the economic benefit of diene product as the optimization index,based on the adaptive mutation particle swarm optimization algorithm,the operation parameters of the cracking furnace are optimized,and the optimal operating parameters are obtained,and the ethylene plant is realized.The optimization goal of maximizing the benefit of diene products.