Application of aritificial neural networks to the optimization of a catalytic reforming unit

• Main Authors: Hsin-fa Huang, 黃信發
• Other Authors: James I.C. Chang
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/03028289908143966870

碩士 === 國立高雄第一科技大學 === 環境與安全衛生工程研究所 === 99 === Catalytic reforming unit converts straight-chain hydrocarbons into cyclic aromatics, which can be blended into gasoline to improve the octane number or produce benzene, toluene and xylene after fractionation. To improve its operating conditions is an important task to oil refiners. This work is to explore the possibility of find the optimal operating conditions using mathematical tools. The compositions of raw materials, products, and operating data such as the reacting temperature, pressure, resident time, etc were between January and October of the year 2010 were first collected and analyzed, and then a artificial neural network program was applied to establish empirical models between the output (product compositions and properties) and the input data (raw materials and operating conditions). Experimental design and multivariate regression analysis were used to obtain a set of simplied models for the study the effects of operating conditions and optimization. The results showed that the models developed by the artificial neural network could accurately predict the product compositions and properties, which could be used for online monitoring. The conditions for maxium profit were found to be at the reacting temperature (WAIT) of 520 C, fractionating temperature of 243 C and the resident time of 0.78 seconds. For a 40,000bbld unit operated at 91.1% capacity (36,450 bbld or MT/d), the revenue would increase 2.46%, which was equivalent to 894 million NT dollars. The procedures developed in this work could be also used for any other production units as long as abundant historic operating data were available.