FCC process modeling and light oil yield optimization research

The fluidized catalystic cracking (FCC) is the most important secondary processing technique in oil refining industry. The authors use the multilayer forward neural network for identifying and establishing the model of FCC reaction process, and the self-correlation function in checking the reliability of the model. In the research of light oil yield optimization the authors develop a series of optimization steps from both the Frank-Wolfe algorithm and approximation programming. The use of this method with real production data can lead to a satisfactory result. It is feasible to apply this optimization method into the coordinator controller in the step-up control system of the light oil production. Adjusting the online industrial procedure parameters according to the optimization object, the coordinator controller can make the whole system run on an optimal state and enhance the yield of light oil.