Sensitivity Analysis and Development of a Set of Rules to Operate FCC Process by Application of a Hybrid Model of ANFIS and Firefly Algorithm

Fluid catalytic cracking (FCC) process is a vital refinery process which majorly produces gasoline. In this research, a hybrid algorithm which was constituted of Adaptive Neuro-Fuzzy Inference System (ANFIS) and firefly optimization algorithm was developed to model the process and optimize the operating conditions. To conduct the research, industrial data of Abadan refinery FCC process were carefully gathered along a definite period. Then a model based on ANFIS was developed to investigate the effect of operating variables including reactor temperature, feed flow rate, temperature of top of main column, and the temperature of bottom of the debutanizer tower on quality and quantity of gasoline, LPG flow rate, and process conversion. Moreover, statistical parameters comprising R2, RMSE, and MRE approved the accuracy of the model. Eventually, validated ANFIS model and firefly algorithm as an evolutionary optimization algorithm were applied to optimize the operating conditions. Also, three different optimization cases including maximization of Research Octane Number (RON) , gasoline flow rate, and conversion were considered. In addition, to obtain maximum target output variables, inlet reactor temperature, temperature of top of main column, temperature of the bottom of debutanizer column, and feed flow rate should be respectively set at 523, 138, 183 ‎°C and 40731 bbl/day. Also, the sensitivity analysis between the input and output variables was carried out to derive some effective rules of thumb to facilitate operation of the process under unsteady state conditions. Finally, the obtained result introduces a methodology to compensate for the negative effect of undesirable variation of some of the operating variables by manipulating the others.