(481h) Multiobjective Optimization of Resid Fluidized Catalytic Cracking Unit through Data-Driven Approach

The sulfur content in marine fuel is limited to 0.5 wt.% from 3.5 wt.% as per recent International Maritime Organization (IMO) regulations. Resid fluidized catalytic cracking unit (RFCCU) is one of the key units in an oil refinery that can desulfurize the residues and help in producing residual bunker fuels like gasoline. Modeling and optimization of RFCCU found to be difficult due to their complexities like time-varying feed properties and computationally expensive objective function. The optimal solution can be obtained in lesser time by approximating a black box formulation as a surrogate model. Over the decades, surrogate models are extensively utilized in simulation, prediction, automation applications. In this paper, a hybrid of the genetic algorithm – radial basis function, termed as SOCEMO (Surrogate Optimization of Computationally Expensive Multi-objective Problems) algorithm, has been utilized to solve the multi-objective optimization problem of RFCCU. SOCEMO uses a variety of sampling methods to generate new points for the evaluation of expensive function evaluation. Subsequently, the seven lump kinetic model is used as a first principle model, which is the primary multi-objective function. SOCEMO has generated good quality non dominated solutions at very low computational cost as compared to genetic algorithm and its variants. The optimum parameters have been found to get maximum yield and conversion of gasoline while minimizing the coke formation. Further, the linear regression and support vector regression models are also constructed to test the quality of the optimal solution obtained within seconds. This algorithm may find applications in other chemical engineering problems like hydrocracking and steam reforming.