Data-driven modeling and global optimization of industrial-scale petrochemical planning operations

In this work, we have developed a data-driven model which is used for the optimization of planning operations of a large petrochemical complex, comprising of a petrochemical plant and two ethylene plants. We have developed unit operation models for all of the processes present within the industrial superstructure, which are integrated with mass-balance, property specification, demand, capacity and unit selection constraints, to form the overall planning problem. The models in this formulation contain parameters which are automatically fitted based on the data obtained from the operation of the plant units. For the dynamic updating of the model parameters, we have developed a user-friendly computational platform which allows the input of new operational data as well as cost, price, demand and specification information for the planning period of interest. Once the parameters are updated and the predictive ability of the models is confirmed, the formed mixed integer nonlinear optimization problem is solved to global optimality, providing the globally optimal flowrates and operating modes which maximize the profit, while simultaneously satisfying specification and demand constraints. Using the developed framework, we have obtained results for multiple case studies proving that the obtained solutions lead to significant improvements in profit when compared to historically applied operating plans.