Parameter estimation and sensitivity analysis for a diesel hydro-processing model

Model-based approaches are essential for the operation, optimization, and control of applications in the process industry. Different structures are often investigated to build representative and robust models, and a set of parameters with the same attributes are required to utilize them effectively. Parameter estimation gets arduous with the increasing complexity of the process, the model, and the size of the parameter space. In this work, a parameter-estimation problem based on a steady-state model of diesel hydrodesulfurization is investigated using gradient-based and gradient-free optimizers. The optimal parameter sets obtained are then assessed in terms of performance and computational time for the different optimizers. Furthermore, the sensitivity of the various parameters is also investigated. Due to the catalytic reactions in this process, some parameters have to be updated depending on the catalyst activity. In addition to the initial estimation, the updated parameters are also studied, and instead of a time-based one, a tolerance-based recalculation schedule is suggested. Finally, the robustness of the final model is analyzed by giving different operating conditions and feed characteristics. The adaptive parameter approach proved better data fitting capabilities by improving the coefficient of determination for temperature predictions.