TIME-SERIES MODEL: HYBRID FIRST-PRINCIPLE WITH OPERATIONAL DATA FOR OPTIMISATION OF AN INDUSTRIAL ENDOTHERMIC REACTOR SYSTEM

Abstract

It is crucial to identify and manage main pain point contributor(s) in the hydrocarbon value chain in maximising revenue and cash generation. Recently, PETRONAS has embarked on the initiative to maximise the production of propane value chain by addressing the run length limitation issue due to increasing coking rate in one endothermic reactor system. Hybrid of the first principle and plant operational data were used and modelled in a time-series manner. Several high impact influential variables were selected and incorporated into the model, i.e. hydrocarbon throughput, hydrogen-to-hydrocarbon ratio, reactor inlet temperature and days-onstream. The hybrid modelling approach is used to find optimum start-of-run operating conditions, control coke formation rate in the presence of variations in operating conditions throughout the run length, and manoeuvre the intermediate operating strategies to achieve the desired run length. The current operation with the implementation of new operating strategies derived from the model shows coke formation reduction up to 70% compared to the historical trending, on-track to achieve the desired run length equivalent to the catalyst expected life, which has never been achieved since in operation. The model can be enhanced in future work to achieve the maximum possible yield throughout the run length and predict the anticipated production with the corresponding operating strategies.