experts. There is a tendency to build all-encompassing, refinery-wide models that focus not only on the FCC but also include many ancillary upstream and downstream pro - cesses. While technically impressive, these models are often too complex for most users to use successfully. The recommendation is for refiners to construct and main - tain multiple models for the FCC unit, with varying levels of complexity and focus. Each must be clearly marked for intended use, validity, and distributed appropriately. To max- imise the effectiveness of a kinetic process model, select the version having the minimum level of detail necessary to com - pletely answer the question. This offers several advantages: • This model requires the least amount of data to execute. • Novice users have a better chance of successful execution. • If trouble occurs, a novice is more likely able to diagnose and correct a simpler model. • A smaller result set simplifies review, increasing the chances of the user recognising issues affecting result qual - ity or confidence. In summary, when modelling, simpler is often better. Catalyst variable best practices Part of the challenge of operating an FCC unit is that some process variables react rapidly, whereas others change gradually. One reason is that FCC units contain large cat - alyst inventories with catalyst replacement rates of only a few per cent per day. During the model calibration (discussed in detail later), kinetic models do not evaluate multiple cases simultane - ously. Calibration only uses data from each case individually. If the catalyst addition rate is unusually high or low on a par - ticular calibration case, the catalyst deactivation calibration factors for that day will be biased. Should that case be used as the basis for a prediction, predicted catalyst activity, cata - lyst addition rate, and equilibrium catalyst (E-cat) metals will be in error. As a best practice, daily catalyst addition rates should be averaged over at least 10 days to reduce variability. The same recommendation applies to the E-cat activity measurements. Vendor-reported E-cat activity results gen - erally have a nominal accuracy of around 1%. Variations within this range can safely be averaged, reducing variation in catalyst deactivation calibration factors. FCC units, especially those processing high-metals feeds, may supplement fresh catalyst additions with purchased E-cat to increase the catalyst replacement rate and flush metals from the unit. Purchased E-cat addition rates are typ - ically 20-50% of the total catalyst addition. For units using purchased E-cat, it is worth confirming that the model accu - rately predicts changes in catalyst activity, as the total cata- lyst addition rate and fresh purchased blend ratio are varied. Purchased E-cat is generally sourced from units with low feed metals and low regenerator severity. When designing a fresh catalyst for these units, hydrothermal stability and metals tolerance are low-priority objectives. However, these are precisely the catalyst features that would benefit the sec - ond user. Consequently, purchased E-cat is typically lower in hydrothermal stability metals tolerance than fresh catalyst formulated specifically for the severe unit. When validating model performance for changing the addition rate and blend
ratio, compare model results to real-world experience. If that is not possible, the following rules of thumb may be useful: • Purchased E-cat stabilities are usually 80-90% of that of the fresh catalyst. • Replacing one unit of fresh catalyst with two units of pur - chased E-cat generally results in constant E-cat activity. Exceptions will occur. Major deviations from these rules of thumb should be thoroughly investigated and understood before trusting the model to accurately predict catalyst activ - ity, addition rate, or optimise fresh/purchased blend ratios. Calibration best practices A frequent question is how often FCC kinetic models should be calibrated. Before answering, it is important to understand what model calibration is and how it works. Calibration is the process of aligning the model’s outputs with the observed unit results using the same independent variable values. This is key in enabling the model to simulate real-world behaviour. During calibration, the user provides both the independ- ent variable values and the observed dependent variable values for a given day. Calibration factors exist within key equations inside the model. Calibration is the simultaneous adjustment of these factors’ values to minimise the differ - ence between the model’s results and the observed results. The beauty of calibration is that the reason for the difference between theory and reality does not matter, as calibration ensures alignment regardless of the cause. Calibration is not a once-and-done process. Calibration factors may change for good reason, such as a unit revamp or unexpected shutdown. When the unit restarts, if events during the shutdown result in changed relationships between independent and dependent variables, and the model cannot reconcile the change based on first-princi - ple inputs, calibration factors must change to maintain the alignment between observations and model results. A second reason calibration factors change is catalyst for - mulation. No commercially available models contain all the variables necessary to fully characterise the catalyst man - ufacturing process across all suppliers. Instead, successful models make little or no attempt to model catalyst design effects; they rely upon calibration factors to capture all the catalyst design effects in the unit. As will be discussed in the final section, this approach enables the model to effectively post-audit catalyst performance without requiring specific proprietary catalyst design and formulation knowledge. Regarding calibration frequency, the best practice is to cal - ibrate the model routinely and as often as possible. Process streams are not always sampled daily. Combined feed and liquid products may be sampled only once or twice per week. It is preferable to sample all streams on the same day. Since E-cat analysis is recommended to be averaged over 10 days or more, it is desirable but not mandatory to sample E-cat simultaneously. Best practice is to calibrate the model on each day for which a complete set of process samples is available. After successful calibration, the next action is not to replace all the previous calibration factor values with the newly cal - culated factors. Instead, the most important action is to com - pare the latest set of calibration factors against the library of historical factor values. Factors will naturally vary due to
23
Catalysis 2026
www.digitalrefining.com
Powered by FlippingBook