Petronas Reformer paper October 2007 H.K. Lim Y. Zak Friedman S.Y. Nam 1 Petrocontrol
Simplified reformer configuration Reactor Furnace Compressor DeC2 DeC3 C2- Regenerator DeC4 Propane Feed Naphtha H2 Butane 2 Petrocontrol
Aromatization + H2 + H2 Main reaction, nearly 100% conversion The naphtha product is very aromatic high octane The reformer is a large producer of hydrogen 3 Petrocontrol
Dehydrocyclization + H2 Difficult reaction, promoted by low pressure NC6 reaction not very effective in semiregenerative units, ~20% conversion CCR units (low P) NC7 converts better, NC8 better yet, ~50% conversion 40-80+% conversion 70-95+% conversion 4 Petrocontrol
Cracking side reaction + Un-dehydrocyclized, low octane components, are cracked 5 Petrocontrol
Aromatics condensation Coke + + H2 Coke gradually accumulates on the catalyst surface Working with high ratio H2 circulation to tilt the balance against coking 6 Petrocontrol
The reaction is endothermic Temperature profile 430 C First reactor Second reactor Third reactor 500 C 7 Petrocontrol
Main reactor controls TC LC Regenerator TC TC FC FC H2 to HCR PC FC FC H2 purge Feed H2 recycle 8 Petrocontrol
Reactor control Control octane by changing WAIT (weighted average reactor inlet temperature) Keep all inlets the same, except as needed to alleviate constraints Control catalyst coking rate by manipulating hydrogen recycle ratio Keep the coking rate within the regenerator capacity Regenerator reliability problem Frequent regenerator trips may dictate temporary throughput cuts 9 Petrocontrol
Manipulated variables Reformer feed flow Maximize feed subject to reactor and NHT throughput constraints. Reactor inlet temperatures Handles for controlling reactor severity Recycle compressor suction valve Handle for controlling the hydrogen recycle ratio 10 Petrocontrol
Control variables 1 Octane number calculation Octane inference is the main CV for controlling the reformate octane Coke on catalyst laydown calculation Coke laydown inference may become an active constraint upon regenerator trip Reactor inlet H2/HC mole ratio Normally the H2/HC ratio is set at 1.5, though it should be increase if there is a coke laydown problem 11 Petrocontrol
Control variables 2 Throughput hydraulic constraints Recycle compressor constraints Reactor temperature profile CVs Profile is to be respected subject to furnace constraints Furnace constrain CVs Certain NHT constraints In rare situations the upstream naphtha hydrotreater may impose throughput limits on the reformer 12 Petrocontrol
Control variables 3 Hydrogen purge Half way through the project economic drives changed. The reformer is now operated solely for the purpose of supplying hydrogen We added a hydrogen purge constraint to keep the purge low 13 Petrocontrol
This control relies on knowledge of octane and rate of coking Octane measurement (or in our case inference) is a must Keep catalyst coking rate within regenerator capacity Steady octane operation improves yield Coking rate inference is good to have Otherwise we must operate at a conservative hydrogen ratio Affects unit efficiency and capacity 14 Petrocontrol
What affects octane? Feed boiling point Conversion increases with molecule size PNA (paraffin naphthene aromatic) Aromatics ride through, they already have high octane Naphthenes convert at 100% to high octane aromatics, but that conversion requires high reactor temperature H2 recycle Partial pressure effect is minor 15 Petrocontrol
WAIT What affects coke make? Coke make increases with reactor temperature, especially in the last reactor H2 recycle Partial pressure effect is major 16 Petrocontrol
Inference of feed boiling point Ideally feed boiling point should be inferred on the upstream crude unit Not possible at Melaka because there are several feed sources Inference is based on debutanizer bottom conditions, corrected for C4 and C5 separation 17 Petrocontrol
Inference of feed PNA Ideally feed PNA is inferred from Feed boiling range Reactor conditions Feed density But the feed density analyzer is not installed yet, and the model assumes a constant PNA distribution Quality of octane inference is acceptable but not great 18 Petrocontrol
Octane trend at Melaka 102 RON_M RON_L 101 100 99 98 17-Jul-06 00 27-Jul-06 00 06-Aug-06 00 16-Aug-06 00 26-Aug-06 00 05-Sep-06 00 15-Sep-06 00 25-Sep-06 00 05-Oct-06 00 15-Oct-06 00 19 Petrocontrol
Octane trend at another location, incorporating density measurement 105 RON_1 RON_LAB 104 103 102 101 100 99 98 97 20 Two Months Petrocontrol
Coke deposit trend [Kg/Hr] 24 Coke burn Coke_M2 23 22 21 20 19 20-Jul-06 00 30-Jul-06 00 09-Aug-06 00 19-Aug-06 00 29-Aug-06 00 08-Sep-06 00 18-Sep-06 00 28-Sep-06 00 08-Oct-06 00 21 Petrocontrol
% coke on catalyst 4.50 4.40 CokeBurnCOC COC_M 4.30 4.20 4.10 4.00 3.90 3.80 3.70 3.60 3.50 20-Jul-06 00 30-Jul-06 00 09-Aug-06 00 19-Aug-06 00 29-Aug-06 00 08-Sep-06 00 18-Sep-06 00 28-Sep-06 00 08-Oct-06 00 22 Petrocontrol
How this application makes money? At normal economic situation Maximize feed at constant octane, considering reactor and regenerator constraints At current depressed economics Minimize feed at constant octane to provide the needed H2 supply while keeping H2 purge at minimum A feed reduction of 5% was observed 23 Petrocontrol
Conclusions Reforming is a sensitive high temperature catalytic process with many constraints APC of the reformer reactor requires reasonable inferences of Reformate octane number Catalyst coke deposit rate APC engineers must adjust applications for the economics of the day 24 Petrocontrol