!!!!! !!! !!! Lubes VDU product property prediction and control. Petrocontrol. Y. Zak Friedman, PhD, Petrocontrol. Syazwan Kamarunzaman, PETRONAS

Transcription

1 Petrcntrl Leader in inferential cntrl technlgy Lubes VDU prduct prperty predictin and cntrl By Y. Zak Friedman, PhD, Petrcntrl Syazwan Kamarunzaman, PETRONAS Azura Bt. Azahar, Seung Yun Nam, PETRONAS Cnpt 34 East 30 th Street, New Yrk, NY Tel: Fax:

2 Overview Lubes VDU prduct prperty predictin and cntrl Lubes VDU (vacuum distillatin unit fr prductin f lube plant feedstck) is different frm a fuels VDU. Figure 1 shws a fuels VDU cnfiguratin. Reduced crude feed cmes frm the CDU (atmspheric crude unit) thrugh a furnace and int the flash zne. There are tw distillate prducts: LVGO (light vacuum gasil) and HVGO (heavy vacuum gasil). LVGO is diesel range material, ging int the diesel pl. HVGO is C r hydrcracker feedstck. There is a pssibility t als draw vacuum wax, thugh nrmally wax is circulated back t the furnace t imprve the separatin. As is cmmn in vacuum clumn designs, the draws are frm ttal draw trays. Part f the draw is pumped arund thrugh heat exchangers t cl the clumn, anther part is pumped dwn as ht reflux, and a third part is taken ut as prduct. Ecnmics usually invlve maximizing vacuum distillates t meet distillate 90% pint targets. The lubes VDU is a mre cmplex clumn, having nt tw but five distillate prducts. Shwn in figures 2 and 3, V1SS, V2SS and V3SS are the main prducts: lube unit feeds fr making 100N, 150N and 500N lube-ils respectively. LVGO, while a valuable diesel material is f lesser value. MVGO is t light t serve as lube unit feed. It is taken int the hydrcracker at a much lwer value. Lube VDU APC ecnmics What are we t cntrl in APC t maximize unit prfitability? 1. The main V1SS, V2SS and V3SS prperty t be cntrlled is distillatin 50% pint, because eventually lube-il viscsity is a functin f feedstck 50% pint. While KW (Watsn K factr), a measure f armatics, als affects viscsity, the lube unit reactrs saturate armatics, mdifying KW, and thus the starting feed KW is incnsequential. 2. The secnd lube feed prperty f imprtance is distillatin 95% pint. The difference between 95% and 50% defines the distillatin range and hence yield f the three lube feeds. 3. Our initial APC strategy maximized ttal yield (V1 + V2 + V3), nly t find ut that relative yields are als f imprtance because in the end the lube plant must satisfy lcal demand. Thse relative yields then became a third prperty r imprtance. 4. Clumn separatin is als f imprtance, and we can manipulate the internal reflux prfile t reduce heavy tail cntaminatin 5. Unfrtunately there are n side strippers and hence light tail cntaminatin is unavidable. Light tail eventually affects the finished lube-il yield, but there is nt much we can d abut it in APC. 6. LVGO is t be maximized t diesel specificatin. Thus, MVGO is minimized between diesel target cutpint and V1SS target 50% pint and cutpint range range. Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 2

3 VDU inferential variables 1. LVGO 90% pint 2. MVGO 90% pint in pen lp 3. V1SS, V2SS, V3SS 50% pints 4. V1SS, V2SS, V3SS 95% pints 5. V1SS, V2SS, V3SS 5% pints in pen lp Given that there are n side strippers and that we are already cntrlling lube feed biling ranges, nthing else can be dne t cntrl 5% pints 6. V1SS, V2SS, V3SS 100ºC in pen lp As stated abve, the specificatin is nt lube feed viscsity but final lube prduct viscsity, and we attempt t predict that final viscsity 7. Tp dew pint inference Fr crrsin prtectin. The verhead flw is mstly steam 8. MVGO internal reflux (that is a partial draw tray with n Pumpdwn arrangement) Main manipulated variables Pumparunds and pumpdwns Manipulated fr prduct cutpint and internal reflux prfile cntrl MVGO draw Cil utlet temperature Manipulate fr cntrl f V3SS 95% pint Bttm steam Manipulated at target raticmparisn f inferential mdels against lab We use tw mnths trends t illustrate the inferential fidelity. Inferential predictins are blue lines, whereas lab data is shwn as cntinuus green lines. These green lab data have a meaning nly at the time when a change f value takes place. That is the time when a new test data pint had been written in. The figures f interest are: Figure 4. LVGO and MVGO 90% pint predictins and lab values Figure 5. V1SS, V2SS and V3SS 50% pint predictins and lab values Figure 6. V1SS, V2SS and V3SS 95% pint predictins and lab values Figure 7. V1SS, V2SS and V3SS viscsity predictins and lab values We can see that the distillatin predictins are quite gd. The viscsity predictins are nt perfect but they d nt need t be, as the intentin is t predict viscsities f the final lube prduct and nt f the lube feedstck. Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 3

4 Operatin under APC We use tw mnths trends t illustrate the inferential fidelity. Inferential predictins are blue lines, whereas lab data is shwn as cntinuus green lines. These green lab data have a meaning nly at the time when a change f value takes place. That is the time when a new test data pint had been written in. The figures f interest are: Figure 4. LVGO and MVGO 90% pint predictins and lab values Figure 5. V1SS, V2SS and V3SS 50% pint predictins and lab values Figure 6. V1SS, V2SS and V3SS 95% pint predictins and lab values Figure 7. V1SS, V2SS and V3SS viscsity predictins and lab values We can see that the distillatin predictins are quite gd. The viscsity predictins are nt perfect but they d nt need t be, as the intentin is t predict viscsities f the final lube prduct and nt f the lube feedstck. APC benefits cme in three flavrs. First, cntrlling the 50% pints and prduct cutpint ranges guarantees stable quality prducts at the crrect yields. That maximizes nt nly utilizatin f the lube plant but als lube yields. Secnd, APC maximizes the yield f diesel at the expense f MVGO. Figures 8 and 9 shw key prduct yields befre and after APC. As can be seen, LVGO yield went frm 14 t 16%, whereas MVGO, came dwn frm 12 t 10%. The price difference between thse tw prducts: diesel versus hydrcracker feed is large, yielding high benefits. Third befre APC the clumn was perating at high wash zne lading and unnecessarily high COT (cil utlet temperature). Figure 10 trends the change in clumn energy use after cmmissining. It can be seen that ver a shrt time COT drpped 5ºC, wash zne flw drpped 10 M 3 /Hr. An added bnus was a drp in flash zne pressure f 5 mmhg. Figure 11 trends tis srt f energy imprvement KPIs ver 300 days, shwing the gradual imprvement under APC. Cnclusins Quality inferences plus understanding f the ecnmics have permitted visible peratinal imprvements f this clumn. The yield f LVGO is maximized against MVGO and lube feed cutpints are cnsistently at targets. The furnace fuel demand drpped abut 3%. APC service factr is high. Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 4

5 Figure 1. Fuels vacuum clumn cnfiguratin TC T ejectrs LVGO pumpdwn LVGO pumparund LC LVGO HVGO pumpdwn HVGO pumparund LC HVGO Furnace LC Vacuum Wax TC LC Asphalt Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 5

6 Figure 2. Lubes vacuum fractinatr upper part Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 6

7 Figure 3. Lubes vacuum fractinatr lwer part Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 7

8 Figure 4. LVGO, MVGO 90% pint trend against lab Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 8

9 Figure 5. V1SS, V2SS, V3SS 50% pint trend against lab Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 9

10 Figure 6. V1SS, V2SS, V3SS 95% pint trend against lab Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 10

11 Figure 7. V1SS, V2SS, V3SS 100ºC trend against lab 12 V1V100_M V1V100_L V2V100_L V2V100_M V3V100_M V3V100_L V3SS viscsity 8 15 V1SS, V2SS viscsity Jun Jun Jul Aug Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 11

12 Figure 8. Key prduct yields befre APC 20 YLVG YMVG YVBOT Jun Jun Jul Jul Jul Aug-2013 Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 12

13 Figure 9. Key prduct yields with APC 20 YLVG YMVG YVBOT Sep Sep Sep Oct Oct-2013 Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 13

14 Figure 10. Energy ptimizatin Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 14

15 Figure 11. Energy imprvement ver 300 days Lubes VDU prduct prperty predictin and cntrl, Nvember 2013, Page 15