1. MPC in Statoil Stig Strand, specialist MPC Statoil Research Center 93  SINTEF Automatic Control Dr. ing 1991: Dynamic Optimisation in State Space Predictive Control Schemes

2. What is MPC? CVs MVs SPs and Limits PID PV SP CO MPC PID PV SP CO
High limit CV1
Set point CV2
MPC
Model
Low limit CV1
PID PV SP CO
PV - Process Value
SP - Set point
CO - controller output
PID - Porportional Integral Derivative, basic feedback controller
CV - Controlled Variable in MPC
MV - Manipulated Variable in MPC
PID controller
MPC controller
1 degree of freedom
More degrees of freedom (# of MVs)
Controls PV to a SP.
Controls CVs to their SP or limits
Has no prediction capability
Has full prediction capability

3. MPC models ex: set of linear SISO step response, superposition  MISO Dynamic response to CV’s (ex AI001) from a unit step in MVs / DVs (ex FIC179)

4. MPC / NMPC / DRTO in SEPTIC
Process u v y x MV DV CV
state
model:
constraints:
discretization:
Prediction horizon
Current t
Controlled variable (n = 9)
Manipulated variable (k = 6)
Set point
Explicit control priorities - solved before the dynamic problem
MV rate of change limits
MV high/low Limits
CV hard constraints (”never” used)
CV soft constraints, CV set points, MV ideal values: Priority level 1
CV soft constraints, CV set points, MV ideal values: Priority level 2
CV soft constraints, CV set points, MV ideal values: Priority level n
CV soft constraints, CV set points, MV ideal values: Priority level 99
Sequence of steady-state QP solutions to solve 2 – 7 (or NLP if nonlinear models)
Then a single dynamic QP to meet the adjusted and feasible steady-state goals (or iterated QP if nonlinear models)
MV blocking  size reduction
CV evaluation points  size reduction
CV reference specifications  tuning flexibility set point changes / disturbance rejection
Soft constraints and priority levels  feasibility and tuning flexibility

5. MPC – nonlinear models
Open loop response is predicted by non-linear model
MV assumption : Interpolation of optimal predictions from last sample
Linearisation by MV step change
one step for each MV blocking parameter (increased transient accuracy)
QP solver as for experimental models (step response type models)
open-loop nonlinear prediction + QP-solution (delta)  linearized solution = first iteration to “exact” solution
Closed loop response is predicted by nonlinear model
compare nonlinear closed loop response to linearized closed loop response, close enough according to convergency criteria?
Iterate solution until satisfactory convergence
Used in 7 applications of the total 100 (appr) in Statoil

6. Depropaniser MPC development Kårstø Gas Plant 2004 Introduction to:
Depropaniser MPC development Kårstø Gas Plant Introduction to: design inferential models – quality estimators – soft sensors step testing commissioning

17. MPC maintenance 25% stop working a few months after commissioning.
40% fail within 3 years after commissioning due to lack of support and monitoring.
25% continue online with reduced benefits.
10% continue online with increased benefits due to good support and monitoring.

18. Kalundborg refinery – Cat reformer Oscillations – why?

19. Missing models – model verification – extended model matrix

20. Total in Statoil: 92 MPC Applications
Snøhvit #4 #7
Åsgard Norne Heidrun
Mongstad
#28
Total in Statoil: 92 MPC Applications
Kollsnes #5 #2
Gullfaks/Tordis
Kårstø
#25
Kalundborg
#21

21. Process Control – an overview
Basic Control (PCS)
(PID, FF,..) (seconds)
Supervisory Control
Model based multivariable control
(MPC) (minutes)
Real Time
Optimisation
(RTO, MPC) (hours)
Planning
Scheduling
(days, weeks)
MPC/RTO
MPC
MPC – Model Predictive Control
RTO – Real Time Optimization
Wikipedia: FC
Basic Control
Manual Control

22. Planning and control layers in oil refining
INTEGRATION OF REFINERY SYSTEMS
CORPORATE PLANNING
REFINERY PLANNING
PRODUCTION SCHEDULING
PLANT OPTIMIZATION
MULTI-UNIT OPTIMIZATION
UNIT OPTIMIZER
EQUIPMENT
OPTIMIZER
MULTI-PERIOD
FUNCTIONS
ADVANCED CONTROL SYSTEM
CONSTRAINT
CONTROL
MULTIVARIABLE
QUALITY
REGULATORY SYSTEM
STEADY-STATE
OR DYNAMIC
DYNAMIC
MONTHS
WEEKS
DAYS
MINUTES
SECONDS
HOURS
PRODUCTION CAMPAIGN & ORDER EXECUTION

23. SEPTIC applications Mongstad 2014

24. Mongstad Refinery - MPC & (D)RTO
INTEGRATION OF REFINERY SYSTEMS

25. Mongstad Refinery – Coker Unit distillate revamp – increased Diesel production

26. Mongstad Refinery – Coker Unit Frac, re-done now Top temperature modelling

27. Mongstad Refinery – Coker Unit Frac, re-done now Bottom temperature modelling

28. INTEGRATION OF REFINERY SYSTEMS
Mongstad Refinery – Coker Unit Frac, re-done now Models implemented today (Wed Nov 16)
INTEGRATION OF REFINERY SYSTEMS

29. INTEGRATION OF REFINERY SYSTEMS
Mongstad Refinery – Coker Unit Frac, re-done now Activated today, 1 hour before drum switch (Wed Nov 16)
INTEGRATION OF REFINERY SYSTEMS

30. CDUOPT (Crude Unit Optimizer) - process scope
Existing MPC's
MPCFVRM
T-150
Crude preheat
12 MV's, 32 CV's
MPCSPLT
T-108, T-112, T-113
17 MV's, 23 CV's
MPCT101
T-101, A-500, A-5100
36 MV's, 62 CV's

31. Crude properties – from crude assays
Classification: Internal

32. Crude properties – smoothed for optimization
Classification: Internal

33. Our strategy RTO General (MPC, RTO) keep it as simple as possible
Scope based on a system analysis, include what we think is needed (can be extended later)
Modelling: we keep in mind that the system should solve
Use MPC, everyone is familiar with that, it is accepted, and it can prioritize between objectives
Optimized solution can be understood
Fast execution of project and running application
use dynamic models
When disturbances with economic impact (as most of them have) are frequent relative to the process dynamics
When model updating during transients is important to get satisfactory execution cycle
General (MPC, RTO)
use internal resources, only
develops competence, ownership, shared goals, and cooperation
prioritize maintenance
there are always things to be improved, if you don't, the application will degrade, and die
return of investment depends on the application performance
maintenance and good performance make confidence