1 Outline Skogestad procedure for control structure design I Top Down Step S1: Define operational objective (cost) and constraints Step S2: Identify degrees of freedom and optimize operation for disturbances Step S3: Implementation of optimal operation –What to control ? (primary CV’s) (self-optimizing control) Step S4: Where set the production rate? (Inventory control) II Bottom Up Step S5: Regulatory control: What more to control (secondary CV’s) ? –Distillation example Step S6: Supervisory control Step S7: Real-time optimization

2 ”Advanced control” STEP S6. SUPERVISORY LAYER Objectives of supervisory layer: 1. Switch control structures (CV1) depending on operating region –Active constraints –self-optimizing variables 2. Perform “advanced” economic/coordination control tasks. –Control primary variables CV1 at setpoint using as degrees of freedom (MV): Setpoints to the regulatory layer (CV2s) ”unused” degrees of freedom (valves) –Keep an eye on stabilizing layer Avoid saturation in stabilizing layer –Feedforward from disturbances If helpful –Make use of extra inputs –Make use of extra measurements Implementation: Alternative 1: Advanced control based on ”simple elements” (decentralized control) Alternative 2: MPC

3 Control configuration elements Control configuration. The restrictions imposed on the overall controller by decomposing it into a set of local controllers (subcontrollers, units, elements, blocks) with predetermined links and with a possibly predetermined design sequence where subcontrollers are designed locally. Some control configuration elements: Cascade controllers Decentralized controllers Feedforward elements Decoupling elements Input resetting/Valve position control/Midranging control Split-range control Selectors

4 Cascade control arises when the output from one controller is the input to another. This is broader than the conventional definition of cascade control which is that the output from one controller is the reference command (setpoint) to another. In addition, in cascade control, it is usually assumed that the inner loop K2 is much faster than the outer loop K1. Feedforward elements link measured disturbances to manipulated inputs. Decoupling elements link one set of manipulated inputs (“measurements”) with another set of manipulated inputs. They are used to improve the performance of decentralized control systems, and are often viewed as feedforward elements (although this is not correct when we view the control system as a whole) where the “measured disturbance” is the manipulated input computed by another decentralized controller.

5 Split Range Temperature Control

6 Note: adjust the location er E0 to make process gains equal E0

7 Sigurd’s pairing rule for decentralized control: “Pair MV that may (optimally) saturate with CV that may be given up” Reason: Minimizes need for reassigning loops Important: Always feasible (and optimal) to give up a CV when optimal MV saturation occurs. –Proof (DOF analysis): When one MV disappears (saturates), then we have one less optimal CV.

8 Use of extra measurements: Cascade control ( conventional) The reference r 2 (= setpoint y s2 ) is an output from another controller General case (“parallel cascade”) Special common case (“series cascade”) Not always helpful… y 2 must be closely related to y 1

9 Series cascade 1.Disturbances arising within the secondary loop (before y 2 ) are corrected by the secondary controller before they can influence the primary variable y 1 2.Phase lag existing in the secondary part of the process (G 2 ) is reduced by the secondary loop. This improves the speed of response of the primary loop. 3.Gain variations in G 2 are overcome within its own loop. Thus, use cascade control (with an extra secondary measurement y 2 ) when: The disturbance d 2 is significant and G 1 has an effective delay The plant G 2 is uncertain (varies) or nonlinear Design / tuning (see also in tuning-part): First design K 2 (“fast loop”) to deal with d 2 Then design K 1 to deal with d 1 Example: Flow cascade for level control u = z, y2=F, y1=M, K1= LC, K2= FC

10 Use of extra inputs Two different cases 1.Have extra dynamic inputs (degrees of freedom) Cascade implementation: “Input resetting to ideal resting value” Example: Heat exchanger with extra bypass Also known as: Midranging control, valve position control 2.Need several inputs to cover whole range (because primary input may saturate) (steady-state) Split-range control Example 1: Control of room temperature using AC (summer), heater (winter), fireplace (winter cold) Example 2: Pressure control using purge and inert feed (distillation)

11 Extra inputs, dynamically Exercise: Explain how “valve position control” fits into this framework. As en example consider a heat exchanger with bypass

12 QUIZ: Heat exchanger with bypass Want tight control of T hot Primary input: CW Secondary input: q B Proposed control structure? qBqB T hot closed

13 qBqB T hot TC Use primary input CW: TOO SLOW Alternative 1 closed

14 qBqB T hot TC Use “dynamic” input q B Advantage: Very fast response (no delay) Problem: q B is too small to cover whole range + has small steady-state effect Alternative 2 closed

15 qBqB T hot TC Alternative 3: Use both inputs (with input resetting of dynamic input) closed FC q Bs TC: Gives fast control of T hot using the “dynamic” input q B FC: Resets q B to its setpoint (IRV) (e.g. 5%) using the “primary” input CW IRV = ideal resting value Also called: “valve position control” (Shinskey) and “midranging control” (Sweden)

16 Exercise Exercise: (a)In what order would you tune the controllers? (b)Give a practical example of a process that fits into this block diagram

17 Too few inputs Must decide which output (CV) has the highest priority –Selectors Implementation: Several controllers have the same MV –Selects max or min MV value –Often used to handle changes in active constraints Example: one heater for two rooms. Both rooms: T>20C –Max-selector –One room will be warmer than setpoint. Example: Petlyuk distillation column –Heat input (V) is used to control three compositions using max-selector –Two will be better than setpoint (“overpurified”) at any given time

18 Control of primary variables Purpose: Keep primary controlled outputs c=y 1 at optimal setpoints c s Degrees of freedom: Setpoints y 2s in reg.control layer Main structural issue: Decentralized or multivariable?

19 Decentralized control (single-loop controllers) Use for: Noninteracting process and no change in active constraints +Tuning may be done on-line +No or minimal model requirements +Easy to fix and change -Need to determine pairing -Performance loss compared to multivariable control - Complicated logic required for reconfiguration when active constraints move

20 Multivariable control (with explicit constraint handling = MPC) Use for: Interacting process and changes in active constraints +Easy handling of feedforward control +Easy handling of changing constraints no need for logic smooth transition -Requires multivariable dynamic model -Tuning may be difficult -Less transparent -“Everything goes down at the same time”

21 Model predictive control (MPC) = “online optimal control” Note: Implement only current input Δu 1 y dev =y-y s u dev =u-u s Discretize in time:

22 Implementation MPC project (Stig Strand, Statoil) Initial MV/CV/DV selection DCS preparation ( controller tuning, instrumentation, MV handles, communication logics etc ) Control room operator pre-training and motivation Product quality control  Data collection (process/lab)  Inferential model MV/DV step testing  dynamic models Model judgement/singularity analysis  remove models? change models? MPC pre-tuning by simulation  MPC activation – step by step and with care – challenging different constraint combinations – adjust models? Control room operator training MPC in normal operation, with at least 99% service factor DCS = “distributed control system” = Basic PID control layer

23 Depropaniser Train 100 – 24-VE TC 1022 LP condensate LP steam 24 LC PC TI LC HA-103 A/B 24-VA PA-102A/B 24 FC TI LC TI TI PC PDC HC 1015 Kjølevann 24-VE TI TI TI PI PD FC TI 1013 Propane Flare Bottoms from deetaniser 25 FI 1003 Manipulated variables (MV) = Set points to PID controllers 24 TI LC LC1001.VYA Disturbance variables (DV) = Feedforward 24 AR 1005 C = C3 E = nC4 F = C5+ Debutaniser 24-VE AR 1008 B = C2 C = C3 D = iC4 Controlled variables (CV) = Product qualities, column deltaP ++ Normally 0 flow, used for start-ups to remove inerts

24 Depropaniser Train100 step testing 3 days – normal operation during night CV1=TOP COMPOSITION CV2=BOTTOM COMPOSITION CV3= ¢ p DV =Feedrate MV1 = L MV2 = Ts

25 Estimator: inferential models Analyser responses are delayed – temperature measurements respond 20 min earlier Combine temperature measurements  predicts product qualities well Calculated by 24TI1011 (tray 39) Calculated by 24TC1022 (t5), 24TI1018 (bottom), 24TI1012 (t17) and 24TI1011 (t39) CV1=TOP COMPOSITION CV2=BOTTOM COMPOSITION

26 Depropaniser Train100 step testing – Final model Step response models: MV1=reflux set point increase of 1 kg/h MV2=temperature set point increase of 1 degree C DV=output increase of 1%. 3 t 20 min MV2 = Ts MV1 = L DV =Feedrate CV1=TOP COMPOSITION CV2=BOTTOM COMPOSITION CV3= ¢ p

27 Depropaniser Train100 MPC – controller activation Starts with 1 MV and 1 CV – CV set point changes, controller tuning, model verification and corrections Shifts to another MV/CV pair, same procedure Interactions verified – controls 2x2 system (2 MV + 2 CV) Expects 3 – 5 days tuning with set point changes to achieve satisfactory performance MV1 = L MV2 = Ts DV =Feedrate CV1=TOP COMPOSITION CV2=BOTTOM COMPOSITION CV3= ¢ p

28 Reflux drum Reflux pumps Heat ex PC Propane Product pumps LCTC To Depropaniser LP Condensate LP Steam Feed from stabilizators FC Flare 0 – 65% % FC LCPC Fuel gas to boilers CV MV DV Another column: Deethanizer Quality estimator LC

29 Top: Binary separation in this case Quality estimator vs. gas chromatograph (use logarithmic composition to reduce nonlinearity, CV = - ln x impurity ) 7 temperatures 2 temperatures =little difference if the right temperatures are chosen

30 The final test: MPC in closed-loop CV1 CV2 CV3 MV1 MV2 DV

31 Conclusion MPC Generally simpler than previous advanced control Well accepted by operators Statoil: Use of in-house technology and expertise successful

32 Outline Skogestad procedure for control structure design I Top Down Step S1: Define operational objective (cost) and constraints Step S2: Identify degrees of freedom and optimize operation for disturbances Step S3: Implementation of optimal operation –What to control ? (primary CV’s) (self-optimizing control) Step S4: Where set the production rate? (Inventory control) II Bottom Up Step S5: Regulatory control: What more to control (secondary CV’s) ? Step S6: Supervisory control Step S7: Real-time optimization

33 Optimization layer (RTO) Purpose: Identify active constraints and compute optimal setpoints (to be implemented by control layer) CV s Process MVs RTO MPC PID Sigurd Skogestad

34 An RTO sucess story: Statoil Mongstad Crude oil preheat train 20 heat exchangers, 5 DOFs (splits), 85 flow andf temperature measurments Max T

35 Symposium Chemical Process Control 6, Tucson, Arizona, 7-12 Jan. 2001, Preprints pp Published in AIChE Symposium Series, 98 (326), pp ISBN (2002).

36 European Symposium on Computer Aided Process Engineering 11, Kolding, Denmark, May 2001, Elsevier, pp

37 Data reconcilation ”All” variables are reconciled: Flows, feed temperatures, UA-values....

38 Optimization: 2% energy reduction In service for 20 years

39 20 heat exchangers, 5 DOFs (splits), 85 flow and temperature measurements Max T Improvements

40 An RTO failure: Complete Statoil Kårstø gas processing plant Plan: 20 + distillation columns, 4 parallel trains, steam system,...

41 Alternative to Real-Time Opimization: Indirect optimization using control layer Use off-line optimization to identify controlled variables (CV): - Active constraints - Self-optimizing variables CV s Process MVs RTO MPC PID

42 Step S7. Optimization layer (RTO) Purpose: Identify active constraints and compute optimal setpoints (to be implemented by supervisory control layer) Main structural issue: Do we need RTO? (or is process self- optimizing) RTO not needed when –Can “easily” identify change in active constraints (operating region) –For each operating region there exists self-optimizing variables

43 Question From: Ruben Marti, Un. Valledolid, Spain Why not combine RTO and control in a single layer with economic cost function (N-MPC = D-RTO)? Why is this not used? What alternatives are there? CV s Process MVs RTO MPC PID