Unique Applications for Embedded Model Predictive

Control Technology Paper 137a

James Beall

Principal Control Consultant

Emerson Automation Solutions

April 2, 2019

Speaker –James Beall

• Principal Process Control Consultant with Emerson Automation Solutions since 2001

• 37+ Years experience with Petrochemical Process Automation

• Specialist in advanced regulatory control and multivariable control

• Eastman Chemical Company, Texas Operations (1981-2001)

Introduction

•What is MPC?

•What is Embedded MPC?

•Example Uses of Embedded MPC

•Questions

“MPC”: Multivariable, Model Predictive Controller

Uses the Past to

Predict the Future

Past Present Future

Modeled

Relationship

Process Input

Process Output

“MPC”: Multivariable, Model Predictive Controller

Energy Flow

(manipulated)

Lime Mud Flow

(optimized and

disturbance)

ID Fan Speed

(manipulated)

Hot-end Temperature

(controlled)

Cold-end Temperature

(controlled)

Excess Oxygen

(controlled and constraint)

Hood Draft Pressure

(constraint)

Kiln Stack Emissions

(constraint)

Lime Kiln Process

Manipulated

Con

tro

lle

d

FC 1-3

Reboil Rate

AT 2-2

A btms comp

TC 1-2

Top Temp

AT 1-2

A ovhd comp

Con

str

ain

t

FC 1-2

Reflux rate

TT 1-4

Reflux temp

TT 1-5

Tower A temp

Disturbance

PT 1-3

Steam PressureFC 1-1

Feed Rate

DP 1-1

Delta P

FEED TGT

A ovhd comp

What is Embedded MPC?

Embedded MPC:

• NO extra databases

• NO database synchronization issues

• NO watchdog timers

• NO fail/shed logic design

• NO custom DCS programming

• NO interface programming

• NO operator interface development

Traditional APC Embedded MPC:

• Can run in DCS controllers/stations

• Redundant and fast (e.g. 1/sec)

• Integrated operator user interface

• Configuration through standard configuration tools

• Automated step testing and Model ID

• Off-line simulation and testing

• Implementation by plant control engineers

Traditional and Embedded MPC

•Traditional MPC• Well developed, full featured

technology

• Higher “minimum” project cost

• Medium to larger req’d for ROI

•Embedded MPC• Less features to “fit” in DCS

• Lower implementation cost

• ROI for smaller

MPC SizeLarge

Medium

Small

Traditional

MPC

Embedded

MPC

Examples of Unique Applications

• “Big Valve, Little Valve”

•Dead time dominant SISO

•Feed Forward and Override

•Waste heat steam generator

•Waste heat recovery – steam export

“Big Valve, Little Valve”

•PID Considerations• Large PV disturbances can

saturate Small valve – loss of control

• ZC should have “gap” control to eliminate potential Large valve limit cycles

Neutralizer

AC

1-1

AC

1-1

AT

1-1

PID Controller

Large

(Coarse)

Small

(Fine)

ZC

1-1

ZC

1-1

Integral only Controller

(CV is Implied Fine

Control Valve Position)

Reagent

CV

“Big Valve, Little Valve”•MPC Considerations

• Large PV disturbances handled by moving both Large and Small valves

• Use constraint control on small valve position to eliminate large valve limit cycles

Neutralizer

AC

1-1

AC

1-1

AT

1-1

MPC Controller

Large(Coarse)

Small(Fine)

Reagent

MPC Design

CV1 Tank Discharge pH

MV1 Small Valve

MV2 Large Valve

“Big Valve, Little Valve”

Successive Load Upsets Set Point Change Trim Valve SP Change

Critical Process Variable

Coarse Valve

Trim Valve

MPC Performance

Deadtime Dominant Loops

Tau

Dead time >= 4*Tau (definition varies)

~4*Tau

Deadtime Dominant Loop

• Reactor coolant supply temperature

• “Mixing” temperature response fast (~15 sec Tau)

• Step test revealed 80 seconds of dead time!

TC

MPC Variables

CV1 Coolant Temperature

MV1 3-Way Valve

Feed Forward with MPC

• Temperature of cooling water impacts outlet temp

• Typically implemented as feedforward to PID (calculate FFWD Gain, dead time, lead-lag

• Use TI-2 as a Disturbance Variable with MPC-simple model

TC-1

TI-2

MPC Variables

CV1 Coolant Temperature

MV1 3-Way Valve

DV1 Cooling Water Temp

Override/Constraint with MPC

• Suppose there is a minimum flow through exchanger tubes, FI-3, to avoid fouling

• Typically done with 2 PID’s and a hi/lo selector

• Implement as a Constraint Variable (LV) with MPC

TC-1

FI-3

TI-2

MPC Variables

CV1 Coolant Temperature

CV2 Minimum Tube Flow

MV1 3-Way Valve

DV1 Cooling Water Temp

Optimize Small Applications

TCJC

Fractionator

PA

Steam

BFW

TC

Debutanizer

Duty

BTU/hr

Goal is to maximize steam production while ensuring sufficient heat for

debutanizer reboiler. Also, handle limited BFW supply.

FI

PI

Optimize Small Applications

Objective: Maximize steam production

Implementation time: 2 days

Benefit: $120K/year

Steam

BFW

MPC

MV

JCDuty

BTU/hr

CV

PI

FI

CV

FI

Heat Recovery

Steam

Steam Drum

Condensate Drum

Primary

Secondary

Fin-Fan Cooler

350 PSIG Steam Header

150 PSIG Steam Header

150 PSIG

Steam Users

Steam Export

SIC SIC

Reactive Distillation

Column

BFW

Heat Recovery

150

PSIG

Steam

Users

Steam

Steam Drum

Condensate

Drum

Primary

Secondary

Fin-Fan Cooler

350 PSIG Steam Header

150 PSIG Steam Header

Steam Export

SIC SIC

Reactive

Distillation

Column

BFW

MPC Variables

CV1 Steam Drum Pressure

CV2 Cond. Drum Pressure

CV3 Steam Export Pressure

CV4 350-150 PRV Position

MV1 Steam Export Valve

MV2 Steam Valves to Cooler

MV3 Total Fan Speed

Heat Recovery• Plant personnel learned about MPC application selection,

MPC design and commissioning

• Identified a steam valve that had excessive leakage at shutoff

• Increase average steam export ~3000 lb/hr = $240K/year, repair of leaking steam valve $200K/yr benefit

• Benefit• Implementation time – 2 weeks• Benefit – $440K/yr• Payback period – 3 weeks

Summary• Traditional MPC is well developed, full function technology but

requires a high minimum investment requiring larger projects to provide ROI

• Embedded MPC has less features but is easy to implement, low implementation cost and executes fast resulting in smaller projects to provide ROI

• Each technology has a “best fit”, so utilize each technology where they provide the best ROI

Questions