Operational Decision Support with Online Models

AVEVA World Conference North America 2019

Chris Boljkovac

Shell Canada, Scotford Complex

1Nov 2019

Agenda

❑ Scotford Complex

❑ ROMeo Systems at Scotford

❑ Motivation for Offline to Online

❑ Examples of Offline to Online (3)

▪ Motivation

▪ Implementation

▪ Benefits

❑ Challenges

❑ What’s Next?

2Nov 2019

Scotford Complex

❑ Ft. Saskatchewan, Alberta, Canada

❑ Refinery, Chemical Plants (2), Upgraders (2), and CCS

❑ 1,300+ employees

❑ 300+ Kb/d crude oils

❑ 3,000 ML/y diesel, 2,000+ ML/y gasoline, 1,000+ ML/y jet fuel

❑ 1,470+ t/d glycols, 1,370+ t/d styrene

❑ 1 mil+ t/y CO2 capture

❑ 65% of energy used by standard comparable complex

One of North America’s most efficient, modern, and

integrated hydrocarbon processing sites

3Nov 2019

ROMeo Systems at Scotford

❑ Refinery (multi-unit RTO)

▪ ADU (FCE, rigorous distillation)

▪ HCU (kinetic reactor model, rigorous distillation)

▪ CCR (kinetic reactor model, component separators)

▪ Aromatics (LP model), GRU (LP model)

❑ Upgraders (unit RTO)

▪ A&V1 (FCE, rigorous distillation)

▪ A&V2 (FCE, rigorous distillation)

❑ Chemicals

▪ Glycols (under evaluation)

4Nov 2019

Motivation

❑ Many operating strategies, safeguarding limits, and performance

investigations are developed from offline studies using a wide variety of

tools (PRO/II, UNISIM, FURN, HTRI, etc.)

❑ These studies typically required many assumptions due to process

variability and uncertainty (e.g. feed quality, equipment performance, etc.)

❑ They quickly become out-of-date

What if we could implement these calculations online?

5Nov 2019

Examples of Offline to Online

2019 - Relief Load Estimation (ADU), PRO/II > ROMeo

2016 - Relief Load Estimation (A&V1, A&V2), PRO/II > ROMeo

2016 - Heater Outlet Vaporization (ADU), FURN+HYSYS > ROMeo

2015 - Diluent Production (ADU/HEX), ROMeo > ROMeo

2012 - Preheat Hydraulics (A&V1), ROMeo+GAMS > ROMeo

2011 - Column Loading (A&V1), PRO/II > ROMeo

“Hey, you have a up-to-date process model of my unit …”

6Nov 2019

Relief Load Estimation – Background

❑ ADU operation was commonly limited by a column relief safeguarding limit

❑ This limited the processing of slops, spot crudes, etc.

❑ Basis was an extensive safeguarding study (2007) that used nominal feed

qualities to develop feed equivalency factors

❑ An offline study using ROMeo was performed (2017) using the same heat

and material balance (HMB) method but up-to-date feed compositions,

equipment performance parameters, etc.

❑ Found that we were significantly overpredicting the relief load

7Nov 2019

Relief Load Estimation – Process Model

❑ Developed process models in PRO/II and ROMeo to estimate relief load

using the current Shell simulation method

❑ Found that the over-prediction was even larger

8Nov 2019

Relief Load Estimation – Implementation

❑ Added relief load process simulation model to the ROMeo model used for

RTO

❑ Developed additional activation and model sequences

❑ Updated the “estimator model” with the most recent estimates of feed

composition (FCE), equipment performance (e.g. exchanger duties), and

current measurements of unit feed flows

❑ Added an alarm at the maximum relief capacity (with safety factor)

❑ Added an watchdog alarm (time with no update)

9Nov 2019

Relief Load Estimation – Implementation

10Nov 2019

EstimatorActivation Sequence

10 min.

EstimatorModel Sequence

ROMeo Model(RTO)

3-4+ hours

ROMeo Model(Estimator)

CopyLast Successful

Reconciled Model(FCE, duties, etc.)

Relief Load Estimation – Benefits

❑ Accurate real-time estimate of relief load at current feed composition

(assays + lab + FCE) and current operating conditions

❑ Stand-alone execution provides both robustness and accuracy (updated by

RTO model parameters, but much less reliant)

❑ High rate-of-execution results in the relief load estimate responding quickly

to feed changes (e.g. clear the alarm, ride the limit)

… but,

❑ Lost the ability to predict relief load for feed/diet changes (e.g. planning)

11Nov 2019

Diluent Production – Background

❑ Refinery intermittently produces make-up Diluent for Upgraders

❑ Increased use of imported feeds (IHB) required a significant change in

make-up Diluent quality and volume

❑ Production were unable to produce Diluent with low C5/C6 or without the

HEX unit to get/keep Diluent pool on-spec

❑ Past simulation and control studies were unsuccessful in finding a solution

❑ The ROMeo model (ADU, HEX product stream) was used offline to

investigate low C5/C6 production strategies

12Nov 2019

Diluent Production – Implementation

❑ The ADU ROMeo model was modified to include penalties for the primary

quality target (C5/C6) and limits (max. nC6, RVP, etc.)

❑ RTO used existing RTO targets in the Stabilizer and De-Isohexanizer (DIH) to

achieve the desired quality of Diluent and maximize production

❑ The HEX unit APC was modified to maximize diluent product subject to a

C7+ Combined Diluent quality constraint

Successful control scheme since HEX unit dynamics are very slow

13Nov 2019

Diluent Production – Coordinated Control

14

Stabilizer Targets

(RTO)

DIH Targets (RTO)

Combined (HEX+ADU) Diluent

HEX Diluent (APC)

Nov 2019

Diluent Production – Penalties

15Nov 2019

Diluent Production – Benefits

❑ Demonstrated offline in late-2014, online by early-2015

❑ Improved quality control (e.g. C5/C6) with simultaneous production

maximization

❑ Allows ADU-only (no HEX) Diluent production (at suitably higher C5/C6

ratios and relaxed qualities)

❑ Used to identify limiting constraints for potential relaxation

The economic benefit of a single 5-day production

run (2015) justified the project

16Nov 2019

Preheat Hydraulics – Background

❑ Preheat exchanger fouling in A&V1 (diluted bitumen, cold-side) is

significant

❑ Production opens manual by-pass valves to maintain feed flow

❑ No guidance was available on which exchangers to by-pass and by how

much?

❑ Developed a hydraulic model of preheat system (in GAMS) based on

pressure surveys, flow bypass and imbalance estimates (using outlet

energy balances), and available instrumentation

Just wanted to know what we could do offline with available data?

17Nov 2019

Preheat Hydraulics – Model

18Nov 2019

Measured

dP (Overall System)

Pre

dic

ted

Preheat Hydraulics – Offline

❑ Hydraulic model used to produce operating points with equal preheat dP’s

for combinations of E-13 and E-14 by-pass flows

❑ ROMeo model predicted a 1.5 to 5.0 C increase

❑ There were significant benefits to capture

19Nov 2019

Preheat Hydraulics – Online

❑ Estimators were built for unmeasured cold-side flows

❑ The hydraulic model was added to the online ROMeo model

❑ By-pass flow targets were added to the RTO Opt case with the constraint to

maintain current overall system dP

❑ Optimal by-pass flow targets were displayed (“manual targets”)

… problems,

❑ Bypass flows were difficult to achieve (manual valves)

❑ The hydraulic model fell out-of-date (pressure surveys required)

Online implementation was of mixed success

Need control valves and P measurements to be successful

20Nov 2019

Challenges

❑ System complexity

▪ RTO models (macro re-factoring, sub-models)

▪ New sequences (activation, execution, housekeeping)

❑ Solution robustness

▪ High availability/frequency

▪ Watchdogs (time-critical)

❑ Lack of predictive model

▪ Correlations for LP/planning tools

So far, have found a way to implement the desired system

21Nov 2019

What’s Next

❑ Plans to expand relief load estimation to additional units

❑ Re-visit AV1 hydraulic model benefits

❑ Plan to enhance ADU heater outlet vaporization implementations (e.g.

additional parameters, increased rate of execution)

❑ Plan to develop predictive relief load tool for use in planning (Phase 2)

❑ Continue to look for new applications

What is limiting our operation?

Can we improve those estimates by calculating them online?

22Nov 2019

Questions

23Nov 2019