Surrogate-Based Process Optimization: A Case Study on ...Makeup/Blowdown RCYL-4 Lean/Rich Exchanger VLVE-100 Sour gas 13 Sweet gas Rich Amine 6 Lean Amine 8 10 11 Q-Pump Makeup Blowdown

Proceedings of the International Conference on Industrial Engineering and Operations Management Dubai, UAE, March 10-12, 2020

© IEOM Society International

Surrogate-Based Process Optimization: A Case Study on Simple Natural Gas Processing Plant

Falah Alhameli1, Mohammed Alkatheri1, Ali Elkamel1,2, Ali Almansoori2, and Peter Douglas1

1Department of Chemical Engineering, University of Waterloo, Waterloo, Ontario, N2L 3G, Canada

2Department of Chemical Engineering, Khalifa University of Science and Technology, SAN Campus, P. O. Box 2533, Abu Dhabi, United Arab Emirates

[email protected], [email protected], [email protected], [email protected], [email protected]

Abstract

Natural gas is a vital component of the world's energy supply. Therefore, optimizing natural gas processes is important to facilitate more efficient and less costly operations. Optimization which relies exclusively on simulation is impractical, due to the enormous computational cost of complex simulations. Alternatively, surrogate models can be constructed from and used in lieu of actual simulation models. The proposed work aims to develop a black-box surrogate-based optimization model that represents the natural gas treatment process which consist of separation, sweetening and dehydration units. A combination of simulation, experimental design, data analytics and mathematical programming was utilized. The problem size was reduced by using a regression model of the system’s behavior. The regression model was based on the simulation and experimental design. This simplified model was then used in a mathematical programming model minimizing a cost function. The methodology is suitable for synthesis of large-scale complex problems with fewer degrees of freedom. Linear regression along with polynomial feature transformation machine learning methods were used to generate the surrogate model. A case study was developed to investigate the impact of increasing H2S composition in the feed gas. It was observed that the feed pressure had the highest influence among the parameters.

1. Introduction

The design and operation of natural gas processing facilities tailored for the production of various natural gas derivatives and pretreatment is considered to be one of the fundamental problems in chemical and process systems engineering. However, planning and superstructure-based mathematical models of any chemical process involve many complex unit equation blocks which can be solved using conversion law or physicochemical engineering fundamentals and available simulation software (Henao and Maravelias, 2011) . Despite the significant developments in realistic unit operation models (i.e. the kind of models featured in commercial process simulators considering non-ideal thermodynamics, kinetics, and transport properties calculations and the availability of commercial process simulators (Gani et al. 2012) (e.g. ASPEN, HYSYS, ProMax, gProms), however, these models cannot be used for within optimization operations (e.g. supply chain management and planning models) due to computational expenses. Data-driven surrogate modelling can be defined as a black-box modelling approach that can utilize available data (e.g. could be collected from commercial software), that can relate relevant inputs to relevant outputs to describe process operations. Such models have been used in the industry and the literature to describe processes by replacing existing expensive models (serve as surrogates to reduce model complexity) and correlations which have not yet been theoretically explained (Li et al. 2016). For example, Jin and et al. in 2000 investigated the advantages and disadvantages of four metamodeling. In 2002, Hetzel replaced some the detailed unit models equality constraints with surrogate models and results showed that the computation time for the refinery-wide optimization with surrogate replacement was reduced by one order of magnitude. Henao and Maravelias integrated surrogate models built from data generated via commercial process simulators into optimization superstructure-based framework. Another recent study (Li et al. 2016). developed a data-driven optimization framework for the production of refinery-petrochemical complex where product yields and properties of a petrochemical complex were predicted based on the abundant data.

1485

mailto:[email protected]:[email protected]



B Given the recent developments of data analytics tools such as big-data analytics (e.g. machine learning and data mining) and the need for inexpensive nonlinear models which can relate relevant inputs to relevant model outputs to represent process operations, the role of data driven surrogate modelling can be extremely valuable. Another reason that thrusts this research to use the big-data analytics tools (i.e. machine learning), is that these tools have proven their ability to generate accurate and computationally efficient surrogate or reduced models (Qin 2014) Thus, applying the data-driven surrogate modelling approach in an optimization framework will reduce the mathematical complexity of the entire optimization framework model and impose a suitable mathematical representation which can be solved by current state-of-art numerical solvers (Henao and Maravelias, 2011). There are still many process system engineering applications which have not yet been studied where surrogate-based data-driven model can replace the constant yields or complex unit equations in optimization frameworks. For example, there are many models for petroleum refining developed by scholars in the last decades (Papageorgiou 2009), whereas, there is a lack in the application surrogate model in natural gas process optimization. These models help to assess and optimize the processes to meet the economic and environmental requirements. Therefore, The proposed work aims to develop a model for gas processing. The rest of this paper is structured as follows: Section 2 states the methodology of the proposed method. Section 3 discusses the results of applying the proposed methodology on a natural gas plant case study. Finally, onclusions are drawn in section 4. 2. Methodology

In this paper, a case study on minimizing overall energy consumption of natural gas process under different scenarios of H2S feed composition while meeting the water content specification of the product gas. A combination of simulation, experimental design, data analytics tools (i.e. machine learning) and mathematical programming frameworks were developed. The conducted case study was carried out for particularly three main units namely; Condensate Stabilizer, Acid gas removal using MDEA(Methyl-Di-Ethanol-Amine) and Dehydration Process using TEG (Tri-Glycol-Ethylene). Figure 1-3 show the schemes of these three processing unit. Table 1 shows the selected process (input) and its range and the performance (output) variables relevant for the current case study.

Table 1. List of identified variables Process Variables (predictors) - X Performance Variables (response) - Y

The temperature for the Sour gas cooler in the separation unit (Condensate Stabilizer). The range is 30-40 oC (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶)

The total energy consumption for the separation condensate Stabilizer and amine sweeting unit ( 𝑠𝑠𝐶𝐶𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝐶𝐶 ) and dehydration unit (𝐶𝐶𝑇𝑇𝑇𝑇)

The temperature for the feed gas in the separation unit. The range is 50-80 oC. (𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠)

Final water content (𝑊𝑊𝑠𝑠𝑊𝑊𝐶𝐶𝐶𝐶)

The pressure for the feed gas in the separation unit. The range is 45-65 bar (𝑃𝑃𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠𝑃𝑃𝐶𝐶𝐶𝐶)

flowrate for the feed gas in the separation unit. The range is 100-600 MMscfd. (𝐹𝐹𝐶𝐶𝐶𝐶𝐹𝐹𝐶𝐶𝑠𝑠𝑊𝑊𝐶𝐶)

Gallons of potentially recoverable liquid hydrocarbons (C2 and heavier) at 60 oF Per 1000 standard cubic feet (Mscf) in the feed gas. The range is 8-11 GPM. (𝑇𝑇𝑃𝑃𝐺𝐺)

The H2S molar composition in % for the feed gas in the separation unit. The range is 0.5-4 % (𝐻𝐻2𝑆𝑆)

The CO2 molar composition in % for the feed gas in the separation unit. The range is 1-5% (𝐶𝐶𝑂𝑂2)

TEG Reboiler Temperature (Reboiler T): is the Reboiler temperature (oF) for the stabilizer in the TEG unit. The range is 380-400 oF. (𝑅𝑅𝐶𝐶𝑅𝑅𝐶𝐶𝑠𝑠𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶)

TEG Feed Temperature (TEG Feed T): is the temperature (oC) for the Sweet gas cooler in the Amine (Acid Gas Removal) unit. The range is 16-38 oC. (𝐶𝐶𝑇𝑇𝑇𝑇𝐹𝐹𝐶𝐶𝐶𝐶𝑇𝑇𝐶𝐶)

TEG Rate: is the ratio of TEG gallons per pounds of water in the gas fed to the TEG unit. The range is 2-5 𝐶𝐶𝑇𝑇𝑇𝑇𝑅𝑅𝑠𝑠𝑊𝑊𝐶𝐶

1486



The generation of simulation samples were based on Design of Experiment method. Box-Behnken design was used for this study to generate the input samples for simulation runs. It has been demonstrated that it performs more efficiently than central composite design and much more efficient than the three-level full factorial designs (Ferreira 2007). The simulation runs for input output variable generation had been executed successfully using ProMax® 3.2. It was chosen among the available software’s because it is suited for gas processing and has a modern model for sweetening unit. Supervised Machine Learning tools (i.e. linear regression and polynomial features generation ‘Supervised learning’) were used to generate surrogate models (i.e. reduced order model) that replace the complicated detailed unit equations in the chemical process optimization framework. Such data-driven models aim to estimate output variables given the input process variables. The simulation data (input-output variables /Process – Performance variables) were analyzed and used to generate regression model (surrogate) using Scikit-learn machine learning library in Python. After that, the developed regression model was integrated into optimization framework where its objective to minimize the total energy consumption while increasing the H2S concertation. Therefore, a data-driven surrogate-Based optimization model was constructed to find the optimal values for the variables corresponding to the minimal total energy.

Feed

Slug catcher

SAT-11

2

3

4

M-P Flash

5

6

Water

8

9

10

Reboiler12

13

Condenser14

15

Stabilizer

20

1

10

Q-Condenser

Q-Reboiler

16

Off gas

18

Q-Condensate Cooler

CMPR-2nd Stage22

Q-2nd Stage Compressor

2nd Stage Cooler

23

2nd Stage Flash

24

25

RCYL-1

26

M-P mixer

27

CMPR-1st Stage28

Q-1st Stage Compressor

1st Stage Cooler

29

Q-1st Stage Cooler

1st Stage Flash

30

31

33

L-P mixer

34

MIX-102

35

MIX-103

37

Sour gas Cooler

36

Q-Sour gas Cooler

Sour gas Flash

Sour gas

39RCYL-3

40

Q-2nd Stage Cooler

PreHeater

Q-PreHeater

Condensate

L-P Flash

Condensate Cooler

Q-Stabilizer Feed Heater

Stabilizer Feed Heater

RCYL-2

XFS1

XFS2

HC Liquids

XFS4

Water from AmineXFS5

Water from TEG

Water mixer

Sour Water

XFS6

Flash gas from Amine

XFS7

Flash gas from TEG

Flash mixer

Flash gases

Figure 1. Separation unit simulation

1487



Absorber

7

1

Cooler

Circulation Pump

Makeup/Blowdown

RCYL-4

Lean/Rich Exchanger

VLVE-100

Sour gas

13

Sweet gas

Rich Amine

6

Lean Amine

8

1011

Q-Pump

Makeup

Blowdown

Q-Cooler

9

Reboiler1

2

Condenser3

4

Regenerator

10

1

2

Q-Condenser

Q-Reboiler

Acid Gas

12

XCHG-101Steam In Steam Out

Rich Flash

18

Flash Gas

Sweet gas cooler

14

Q-Sweet gas cooler

Sweet gas Flash

HC Liquids

Wet gas

Water

XFS1

XFS2RCYL-5

5

XFS3

XFS4

XFS6

Figure 2: Simulation of Amine unit

Reboiler20

3

Condenser

1011

Glycol Regenerator

4

1

2

Q-101

Reboiler Q

Glycol Contactor

2

1TEG Makeup

Q-Recycle

Q

TEG Recycle

Reflux Coil

Gas/Glycol HEX

Cross Exchanger

Glycol Pump

Wet Gas Feed

TEG

5 6

8

9

Lean TEG

15

16171819

4 Dry Gas

Flash Gas

Water Vapor

Q-100

Pump HpBlowdown

Flash drum

XFS3

XFS5

XFS7

Figure 3: TEG unit simulation

3. Result and Discussion

1488



In this case study, the main target is to minimize the total for the three-processing plant energy. The total energy of the three plant has been divided into two parts: 1) TEG, which is total energy for TEG unit. This is because its value is very small compared to the separation or amine units so it is usually rolled out unless analyzed separately. 2) Sepamine, which is the total energy for the separation and amine units. This is because they have common cross streams. The total energy is the summation of TEG and Sepamine. The splitting treatment is very crucial for error reduction. Figure 4-show the normal probability plot and a comparison between the fitted and simulated values for sepamine, TEG and water content in the gas product respectively. Since the data almost follow a straight line for all normal probability plots, the assumption of normal distribution is validated. Moreover, Figure 5, 7, and 9 show good matching between the fitted and simulated values so the models for sepamine, TEG and Final water content are representative. It is worth mentioning that all models have R2 values of 0.95 or higher. The objective of conducting this case study is to illustrate the surrogate model application into natural gas process optimization. The increase of H2S composition in the natural gas is reported due to the presence of micro bacteria [9]. Therefore, it is important to investigate the effect H2S increase for various GPM. In this optimization problem the final water content was set to be 4 lb/MMscf. Moreover, the flowrate and CO2 composition were kept constant (flowrate = 350 MMscfd and CO2 = 3%). The chosen H2S compositions for the study are 0.5, 2.25 and 4 % with 8, 9.5 and 11 GPM so the total number of runs was 9. The optimization model was executed using GAMS. Following is the optimization model:

𝑠𝑠𝑠𝑠𝑠𝑠 𝑠𝑠𝐶𝐶𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝐶𝐶 + 𝐶𝐶𝑇𝑇𝑇𝑇 Subject to the following constraints

𝐶𝐶𝐶𝐶𝑊𝑊𝑠𝑠𝐶𝐶 = 𝑠𝑠𝐶𝐶𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝐶𝐶 + 𝐶𝐶𝑇𝑇𝑇𝑇 𝑠𝑠𝐶𝐶𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝐶𝐶 = 𝑓𝑓(𝑥𝑥)

𝐶𝐶𝑇𝑇𝑇𝑇 = 𝑔𝑔(𝑥𝑥) 𝑊𝑊𝑠𝑠𝑊𝑊𝐶𝐶𝐶𝐶 = ℎ(𝑥𝑥)

𝑊𝑊𝑠𝑠𝑊𝑊𝐶𝐶𝐶𝐶 ≤ 4 (𝐹𝐹𝑠𝑠𝑊𝑊𝐶𝐶𝐶𝐶𝑤𝑤𝐶𝐶𝑠𝑠𝑊𝑊𝐶𝐶𝑠𝑠𝑊𝑊 𝐶𝐶𝐶𝐶𝑠𝑠𝑊𝑊𝐶𝐶𝑠𝑠𝑤𝑤𝑊𝑊𝑠𝑠𝐶𝐶𝑠𝑠) 30 ≤ 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 ≤ 40

50 ≤ 𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠 ≤ 80 45 ≤ 𝑃𝑃𝐶𝐶𝐶𝐶𝑠𝑠𝑠𝑠𝑃𝑃𝐶𝐶𝐶𝐶 ≤ 65

𝑇𝑇𝑃𝑃𝐺𝐺 = 8, 9.5, 11 (𝐶𝐶𝐶𝐶 𝑅𝑅𝐶𝐶 𝑤𝑤ℎ𝑠𝑠𝑠𝑠𝑔𝑔𝐶𝐶𝑇𝑇 𝑓𝑓𝐶𝐶𝐶𝐶 𝐶𝐶𝑠𝑠𝑤𝑤ℎ 𝐶𝐶𝑃𝑃𝑠𝑠) 𝐻𝐻2𝑆𝑆 = 0.5, 2.25, 4 (𝐶𝐶𝐶𝐶 𝑅𝑅𝐶𝐶 𝑤𝑤ℎ𝑠𝑠𝑠𝑠𝑔𝑔𝐶𝐶𝑇𝑇 𝑓𝑓𝐶𝐶𝐶𝐶 𝐶𝐶𝑠𝑠𝑤𝑤ℎ 𝐶𝐶𝑃𝑃𝑠𝑠)

𝐶𝐶𝑂𝑂2 = 3 380 ≤ 𝑅𝑅𝐶𝐶𝑅𝑅𝐶𝐶𝑠𝑠𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 ≤ 400

16 ≤ 𝐶𝐶𝑇𝑇𝑇𝑇𝐹𝐹𝐶𝐶𝐶𝐶𝑇𝑇𝐶𝐶 ≤ 38 2 ≤ 𝐶𝐶𝑇𝑇𝑇𝑇𝑅𝑅𝑠𝑠𝑊𝑊𝐶𝐶 ≤ 5

Where 𝑓𝑓(𝑥𝑥), 𝑔𝑔(𝑥𝑥) and ℎ(𝑥𝑥) are the surrogate models generated from regression for 𝑠𝑠𝐶𝐶𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝐶𝐶, 𝐶𝐶𝑇𝑇𝑇𝑇 and 𝑊𝑊𝑠𝑠𝑊𝑊𝐶𝐶𝐶𝐶 response variables respectively and 𝑥𝑥 are the process variables (See Table 1). The following Table 2 shows the optimization results. The total energy increases almost linearly with H2S increase as shown in Table 2 and Figure 10. This is due to the increase of circulation rate of amine to treat that increase of H2S. Moreover, as it can be noticed in Figure 10 the total energy increases with GPM increase. This is because more compression and cooling duties in the separation unit are required in presence of more heavy hydrocarbons. Furthermore, From Table 2 it can be seen that the optimal values for Cooler T and Temp are the lower bounds because the absorption (amine and TEG) is favoured at low temperature. However, theses temperatures are usually limited by hydrate formation and heavy hydrocarbons condensation. Figure 11 shows the variation of the optimal pressure as a function of H2S content for different GPM content. As noticed, the optimal pressure increases with H2S increase. This is explained with the aid of Figure 12 (total energy for separation and amine units Vs. pressure). As the pressure increases, the total energy of separation and amine units decreases. This is because increasing the pressure enhances the stripping of gaseous species in the slug catcher resulting in decreasing the flowrate of the liquid stream from the slug catcher. Consequently, the duties for compression and cooling drop. In addition, the absorption process within amine unit is easier when pressure is higher. Therefore, the amine circulation rate decreases results in less the total energy for amine unit. However, pressure rising beyond certain range increases the will increase energy consumption. That can be clearly seen in Figure 12 for the pressure range of 56-65 bar. Also, from Table 2, It is worth mentioning that the optimal pressure increases when GPM increases for the same H2S composition. This is because increasing GPM corresponds to different phase envelopes for the feed gas. As shown in Figure 13, the optimal TEG feed temperature is proportionally increasing with H2S

1489



concentration. However, TEG feed temperature decreases when the concentration of H2S is 4 % and GPM is11 because the pressure increase is not enough to meet the water specification. Conclusion A surrogate-based optimization framework was applied to a natural gas processing plant. The gas processing plants consists of separation, sweetening and dehydration units. The objective of this case study was to minimize the energy consumption while meeting pipeline gas quality. Box-Behnken design was adopted to generate simulation samples. ProMax® 3.2 was used to simulate the plant. Multiple linear regression combined with polynomial feature transformation was used to generate the surrogate models. The developed regression model was integrated into mathematical programing model along with desired objective and imposed operating conditions constraint. The chosen variables for the model were 10. They represent the gas feed conditions and units’ parameters. The total runs were 170. They were successfully implemented and analysed. Total energy of the plant and water content for the product gas models were obtained. This case study was conducted to investigate the impact of H2S composition increase in the feed gas. The models were used for the case study with the objective of total energy minimization and constraint of 4 lb/MMscf for water content in the product gas. It was noticed that the feed pressure had the highest influence among the other parameters. Table 2: Optimal variables for different H2S with various GPM

Fixed Variables Optimal Variables

Run #

Flowrate (MMscfd)

GPM H2S (%)

CO2 (%)

Pressure (bar)

Cooler T (oC)

Temp (oC)

Reboiler T (oF)

TEG Feed T (oC)

TEG Rate

Total (MMBtu/h)

1 350 8 0.5 3 45 30 50 400 35.48 5 263.405

2 350 8 2.25 3 48.819 30 50 400 36.59 5 519.897

3 350 8 4 3 53.891 30 50 400 38 5 764.753

4 350 9.5 0.5 3 45.91 30 50 400 35.75 5 329.971

5 350 9.5 2.25 3 50.999 30 50 400 37.20 5 573.839

6 350 9.5 4 3 56.404 30 50 400 38 4.46 809.382

7 350 11 0.5 3 48.058 30 50 400 36.37 5 391.819

8 350 11 2.25 3 53.212 30 50 400 37.81 5 626.237

9 350 11 4 3 60.353 30 50 400 37.49 3.65 851.655

200

300

400

500

600

700

800

900

0 1 2 3 4 5

Tota

l Ene

rgy

(MM

Btu/

h)

H2S Composition (%)

GPM = 8GPM = 9.5GPM = 11

40

45

50

55

60

65

0 1 2 3 4 5

Pres

sure

(bar

)

H2S Composition (%)

GPM = 8GPM = 9.5GPM = 11

1490



Figure 10: Optimal total energy Vs. H2S

Figure 11: Optimal pressure Vs. H2S

Figure 12: Total energy for separation and amine units Vs Pressure for H2S = 4%, Cooler T = 30 oC, Temp = 50 oC, Flowrate = 350 MMscfd, GPM = 9.5,

CO2 = 3%, TEG Feed T = 38 oC

Figure 13: Optimal TEG Feed T Vs. H2S

806

808

810

812

814

816

818

820

45 50 55 60 65

Tota

l Ene

rgy

(MM

Btu/

h)

Pressure (bar)35

36

37

38

39

0 1 2 3 4 5

TEG

Feed

Tem

p (o

C)

H2S Composition (%)

GPM = 8GPM = 9.5GPM = 11

1491



403020100-10-20-30

99.9

99

9590

80706050403020105

1

0.1

Residual

Perc

ent

Normal Probability Plot(response is Sep+Amine)

Figure 4: Normal probability plot of total energy for

separation and amine units

Figure 5: Comparison between fitted and

simulated values of total energy for separation and amine units

0.20.10.0-0.1-0.2-0.3

99.9

99

9590

80706050403020105

1

0.1

Residual

Perc

ent

Normal Probability Plot(response is TEG)

Figure 6: Normal probability plot of total energy of

TEG unit


simulated values of total energy for TEG unit

0.50.40.30.20.10.0-0.1-0.2-0.3-0.4

99.9

99

9590

80706050403020105

1

0.1

Residual

Perc

ent

Normal Probability Plot(response is Final water content (lb/MMscf))

Figure 8: Normal probability plot of water content in

the product gas


simulated values of water content in the product gas

y = 0.9998xR² = 0.9992

0

450

900

1350

1800

0 450 900 1350 1800

Fitt

ed v

alue

s (M

MBt

u/h)

Simulated values (MMBtu/h)

Sepamine

y = 0.9981xR² = 0.9938

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 1 2 3 4 5

Fitt

ed v

alue

s (M

MBt

u/h)

Simulated values (MMBtu/h)

TEG

y = 0.9992xR² = 0.9956

0

1

2

3

4

5

6

7

8

0 2 4 6 8

Fitt

ed v

alue

s (lb

/MM

scf)

Simulated values (lb/MMscf)

Final water content

1492



References Henao C. A. and Maravelias C. T., "Surrogate-based superstructure optimization framework," AIChE Journal, vol.

57, pp. 1216-1232, 2011. Gani R., Cameron I., Lucia A., Sin G., and Georgiadis M., "Process Systems Engineering, 2. Modeling and

Simulation," in Ullmann's Encyclopedia of Industrial Chemistry, ed. 2012 A. Technology, "Aspen Plus 11.1," ed, 2001. "ProMax®," ed. Texas, USA: Bryan Research & Engineering. "GProms," ed: Enterprise, Systems, Process. Qin S. J., "Process data analytics in the era of big data," AIChE Journal, vol. 60, pp. 3092-3100, 2014. Li J., Xiao X., Boukouvala F., Floudas C. A., Zhao B., Du G., et al., "Data-driven mathematical modeling and global

optimization framework for entire petrochemical planning operations," AIChE Journal, vol. 62, pp. 3020-3040, 2016.

Jin R., Chen W., and Simpson T., "Comparative studies of metamodeling techniques under multiple modeling criteria," in 8th Symposium on Multidisciplinary Analysis and Optimization, ed: American Institute of Aeronautics and Astronautics, 2000

Matthew Joseph Hetzel B. S., "Refinery-wide optimization using neural network surrogate models," Master of science Dissertation Chemical engineering, Texas Tech University, 2002.

Papageorgiou, L. G., "Supply chain optimisation for the process industries: Advances and opportunities, "Computers and Chemical Engineering, 33 (2009 1931–1938

Ferreira, S.L.C. and et al, "Box-Behnken design: An alternative for the optimization of analytical methods, " Analytica Chemical Acta, 597 2007 179-186

Biographies Mohammed Alkatheri holds a BSc degree in Chemical Engineering from United Arab Emirates University, and MSc degree in Chemical Engineering from the Petroleum Institute in Abu Dhabi. During his MSc, he developed research on Modelling and simulation of kinetics and single particle growth for the heterogeneous polymerization of Ziegler-Natta catalyst. From 2015 – 2017, he worked as a research assistant at the Petroleum Institute where he studied the economics of different ultra-sour natural gas sweetening processes, assessed sweeting of ultra-sour natural gas using hybrid process and carried out green-house gases life cycle assessment for the United Arab Emirates electricity sector. In May 2017, he joined PhD program in Chemical Engineering at University of Waterloo. His PhD research is focusing on the application and integration of big-data tools (i.e. Artificial Intelligence and Machine Learning) in chemical process optimization and process system engineering. The scope of his PhD project is to address the challenges associated with chemical engineering process design and operation, namely, uncertainty handling, parameter estimation and unit process equation complexity. Therefore, high-level optimization tasks such as planning and scheduling will highly benefit from information mined from massive data, since optimization has always been based on the interchange between models and data. Falah Alhameli is currently a research and development engineering at Abu Dhabi national Oil Company (ADNOC). He earned a B.S. and an M.S. in Chemical Engineering from the Petroleum Institute (now part of Khalifa University of Science & Technology) and a PhD from the University of Waterloo. He has published journal and conference papers. Dr Alhameli has completed research projects related to gas processing, planning of power production, and renewable energy integration in gas and oil industry. His current research interests focus on big data analytics and integration in multiscale decision making in oil and gas operations. Ali Elkamel is a Professor of Chemical Engineering. He holds a BSc in Chemical Engineering and BSc in Mathematics from Colorado School of Mines, MSc in Chemical Engineering from the University of Colorado-Boulder, and PhD in Chemical Engineering from Purdue University – West Lafayette, Indiana. His specific research interests are in computer-aided modelling, optimization and simulation with applications to energy production planning, carbon management, sustainable operations and product design. Professor Elkamel is currently focusing on research projects related to energy systems, integration of renewable energy in process operations and energy production systems, and the utilization of data analytics (Digitalization), machine learning, and Artificial Intelligence (AI) to improve process and enterprise-wide efficiency and profitability.

1493



Prof. Elkamel supervised over 90 graduate students (of which 35 are PhDs) and more than 30 post-doctoral fellows/research associates. Among his accomplishments are the Research Excellence Award, the Excellence in Graduate Supervision Award, the Outstanding Faculty Award, the Best teacher award, and the IEOM (Industrial engineering and Operations Management) Outstanding Service and Distinguished Educator Award. He has more than 280 journal articles, 141 proceedings, and 33 book chapters. He is also a co-author of four books; two recent books were published by Wiley and entitled Planning of Refinery and Petrochemical Operations and Environmentally Conscious Fossil Energy Production. Ali Almansoori is Professor of Chemical Engineering at Khalifa University in Abu Dhabi. During his profession, Dr. Almansoori held several administrative positions including: the Coordinator of President’s Duties, Dean of Engineering, and Chair and Deputy Chair of the Chemical Engineering Department. He also was the Interim Senior Vice President for Academic Affairs during the merge between PI, Masdar Institute, and Khalifa University of Science, Technology, and Research. His main research interest is in the area of Process Systems Engineering with the focus on energy systems design, simulation, modelling and optimization. He also conducts general research in the area of renewable energy and fuel cell technology with applications to the oil and gas industry. He has published numerous articles in renowned refereed journals and conference proceedings. He also delivered several presentations in international conferences and is the author of a few book chapters. Furthermore, he serves as a reviewer for reputable international journals in the area of energy and process systems. Peter Douglas is the Associate Dean of Engineering (Undergraduate Studies) and a Professor of Chemical Engineering at the University of Waterloo. He was previously the Director of the University of Waterloo United Arab Emirates Campus in Dubai from 2009 to 2013, the Associate Dean of Engineering (Computing), and the Associate Dean of Engineering (Graduate Studies). Professor Douglas was a founding member of WISE the Waterloo Institute for Sustainable Energy at UWaterloo. His primary research area of interest is in the development and application of PSE technology to industrial processes including process modelling, simulation, control and optimization. He is currently working on simulation and optimization issues related to the mitigation and capture of carbon dioxide from large scale emitters. Professor Douglas has consulted on a world-wide basis for many clients and has worked in Canada, Australia, Malaysia, Thailand, the UAE. Additionally, he is a co-inventor of the Dryer Master online measurement and control systems for the food processing industry; such systems are finding widespread use in Canada, USA, Europe and Asia. In addition to his research work, Professor Douglas has co-authored more than 200 related research publications and has supervised more than 80 postgraduate students

1494

ConclusionReferencesBiographies

Documents

Surrogate-Based Process Optimization: A Case Study on ...Makeup/Blowdown RCYL-4 Lean/Rich Exchanger VLVE-100 Sour gas 13 Sweet gas Rich Amine 6 Lean Amine 8 10 11 Q-Pump Makeup Blowdown