Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Energy 203 (2020) 117770

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

On the optimization of water-energy nexus in shale gas network


under price uncertainties
Doris Oke a, Rajib Mukherjee b, Debalina Sengupta b, Thokozani Majozi a, *,
Mahmoud El-Halwagi b, c
a
School of Chemical and Metallurgical Engineering, University of the Witwatersrand, 1 Jan Smuts Avenue, Braamfontein, Johannesburg, 2000, South Africa
b
Gas and Fuels Research Center, Texas A&M Engineering Experiment Station, College Station, TX, 77843, USA
c
Chemical Engineering Department, Texas A&M University,College Station, TX, 77843-3122, USA

a r t i c l e i n f o a b s t r a c t

Article history: This study develops a framework for water-energy nexus optimization in shale gas production and
Received 27 September 2019 distribution network under uncertainty. Sustainable design of the network is achieved through treat-
Received in revised form ment of wastewater using thermal membrane distillation, whereby a comprehensive design model is
21 March 2020
integrated within the network to account for energy requirement of the unit. The proposed model also
Accepted 1 May 2020
Available online 6 May 2020
accounts for the problem of scheduling hydraulic fracturing using a continuous time formulation. Various
uncertainties are associated with the network. Among different uncertain variables, uncertainties
associated with price and demand are crucial as they can affect the optimal configuration significantly.
Keywords:
Shale gas
Incorporation of uncertainty in the model seeks to meet the demand of natural gas consumers whilst
Stochastic optimization accounting for the uncertainties associated with the price of final products. Uncertainties are modelled
Scheduling via randomly generated scenarios using beta distribution of the price generated using historical data. The
Uncertainty stochastic model is applied to a case study through the maximization of net profit. Three different
scenarios are considered for analysis. Results from the three different scenarios show 14.39%, 11.49%, and
12.34% increase in profit respectively as compared to the deterministic approach. Solving all the sce-
narios together gives an ensemble-average solution with 13.74% increase in expected profit. Savings in
the freshwater requirement for fracturing and the energy associated with water management amounted
to 23.2% and 42.7%, respectively.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction encompasses different integrated processes. Among these inte-


grated processes are wellpad drilling and fracturing, water acqui-
The increase in energy demand globally which is due to sition and wastewater management, gas separation, transportation,
industrialization and technological advancement has brought storage and distribution to various markets and end users [7e10].
about a rise in the utilization of natural gas worldwide [1,2]. In the Therefore, in designing the network, consideration must be given
US, natural gas production to meet energy demand comes mostly to integrate these processes in a unified and optimal manner for full
from shale resources through horizontal drilling and hydraulic economic benefit. In addition, attached to most of these integrated
fracturing combined technologies [3e5]. Of the overall dry gas processes are inherent uncertainties, which can have a significant
production in the US in 2017, production from shale resources is effect on the overall process design and add more complexity to the
estimated to be 60% [6]. process [6]. Among these are the uncertainty related to acquisition
Given the impact of shale resources on the natural gas produc- of water for fracturing and flowback water production after hy-
tion growth, special interest has been on how the process can be draulic fracturing, uncertain estimated of recovery, and the uncer-
coordinated to maximize full economic potential. The shale gas tain market demand and price of final product. However, the
production to distribution comprises a complex network, which uncertainty related to the demand and prices of final product can
influence the total economic performance of the whole supply
chain network in a noticeable way [6]. Therefore, it is of paramount
* Corresponding author. importance to account for uncertainties involved in designing an
E-mail address: thokozani.majozi@wits.ac.za (T. Majozi).

https://doi.org/10.1016/j.energy.2020.117770
0360-5442/© 2020 Elsevier Ltd. All rights reserved.
2 D. Oke et al. / Energy 203 (2020) 117770

optimal shale gas supply network for full economic benefit. A network system that integrates multiple processes under
A considerable number of research have been done in literature uncertainty are generally solved by solving all scenarios simulta-
to address the complexity associated with shale gas production neously where the scenarios are weighted individually by the
network using mathematical optimization approach [7e16]. How- probability density of that scenario [23,24]. In the present work, we
ever, most of these literature did not give consideration to uncer- have defined each scenario comprising planning, scheduling, pro-
tainty. In light of this, Yang et al. [17] proposed a two-stage duction and distribution of shale gas from the set of wellpads for
stochastic model that considered the uncertainty related to water the entire time horizon and the probability of each scenario are
acquisition in hydraulic fracturing activities. Gao and You [18] equally likely. Three such scenarios are generated and each scenario
accounted for the uncertainty in relation to the estimated ultimate is solved individually as well as together to obtained the expected
recovery. Lira-Barraga n et al. [19] formulated a mathematical value. Result from each scenario as well as the expected value is
model that account for the uncertainty in freshwater requirements compared with that from the deterministic solution.
for hydraulic fracturing as well as flowback water generation.
Chebeir et al. [20] formulated a mathematical relation that accounts
for the market uncertainties. Chen et al. [21] captured the uncer-
tainty related to the life cycle economic and environmental per- 2. Problem statement
formance of shale gas supply chain using a multicriteria decision
making approach. Gao et al. [6] proposed a two-stage optimization Given are:
that account for both market demand uncertainties and estimated
ultimate recovery. The proposed model is based on data-driven  Sets of wellpad S to be stimulated through hydraulic fracturing
robust optimization approach. Recent study by Al-Aboosi and El- with water requirement supplied through freshwater from any
Halwagi [22] considered an approach that integrate different sys- of the given two sources in which the supply from one of these
tems operating under uncertainty to account for water-energy sources is uncertain or reused water from the storage tank. The
nexus associated with water management on sites where shale cost of withdrawal and transportation (piping/trucking) based
gas/oil is produced. The proposed approach combined solar energy on the distance between the sources and the wellpads are
and fossil fuels to promote the design sustainability of the system known.
while accounting for the stochasticity associated with these energy  A network of treatment technology R in which the flowback
sources. Probabilities are appointed based on discrete approxima- water generated fortnight after fracturing is treated to condi-
tions of continuous distribution for finite number of probabilistic tions acceptable for reuse.
scenarios generation. Three scenario tree modes using three-point  A set of processing plant P where the produced gas is trans-
Pearson-Turkey approximation technique was used to represent ported for major separation to take place. At the processing
the uncertainties. Despite the usefulness of the aforementioned plant, the NGL is separated from natural gas and sent to the
studies, uncertainty related to the price of final products to the storage of which the minimum and maximum capacity is
market have not been explicitly dealt with. The work of Chebeir known.
et al. [20] accounted for market uncertainties in a general form by  Sets of power plant W in which portion of the processed gas is
considering only uncertainty related to natural gas liquid (NGL) and sent for power generation depending on the given demand.
natural gas price to the general market. However, market un-  Portion of the natural gas is also transported to a set of under-
certainties as related to all the possible end use of natural gas to ground storage for seasonal backup based on the given mini-
different customers have not been considered. This includes the mum and maximum capacity storage.
uncertainty related to the prices of natural gas, NGL and electricity  Sets of distribution company B that allocate natural gas to the
to different end users such as industrial, residential and commercial consumers with known minimum and maximum demand.
customers. The prices of these final products vary and difficult to  Sets of industrial customers I, residential customers E and
control their change in the entire time horizon of interest and this commercial customers C. The energy demand of these cus-
may have a huge impact on designing an optimal shale gas tomers can be satisfied by electricity from the set of trans-
network. mission line K or natural gas from the distribution company.
Therefore, the proposed framework considers the uncertainties  NGL, electricity and natural gas price to various customers.
related to the prices of all natural gas end use and the effect on the These are considered uncertain and are described through
shale gas system optimal design and operation is investigated. The analysis of historical price variations for each customer. A set of
proposed model also accounts for the scheduling problem in rela- scenarios is generated depending on the statistical distribution
tion to hydraulic fracturing activities. This is an extension of Oke in relation to the nature of the occurrence.
et al. [16] with the incorporation of uncertainties in the model to
ensure robustness of the designed network. Uncertainties are The superstructure representation depicted in Fig. 1represents
modelled via randomly generated scenarios using samples from the key elements of the problem to be solved.
beta distribution of the price, based on historical data. The major Associated with the total network are different uncertainties as
distinction compared with the work of Al-Aboosi and El-Halwagi shown in the super structure. Among these are: uncertainties
[22] in relation to uncertainty modelling is the sampling tech- associated with supply of water from the interruptible source,
nique used in this study. This study uses past data to find the dis- volume of water expected as flowback after fracturing operation,
tribution which is used for Monte Carlo sampling for the probability price of NGL, natural gas demand and price to different consumers,
of occurrence of each scenario. This study does not necessarily and price of electricity. However, it is worth mentioning that this
consider the distribution to be normal as the scheduling problem study only considers the uncertainty associated with price of NGL,
related to production of shale gas is accounted for and each sce- natural gas and electricity to different customers.
nario may appear randomly in the entire time horizon. The work of The objective is to determine the optimal design and operation
Al-Aboosi and El-Halwagi [22] on the other hand used three-point strategies of the shale gas network needed to maximize the ex-
approximation technique (low, medium, high) for probability of pected total profit under the influence of price uncertainties.
occurrence. Some of the assumptions made in the model formulation based
on existing literature on the subject are as follows:
D. Oke et al. / Energy 203 (2020) 117770 3

 The given time horizon is discretized unevenly into set of time 3.2. Propagation of uncertainty in the shale gas network
point n happening at unknown time.
 At each specified time point n, the uncertain price of NGL, nat- Optimization under uncertainty involves the propagation of
ural gas and electricity to various end users can be randomly uncertainties through the model to determine the impact of un-
sampled from the distribution of the corresponding price as certain parameters on the operation and design of the network as
obtained from the historical data. well as the effect on the objective function. In this work, un-
 Each scenario comprises the entire time horizon. certainties are modelled via randomly generated scenarios. At
 All the scenarios considered have the same probability of different time point n in a given scenario, the price of NGL, elec-
occurrence. tricity and natural gas are randomly selected from beta distribution
 Shale gas composition of each wellpad is known. [25] of the price, based on available monthly data from the past.
 Flowback water generated few days after fracturing is consid- Three different scenarios are generated from past data. It is worth
ered and assumed a fraction of the initial water used in frac- mentioning that the scenario generation is limited to three based
turing operation. on the time horizon of interest considered. The aim is finding the
 The number of wells in each wellpad is known and daily frac optimal configuration of the shale gas network and evaluate how
stages are fixed at 4. the network is affected if the market related uncertainty is
 The availability time of each wellpad is known. accounted for.
 The natural gas and NGL demands are known and remained The mathematical formulation is built on the superstructure
unchanged for the entire time horizon of interest. given in Fig. 1. It entails scheduling constraints, material balances
 The production profile of shale gas for each wellpad is known. constraints, bounding and capacity constraints, and the water
 The regenerator is situated onsite and can be moved within each treatment unit design constraints.
wellpad.
3.2.1. Scheduling constraints
Capturing the time related aspect in shale gas production pro-
cess is essential and this is accounted for by scheduling constraints.
3. Model formulation There are two important stages where capturing the time is of
necessity, these are the hydraulic fracturing stage which entails the
The proposed mathematical model is formulated based on wellpads scheduling and the flowback water management stage
continuous time formulation approach for scheduling using rep- which include the flowback water storage scheduling and the water
resentation that depends on the state task network by discretizing treatment unit scheduling. As the detailed presentation of the
the time horizon of interest unevenly into time point n, which constraints related to scheduling have been provided in Oke et al.
occurs at an unknown time. [16,26] a brief description related to wellpads scheduling is pro-
vided here.

3.1. Deterministic approach for shale gas network


3.2.1.1. Scheduling of wellpads. The allocation constraint specified
The detail of the deterministic approach can be found in Oke in Eq. (2) indicates that fracturing of each wellpad s has to take
et al. [16]. In the deterministic model, NGL, electricity and natural place precisely once at a particular time point n in scenario j in the
gas price to different consumers are represented by an average time horizon.
value and used for analysis of the supply chain network and eco- X
nomic performance. The objective aimed at profit maximization by ws;n;j ¼ 1 cs2S; j2J (2)
n
considering revenue generation from NGL, natural gas and elec-
tricity and also accounting for the cost related to the whole network Eq. (3) specifies that the occurrence time ttn of time point n in
as given in Eq. (1). scenario j must agree with the time ATs at which wellpad s is
available. Where binary variable ws;n;j shows if fracturing of well-
maxprofit ¼Retotal pad s happens at time point n in scenario j.
 
X  
 Ct fw þCt ww þCt imp þCt wdp þCt pro þCt trpt þCt strg þCt power ttn;j  ATs ws;n;j  H 1  ws;n;j cn 2 N; j2J (3)
s
(1)
The equation accounts for the following terms:

(i) Overall returns from the sale of natural gas, electricity, and 3.2.2. Material balance constraints
natural gas liquid,Retotal (ii) fresh water cost, Ct fw (iii) cost of
3.2.2.1. Freshwater and wastewater balance
flowback management,Ct ww (iv) cost of impoundmentCt imp
3.2.2.1.1. Around wellpad s. Water requirement by each wellpad
(v) cost related to drilling of wellpad and production of shale fw
gas, Ct wdp (vi) cost of gas processing plant,Ct pro (vii) trans- in scenario j is provided by freshwater from the impoundment fs;n;j
ww ,
portation cost for shale and natural gas,Ct trpt (ix) storage cost and/or recycled water from the frac tank fs;n;j as given by Eq. (4).
for natural gas and ngl,Ct strg (x) cost of power
generation,Ct power . fw ww
fs;n;j ¼ fs;n;j þ fs;n;j cs2S; n2N; j2J; n  2 (4)

The objective function is subject to equality and inequality fbw


Eq. (5) stipulates that the produced flowback water fs;n;j after
constraints. However, as the future occurrence cannot be predicted,
fracturing operation in scenario j can be disposed into the injection
the present approach is formulated by taking into account the trtb is the amount of
market uncertainty in terms of price of final products to different well or treated by regenerator R. Where fs;n;j
consumers. dis is the amount to be disposed.
wastewater to be treated and fs;n;j
4 D. Oke et al. / Energy 203 (2020) 117770

Fig. 1. Superstructure representation.

dis at time point n in


water sent from wellpad s to injection well fs;n;j
fbw trtb dis
fs;n;j ¼ fs;n;j þ fs;n;j cs2S; n2N; j2J (5) con as specified
scenario j and the concentrate from the regenerator fn;j
in Eq. (11)
3.2.2.1.2. Around impoundment t. Eq. (6) describes the total X
tdis dis conc
fn;j ¼ fs;n;j þ fn;j cn2N; j2J (11)
water use ifw
t;n;j
at a given time point in scenario j from each s
impoundment t where is the piping link between impoundment t
fw
and wellpad s. The overall quantity of water fpa;n;j withdrawn from
interruptible source a at a given time point in scenario j is defined
in Eq. (7) where TPa;t is the piping connection between impound- 3.2.2.2. Gas balance
ment t and source a. The amount of freshwater to be withdrawn 3.2.2.2.1. Around the wellpad and processing plant. The shale gas
from river a at a particular time point in scenario j should not production prs;n;j of wellpad s at time point n in scenario j is given in
exceed its capacity range as specified in Eq. (8), where binary var- Eq. (12) whereps is a parameter that indicate the gas production of
iable ya specifies if river a exists and Fapump;max is a parameter rep- wellpad s.
resenting the maximum water withdrawal allowed from river a.
prs;n;j ¼ ps ws;n;j cs2S; n2N; j2J (12)
The volume of impoundment t at each time point n in scenario j is
pump
given by Eq. (9), where ft;n;j indicates the volume of freshwater The produced shale gas is sent to various processing plants as
truck is
provided by piping to the corresponding impoundment and ft;n;j specified in Eq. (13).
the volume of freshwater provided by trucking. X
prs;n;j ¼ sps;p;n;j cs2S; n2N; j2J (13)
X p
ifw
t;n;j
¼ fw
fs;n;j ct2T; n2N; j2J (6)
s2TPs;t The natural gas production tmp;n;j at processing plant depends
on the average composition of methane MCs of the shale gas and
pump
X the efficiency of the processing plant PE as stated in Eq. (14).
ft;n;j ¼ fpfw ct2T; n2N; j2J (7)
a;n;j X
a2TPa;t tmp;n;j ¼ sps;p;n;j MCs PE cp2P; n2N; j2J (14)
s

fpfw
a;n;j
 Fapump;max ya ca2A; n2N; j2J (8) At the processing plant, the produced natural gas after pro-
cessing is equal to the quantity sent to various power plants
pump fw
pwp;w;n;j , distribution companies pdp;b;n;j and underground reser-
truck
vit;n;j ¼ vit;n1;j þ ft;n;j  it;n;j þ ft;n;j ct2T; n2N; j2J (9) voir pup;u;n;j as stated in Eq. (15).
X X
3.2.2.1.3. Around regenerator R and injection well d. The water
tmp;n;j ¼ pwp;w;n;j þ pdp;b;n;j
w b
balance around the treatment unit is given in Eq. (10). It specifies X
that the quantity of water sent to the regenerator trtb
fs;n;j at time point þ pup;u:n;j cp2P; n2N; j2J (15)
u
n in scenario j is equal to the volume collected as permeate and sent
ww and the volume collected as concentrate and
to the frac tank ft;n;j
con . 3.2.2.2.2. Around underground reservoir u. The gas balance
sent to disposal ft;n;j
around the underground reservoir u at time point n in scenario j is
X X given in Eq. (16) where pup;u;n;j is the quantity of gas sent from the
trtb ww conc
fs;n;j ¼ fs;n;j þ fn;j cn2N; j2J (10) processing plant to underground reservoir and uwu;w;n;j is the
s s
amount of natural gas transported from underground reservoir to
tdis is the sum of
The overall quantity of water sent to disposal fn;j power plant w at time point n.
D. Oke et al. / Energy 203 (2020) 117770 5

X X X
vund;res
u;n;j
¼ vund;res
u;n1;j
þ pup;u;n;j  uwu;w;n;j cu2U; n2N; j2J rec
Pn;j ¼ loss
tew;n;j  Pn;j cn2N; j2J (23)
p w w

(16)
3.2.2.3.2. At the power substation. At the substation, the
received power is sent to various consumers as indicated in Eq. (24).
3.2.2.2.3. Around distribution company b. The quantity of natu-
ral gas transported to distribution Company b is delivered to pind com res
i;n;j , pc;n;j , pe;n;j are variables, that signify power sent to industrial,
various customers as specified in Eq. (17). indb;i;n;j , comb;c;n;j , resb;e;n;j commercial and residential consumers respectively. Depending on
represent the amount of gas distributed to the respective com- the voltage requirement of each customer, the power is stepped
mercial, industrial, and residential customers. down to lower voltages as specified in Eq. (24) e(27) where V com ,
X X X V ind , V res are parameters representing the voltage requirement for
pdp;b;n;j ¼ indb;i;n;j þ comb;c;n;j commercial, residential and industrial customers respectively.
p i c
X X X X
rec ind com res
þ resb;e;n;j cb2B; n2N; j2J (17) Pn;j ¼ Pi;n;j þ Pc;n;j þ Pe;n;j n2N; j2J (24)
e i c e

P ind
3.2.2.2.4. Around the NGL storage. The natural gas liquid pro- Pi;n;j
duction nglp;n of each processing plant depends on the average NGL
trans;ind
In;j ¼ i pffiffiffi cn2N; j2J (25)
V ind PF 3
composition NCs in the shale gas as given in Eq. (18).
X P com
nglp;n;j ¼ sps;p;n;j NCs PE cp2P; n2N; j2J (18) Pc;n;j
trans;com
s In;j ¼ c pffiffiffi cn2N; j2J (26)
V com PF 3
Eq. (19) represents the natural gas liquid balance at the pro-
cessing plant p at a given time point n in scenario j where
P res
Pe;n;j
nglp;n;j indicates the quantity of NGL produced at time point n and trans;res
In;j ¼ e pffiffiffi cn2N; j2J (27)
nlsp;n is the quantity of NGL sold at time point n in scenario j. V res PF 3

ngl ngl
vp;n;j ¼ vp;n1;j þ nglp;n;j  nlsp;n;j cp2P; n2N; j2J (19)
3.2.3. Bounding and capacity constraints
3.2.3.1. Processing plant p. Eq. (28) is the capacity constraint for the
3.2.2.3. Electricity balance processing plant. It specifies that the overall quantity of shale gas
3.2.2.3.1. At power plant w. Electricity generation tew;n;j at po- that each plant processes at time point n in scenario j should not
wer plant w is a function of the sum of gas volume sent to power surpass its capacity range. The processing plant capacity is repre-
plant w from processing plant p and underground reservoir u at sented by a continuous variable ccapp;j
.
time point n in scenario j and parameter EG, which represents the X cap
amount of electricity generated per unit shale gas input. This is sps;p;n;j  cp;j p 2 P; cn2N; j2J (28)
represented by Eq. (20) as: s

! The existence of a processing plant should be bounded by its


X X
tew;n;j ¼ EG pwp;w;n;j þ uwu;w;n;j cw 2 W; n 2 N; j2J corresponding capacity range as specified in Eq. (29)
p u
cap
(20) cpmin yp;j  cp;j  cpmax yp;j cp2P; j2J (29)

The voltage of the generated power at the power plant is low ngl
The amount of natural gas liquids stored vp;n;j at a particular
and required stepping up to a higher voltage before transmission to
processing plants p in scenario j should not surpass its storage ca-
various consumers over a long distance. Using higher voltage for
pacity V ngl;max as given in Eq. (30) where yp;j signifies the existence
transmission reduces losses associated with transmission. The
trans at time point n in scenario j is of a processing plant.
current on the transmission line In;j
given in Eq. (21). V trans , PF are parameters that indicate the voltage ngl
vp;n;j  V ngl;max yp;j cp2P; n2N; j2J (30)
of transmission and power factor respectively.
P The quantity of natural gas liquids sold from all the processing
tew;n;j plants at time point n in scenario j cannot surpass the demand
In;j ¼ w
trans
pffiffiffi cn2N; j2J (21) NLmax as represented by Eq. (31)
V trans PF 3
X
The power loss during high voltage transmission ploss
n;j is indi-
nlsp;n;j  NLmax cn2N; j2J (31)
p
cated in Eq. (22) while the received power at the substation prec
n;j
is given in Eq. (23) where Rtrans is a parameter representing the
resistance of transmission.
 2 3.2.3.2. Power plant w. The total quantity of natural gas delivered
loss trans
Pn;j ¼ In;j Rtrans cn 2 N; j2J (22) to individual power plant w from underground reservoir u and
processing plant p in scenario j should not surpass the
demandW max of the power plant as specified in Eq. (32)
6 D. Oke et al. / Energy 203 (2020) 117770

X X in Eq. (42) e (44)


pwp;w;n;j þ uwu;w;n;j  W max yw;j w2W; cn2N; j2J X
p u indb;i;n;j  INDmax i2I; cn2N; j2J (42)
(32) b

At the power plant, the total amount of power generated that is X


transmitted by individual transmission line k must not surpass the comb;c;n;j  COMmax c2C; cn2N; j2J (43)
b
capacitytck of the transmission line as specified in Eq. (33).
X X
tew;n;j  tck;j ck2K; n2N; k2K (33) resb;e;n;j  RESmax b2B; cn2N; j2J (44)
w e

If a transmission line exists, its transmission capacity should be


bounded by the corresponding range as represented by Eq. (34)

kmin yk;j  tck;j  kmax yk;j ck2K; j2j (34) 3.2.3.5. Pipeline. Eqs. 45e49 give the capacity of all pipeline
transporting gas. cSpipe;min and cSpipe;max signify the minimum and
Eq. (35) e (37) specified that the quantity of power transmitted
maximum capacity of pipeline transporting shale gas, while
to various end users should not exceed the customer's demand.
cNpipe;min and cNpipe;max represent the minimum and maximum ca-
Based on the information from literature [2],M ind ,M com and M res
pacity of pipeline transporting natural gas respectively.
represent the percentage of power generated per time that is
transmitted to industrial, commercial and residential customers sp pipe sp
respectively. cSpipe;min ys;p;j  cs;p;j  cSpipe;max ys;p;j s2S; p2P; j2J (45)
X
ind
Pi;n;j  Mind Pn;j
rec
cn2N; j2J (35) pw pipe
cNpipe;min yp;w;j  cp;w;j  cNpipe;max yp;w;j
pw
p2P; w2W; j2J
i
(46)
X
com
Pc;n;j  M com Pn;j
rec
cn2N; j2J (36)
c cNpipe;min ypu
p;u;j
 cpipe
p;u;j
 cNpipe;max ypu
p;u;j
p2P; u2U; j2J (47)
X
res
Pe;n;j  Mres Pn;j
rec
cn2N; j2J (37) pd pipe pd
e cNpipe;min yp;b;j  cp;b;j  cNpipe;max yp;b;j p2P; b2B; j2J (48)

pipe
cNpipe;min yuw
u;w;j  cu;w;j  c
Npipe;max uw
yu;w;j u2U; w2W; j2J
3.2.3.3. Underground reservoir u. The overall quantity of natural
(49)
gas stored vund;res
u;n;j
at underground reservoir u at time point n in
The quantity of gas transported by various pipeline from one
scenario j must not surpass the capacity of the storage V und;res;max as point to the other at time point n in scenario j is limited by the
specified in Eq. (38). Eq. (39) indicates that the quantity of gas
capacity of the pipeline as specified in Eq. (50) e (54)
injected pup;u;n;j in each reservoir cannot exceed its injection ca-
pacity IC max;und while Eq. (40) states that the amount of gas with- pipe
sps;p;n;j  cs;p;j s2S; p2P; cn2N; j2J (50)
drawn uwu;w;n;j from each underground reservoir u at time point n
should not surpass its withdrawal ability WC max;und. Where binary
variable yu;j specifies the existence of underground reservoir u in pwp;w;n;j  cpipe
p;w;j
p2P; w2W; cn2N; j2J (51)
scenario j.
pup;u;n;j  cpipe
p;u;j
p2P; u2U; cn2N; j2J (52)
vund;res
u;n;j
 V und;res;max yu;j u2U; cn2N; j2J (38)

X pdp;b;n;J  cpipe
p;b;J
p2P; b2B; cn2N (53)
pup;u;n;j  IC max;und yu;j u2U; cn2N; j2J (39)
p
pipe
uwu;w;n;j  cu;w;j u2U; w2W; cn2N; j2J (54)
X
uwu;w;n;j  WC max;und yu;j u2U; cn2N; j2J (40)
w

3.2.3.6. Impoundment t. The capacity of impoundment t is bounded


3.2.3.4. Distribution company b. Eq. (41) states that the quantity of by the corresponding capacity range as specified in Eq. (55)
natural gas received by distribution company b from processing
plant p should not exceed the demand Bmax . Where binary variable T min  ct;j  T max ct2T; j2J (55)
yb signifies if a distribution company exists.
X The impoundment volume at any time point n in scenario j
pdp;b;n;j  Bmax yb;j b2B; cn2N; j2J (41) should not exceed its storage capacity as stated in Eq. (56)
p
vit;n;j  ct;j ct2T; n2N; j2J (56)
The quantity of natural gas delivered to various customers from
distribution company b should not surpass the demand as specified
D. Oke et al. / Energy 203 (2020) 117770 7

3.3. Membrane distillation (MD) model


00 0 11
The study by Elsayed et al. [27] serves as the basis on which the XXX ft XX st;ww
Ctjww ¼ @@OC st;ww @
vs;n;j þ vn;j AA
MD model adopted in this work is built. This model will not be s n n
j j
discussed here as it has been presented in Oke et al. [26]. 0 0 1
XXX XX perm
þ @OC pump;ww @ trtb
Ds fs;n;j þ fn;j A
3.4. Objective function s n j n j
1
X  
The objective aims at maximization of the total profit consid- þ CC pipe Ds A þ 58:5Am þ 1115ff feed
ering the uncertainty in the price of final products. The total profit s
for each scenario is determined by considering revenue for each  
scenario j minus the total cost for each scenario. þ ð1411 þ 43ð1  LRÞ þ 1613ð1 þ uÞÞff feed þ AOT QOC ht
0 11
XX XXX
maxprofitðjÞ¼Retotal
j @
þ OC dis
fdn;j þ OC ww;truck
Dds fdn;j AA
  n j s n j
fw ww wdp pro shalegas natgas strg power
 Ct þCt þCtj þCtj þCtj þCtj þCtj þCtj
(60)
(57)
Cost associated with the drilling of wellpads and production of
The revenue generation Retotal of each scenario comprises of the shale gas Ctj
wdp
for each scenario j is defined in Eq. (61). Where-
revenue from the sale of natural gas, NGL, and electricity as given in
Eq. (58): ucdrill
s;n represents the unit cost associated with drilling wellpad s at
time point n, NSs signifies the number of wells in wellpad s, ucprod
s;n

!
X gas;ind
XX gas;com
XX gas;res
XX
Retotal
j ¼ SPn;j indb;i;n;j þSPn;j comb;c;n;j þSPn;j resb;e;n;j
n b i b c b e
! !
X ngl
X X elec;ind
X elec;com
X X
þ SPn;j nlsp;n;j þ SPn;j ind
Pi;n;j þSPn;j com
Pc;n;j þSPjelec;res res
Pe;n;j (58)
n p n i c e

represents the cost associated with production of a unit of shale gas


at wellpad s
Cost associated with freshwater management, which includes
the cost of freshwater withdrawal, transportation, and storage is
given in Eq. (59).

0 0 1 0 0 1
XXX XXX pump XXX XXX pump XX
Ctj ¼ @OC wd;fw @ ft;n;j A þ @OC truck;fw truck @
CC pipe Da;t A
fw truck
ft;n;j þ Dt ft;n;j þ OC pump;fw ft;n;j þ
t n j t n j t n j t n j a t
1 0 11
XX
 A þ @CC base;imp þ IC inc;imp ct;j AA
t j

(59)

!
wdp
XX prod

Ctj ¼ ucdrill
s;n NSs þ ucs;n prs;n;j (61)
The cost of flowback water management, which includes s n
pumping, storage, treatment and disposal cost is represented by Eq.
The processing plant cost for each possible scenario j which
(60)
comprises of the capital and variable cost, is represented by Eq. (62)
8 D. Oke et al. / Energy 203 (2020) 117770

as:
0 0 !
sfp 1 1 XX X X
pro @
X ccap chem pro XXX strg
Ctj ¼ OC inj;natgas
pup;u;n;j þ OC wd;natgas uwu;w;n;j
Ctj ¼ CC ref ;pro @ p A þOC pro
sps;p;n;j A
rcppro rchempro u n p w
p s p n !
XX ngl
(62)  þ OC st;ngl vp;n;j
p n
The cost of transporting shale gas from the wellpads to the
processing plants for each scenario j is given in Eq. (63) (65)

0 0 pipe sft 1 The cost of power generation and transmission Ctjpower for each
XX cs;p pipe scenario j is given in Eq. (66) as:
shalegas
Ctj ¼ @CC ref ;pipe
Ds;p @ A chem
rcp pipe rchempipe
s p
1 (63) !!
XX X X
XXX power
Ctj ¼ OC power;gen
pwp;w;n;j þ uwu;w;n;j
þ OC pipe
Ds;p sps;p;n;j A
w n p u
s p n !
XX
natgas
The natural gas transportation cost Ctj , which comprises of þ OC ele;trans tew;n;j
w n
the cost of transporting natural gas from processing plant to various
(66)
end point is represented in Eq. (64).

0 0 pipe sft 1 1
X X cp;w chempipe XXX
natgas
Ctj @
¼ CC ref ;pipe
Dp;w @ A þ OC pipe
Dp;w pwp;w;n;j A
p w rcppipe rchempipe p w n
0 0  1 1
pipe sft
B ref ;pipe XX B cp;b XXX
pipe
C chem C
þ@CC Dp;b @ pipe
A pipe
þ OC pipe Dp;b pdp;b;n;j A
p b rcp rchem p b n

0 0 pipe sft 1 1
XX cp;u chempipe XXX
þ@CC ref ;pipe Dp;u @ A þ OC pipe Dp;u pup;u;n;j A
p u rcppipe rchempipe p u n
0 0 pipe sft 1 1
XX cu;w chempipe XXX
@
þ CC ref ;pipe
Du;w @ A þ OC pipe
Du;w uwu;w;n;j A (64)
u w rcppipe rchempipe u w n

3.5. Solution approach


The cost related to natural gas and natural gas liquid storage for
each scenario j, which comprises of the cost of injecting and The solution strategy adopted consist of three stages as shown
withdrawing of natural gas, and NGL storage cost is given in Eq. in Fig. 2.
(65). The first stage involves the consideration of natural gas, NGL and
electricity price seven years historical data. The data can be used to
analyse the sample data statistically in order to get the standard

Fig. 2. Proposed solution approach.


D. Oke et al. / Energy 203 (2020) 117770 9

Table 1
Wellpad data.

Wellpads S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14

Earliest frac day 1 42 133 133 497 497 140 77 133 133 133 497 497 497
# of stages 120 120 210 140 140 210 434 492 280 175 280 175 70 350
# of wells 10 4 6 4 4 6 14 12 8 5 8 5 2 10

Table 2
Shale gas, Water and regeneration network data.

Parameter Value

transition time from one wellpad to the other (day) 5


Water demand for each frac stage (m3) 1271.9
Efficiency of processing plant 0.97
Cost of storage ($/m3) 0.59
Cost of trucking freshwater ($/m3) 29.35
Pumping cost (freshwater) ($/m3) 15.93
Incremental cost of impoundment ($/m3) 7.40
Wastewater disposal cost ($/m3) 134.18
Pumping cost (wastewater) ($/km/m3) 0.28
Base capital cost of Impoundment ($) 360,000
Pipeline cost ($/km) 202,000
Cost of withdrawing freshwater ($/m3) 0.99
Reference processing plant capital investment ($) 21,310,000
Pipeline chemical engineering plant cost index 881.9
Processing plant chemical engineering plant cost index 574
Shale gas methane composition 0.8e0.9
NGL composition in shale gas 0.1e0.2
Reference year Chemical engineering plant cost index for pipeline 887.6
Reference year chemical engineering plant cost index for processing plant 567.3

Fig. 3. Natural gas price (a) commercial customers (b) residential customers (c) industrial customers.

deviation, mean value and the variance, thereby, predicting the generated using probability density function like Gamma, Normal,
uncertain data. The dispersion of the data can be analysed using the Beta or Exponential distribution. The probability function will help
obtained statistical parameters and a set of scenarios can be to obtain random data which shows the probability that the
10 D. Oke et al. / Energy 203 (2020) 117770

Fig. 4. Electricity price (a) commercial customers (b) residential customers (c) Industrial customers.

Fig. 5. Natural gas liquid price.

random variable will take a specific value, thereby giving a future X  


fw wdp pro
prediction of the uncertain variable. This study used Beta distri- maxprofit ¼ pj Retotal
j  Ctj þ Ctjww þ Ctj þ Ctj
bution based on historical data [28]. j
In stage 2, the generation of multi-scenario framework is

þ Ctjshalegas þ Ctjnatgas þ Ctjstrg þ Ctjpower (67)
considered, whereby the probability density function obtained in
stage one is used to generate random sample prices for each of
natural gas, NGL and electricity customers to obtain the scenarios.
However, the number of scenario generation in this study is based
on the time horizon of interest considered. Each scenario is solved 4. Case study
individually for economic objective and a possible configuration for
each individual scenario is obtained. An illustrative example adapted from Yang et al. [29] was used
The last stage involves solving all the scenarios together to test the applicability of the proposed framework as shown in
simultaneously as indicated in Fig. 2 and the supply chain network Table 1. The wellpads are distributed in a given region in two
in the uncertainty environment is obtained. The objective in this clusters. The two areas are surrounded by five small rivers (inter-
stage is to maximize the average total profit (ensemble average) of ruptible sources), and one uninterruptible source from where
all scenarios given. It is to be mentioned that we assumed all sce- freshwater can be withdrawn and the time horizon of interest is 3
narios have equal probability of occurrence. The ensemble average years.
profit is obtained by the sum of profit multiplied by the probability In addition, three sets of processing plants, four underground
of that scenario as given in Eq. (67), where pj is the probability of reservoirs, five natural gas distribution companies, four gas turbine
scenario j. combined cycle power plants with generation efficiency of 50%
based on lower heating value (LHV) and three sets of transmission
lines (high voltage) were considered. The necessary input
D. Oke et al. / Energy 203 (2020) 117770 11

Fig. 6. Natural gas price scenario (a) residential customers (b) industrial (c) commercial customers.

Fig. 7. Electricity price scenario (a) industrial customers (b) commercial customers (c) residential customers.

parameters are provided in Table 2 whilst the data related to were taken from historical data for the last seven years as shown in
flowback water and concentration profile for a particular wellpad in Figs. 3e5 [30].
the first few days following wellpad fracturing, and the estimated
gas production for the wellpads were obtained from Yang et al. [29]. 5. Results and discussions
The data pertaining to optimization of shale gas were taken from
Gao and You [9] and the design parameters of the regeneration The optimization formulation was coded in GAMS and solved
process were adapted from Elsayed et al. [27]. The monthly average using BARON. Comparison is made between the results obtained
price for electricity, NGLs and natural gas to different end users using stochastic approach and that from deterministic approach.
12 D. Oke et al. / Energy 203 (2020) 117770

Fig. 8. Natural gas liquid price scenario.

Table 3
Cost analysis for each scenario.

Scenario Revenue $ (  109) Total Cost $ (  109) Profit $ (  109)

1 52.78 42.08 10.70


2 52.37 42.02 10.35
3 52.48 42.02 10.45
Deterministic Solution 51.15 41.98 9.16

Table 4 adopted, each scenario is solved individually followed by their


Revenue breakdown. ensembled average.
Scenario 1 Scenario 2 Scenario 3 Deterministic

Natural gas 41.24% 40.65% 40.9% 22.84% 5.1. Cost analysis


Electricity 46.32% 46.77% 46.57% 65%
Natural gas liquid 12.44% 12.58% 12.52% 12.14%
Tables 3 and 4 show the obtained results and variation in each
scenario in terms of cost and revenue respectively. The effects of
price uncertainty on the profit is compared with that from the
Three different scenarios were generated from the past data with deterministic solution. From the tables, the expected profit is found
each scenario comprising of 14 time points. Figs. 6e8 show the to be $10.70 billion with total cost of $ 42.08 billion for scenario 1.
scenario generation for each customer with observation that The revenue generated amount to $ 52.78 billion. The sale of nat-
different price profiles can be obtained for each customer ural gas constituted 41.24% of the total revenue while 46.32% and
depending on the different random values. In the solution sequence 12.44% comes from electricity and natural gas liquid sale

Fig. 9. Detail cost breakdown.


D. Oke et al. / Energy 203 (2020) 117770 13

Table 5
Network configuration.

Scenario Processing plant Underground Reservoir Distribution company Power Plant Transmission line

1 3 2 4 2 3
2 3 1 4 2 3
3 3 1 3 2 3
Deterministic 2 1 4 2 3

respectively. As can be seen from the table, the highest revenue as $19.9 million and $21.9 million respectively. Natural gas liquid
well as cost is observed in scenario 1. The profit realised in scenario storage in each of these scenarios is necessary in order to take full
2 amount to $10.35 billion while the revenue and total cost are advantage of the fluctuation in the price of NGL throughout the
$52.37 and $ 42.02 billion respectively. This scenario represents the time horizon to promote revenue maximization. This results in
lowest in terms of revenue as well as profit. 46.77% of revenue 2.4%, 3.5% and 3.04% increase in the percentage that revenue
generated in this scenario arise from the sale of electricity while generated from NGL contributes to the total revenue as shown in
40.65% and 12.58% are realised from the sale of natural gas and NGL Table 3 for scenario 1, 2 and 3 respectively compared to the
respectively. For scenario 3, the total profit is $10.45 with revenue deterministic solution. These results are in agreement with other
and total cost of $52.48 and $ 42.02 billion respectively. In this published literature [20].
scenario, the sale of electricity account for 46.57% of the total rev- In addition, increase in the cost of natural gas transportation is
enue, while the sale of natural gas and NGL constituted 40.9% and also observed in the supply chain networks of the three scenarios.
12.52% respectively. The natural gas transportation cost increased from $47.1million in
The profit realised in the case of the deterministic solution is the deterministic approach to $87million in scenario 1,
$9.16 billion with revenue and total cost of $51.15 and $ 41.98 billion $52.7million scenario 2 and $48.8million in scenario 3 respectively.
respectively. Significant part of the revenue generated in the This is because of higher capacity of pipeline required to allocate
deterministic solution comes from the sale of electricity with 65% of natural gas to the distribution companies in order to take advan-
the total revenue, 22.84% from natural gas sale and the sale of NGL tage of natural gas price fluctuation to different customers for
constitutes the remaining 12.14%. Although, this approach indicates revenue maximization. This effect can be observed in the per-
a slightly lower cost, the profit realised is lower compared to sce- centage that sale of natural gas contributed to the revenue gener-
nario 1 to 3 due to lower revenue generation. Results from the three ation in the three scenarios compared to deterministic solution as
different scenarios show 14.39%, 11.49%, and 12.34% increase in the shown in Table 3. More gas is transported to the distribution
profit realised, respectively, compared to the deterministic companies taking advantage of variation in price, thereby the
approach. This shows the significant of taking into consideration percentage of revenue generated from natural gas sale increased
the uncertainties associated with price of natural gas, electricity from 22.84% to 41.24%, 40.65% and 40.90% for scenario 1e3
and NGL compared to assuming an average value for a robust respectively.
solution.
The detailed cost breakdown for each scenario and the deter- 5.2. Optimal network configuration
ministic solution is given in Fig. 9. The figure shows the close
similarity in the cost associated with each scenario as well as the The established scenarios give opportunity for assessment of
deterministic solution. Shale gas production cost various possible solution for the network (optimum number of
(drilling þ production) which is the highest of the total cost lies processing plant, underground reservoir, distribution company and
between 51.5 and 51.7% followed by the processing plant cost which the transmission line). Table 5 shows the obtained configuration
accounts for between 41.7 and 42% of the total cost. The cost from each scenario in comparison to the deterministic solution. The
incurred in generating and transmitting power constitute 4% and Table shows that scenario 1e3 requires setting up 3 processing
1.8% respectively, while cost associated with water management as plants each while 2 processing plants are set up in the deterministic
well as storage and transportation of gas are negligible. solution. For the underground reservoir, scenarios 2 and 3 as well as
However, it is important to do a critical analysis of the networks the deterministic solution require 1 underground reservoir for
in terms of revenue vs cost. The increase in the profits realised in natural gas storage while scenario 3 require 2. The number of dis-
the three scenarios can be attributed to increase in revenue but at tribution companies require in scenario 1 and 2 are the same as the
the expense of total cost as shown in Tables 2 and 3 The extra cost deterministic while scenario 3 require less. All the scenario as well
associated with each scenario compared to deterministic solution is as the deterministic require the same number of power plants and
attributed to the extra cost incurred in natural gas liquid storage transmission lines to be set up.
and pipeline infrastructure. For instance, the deterministic Table 6 shows the capacity of processing plant require in each
approach incurred zero cost for NGL storage, as no NGL is stored scenario, all the three scenarios as explained earlier require the
which may be as a result of constant price assumed in the network same number of processing plants to be set up but with different
while in scenario 1e3, NGL storage cost amount to $33.1million, processing capacities. For scenario 1, the capacity of each

Table 6
Capacity of processing plants.

Scenario Processing plant 1 m3(  109) Processing plant 2 m3(  109) Processing plant 3 m3(  109)

1 2.15 2.36 3.03


2 2.36 2.36 2.82
3 2.26 7.08 1.25
Deterministic 5.01 5.39 N/A
Stochastic N/A 5.18 3.46
14 D. Oke et al. / Energy 203 (2020) 117770

Fig. 10. Fracturing schedule (Deterministic Solution).

Fig. 11. Fracturing schedule (Scenario 1).

Fig. 12. Fracturing schedule (Scenario 2).

processing plant ranges between 2.15e3.03  109 m3. There is very processing plants with capacity 5.18  109 m3 and 3.46  109 m3
close range in the capacities of the processing plant set up in sce- respectively while the capacity of processing plants in the deter-
nario 2 with value of 2.36e2.82  109 m3 while scenario 3 require ministic solution are 5.01  109 m3 and 5.39  109 m3. This leads to
processing plant that range from 1.25e7.08  109 m3. Comparing 0.66% reduction in the cost associated with processing in the un-
the scenarios with the deterministic solution show that the certainty environment. This shows that the price uncertainties have
deterministic configuration requires less processing plant but with effect on the network configuration. It is to be mentioned that the
higher capacity each. However, every change in capacity is impor- processing plant selection in the two cases mainly depends on the
tant and must be considered, as an oversized or undersized design distances to the wellpads.
will have effect on both the investment cost and the operation.
Therefore, analysis of the network under uncertainty environment
was carried out by solving all scenarios simultaneously in order to 5.3. Fracturing schedule
obtain optimal number of processing plants and pipeline capacities.
In the uncertainty environment, the optimal network requires 2 The fracturing schedule for the deterministic solution as well as
each scenario are given in Figs. 10e13. The schedule in all cases
D. Oke et al. / Energy 203 (2020) 117770 15

Fig. 13. Fracturing schedule (Scenario 3).

depend majorly on water availability for fracturing as in most cases involve in determining the optimal network. This indicated that
the wellpads are stimulated in a later stage. In the case of deter- coordinating the distribution operation of final products in an
ministic solution and scenario 2, wellpad S1 which is one of the optimal way can promote significant economic improvement of the
wellpad with least number of stages is fractured first at an earlier shale gas supply chain network. The higher revenue realised in the
stage and the wellpads that followed in both cases i.e S8 and S7 are optimization under uncertainty is at the expense of additional costs
wellpads with high stages thus require high quantity of water. but leads to higher profit.
Therefore, the fracturing activities is delayed to allow for enough Based on the results of this study, the oil and gas companies as
water in the impoundments. Scenarios 1 and 3 chose to commence well as the economic development organizations can use the
fracturing at later stage also starting with wellpads that require framework as a basic tool for the evaluation and implementation of
lower volume of water for fracturing. In all cases, water availability essential policies to prevent likely negative effects that may arise
is the determining factor as to when the fracturing activities from market uncertainty. Coordinating the post-production oper-
commence, however, the sequence of wellpads fracturing facilitates ations especially natural gas and NGL in terms of storage and
maximization of wastewater reuse and fracturing wellpads from allocation to various end users in an optimal way can result in a
the same region after each other will facilitate flexibility in terms of significant economic improvement. In addition, the results of this
logistic. It is worth mentioning that the wellpads sequencing in study and our previous studies [16,19] have clearly indicated that
scenario 1e3 also take into consideration the fluctuation in the for full economic potential of the shale gas enterprise, the decision
prices of final products to maximize revenue generation. In terms of makers must coordinate the system as a unified process and not as
water and energy requirements of the network, the same quantity a standalone. The results from literature example have shown that
of water for fracturing and energy requirement of the membrane consideration of uncertainty is important in shale gas production
process is applicable in both the deterministic as reported in Oke development to avoid unnecessary loss that may arise from un-
et al. [16] and stochastic solution as the number of wellpads frac- certain prices of final products. However, it is necessary for real case
tured are the same in the two approach. Savings in the freshwater studies to be considered so that the applicability of the model in
requirement for fracturing and in energy associated with water real shale gas project can be ascertained.
management amounted to 23.2% and 42.7%, respectively.

Funding
6. Conclusions
This research was supported by the National Research Founda-
A multi-scenario model has been proposed to capture the effect tion (NRF), South Africa.
of uncertainties in the price of electricity, natural gas and NGL to
different consumers on the design and operation of shale gas Declaration of competing interest
production to distribution network. Continuous time formulation
for scheduling is proposed to tackle the scheduling problem asso- The authors declare that they have no known competing
ciated with the shale gas network. The uncertainties in the final financial interests or personal relationships that could have
products prices are modelled via randomly generated scenarios appeared to influence the work reported in this paper.
using beta distribution of the price, based on historical data where
each month of a year represent a sample data. The probability
density function is used to establishing scenarios with randomly CRediT authorship contribution statement
generated price values per time point. The applicability of the
model and the solution approach is examined using an example Doris Oke: Conceptualization, Methodology, Data curation,
from literature. The results show that the incorporation of uncer- Software, Writing - original draft, Visualization, Investigation,
tainty leads to more robust and profitable operation and design of Validation, Writing - review & editing. Rajib Mukherjee: Concep-
the shale gas network compared to the deterministic approach. The tualization, Methodology, Data curation, Visualization, Validation,
effect of this is more profound in the revenue generation as well as Software, Investigation, Writing - review & editing. Debalina
the optimal supply chain network. Results from the three different Sengupta: Visualization, Validation, Investigation, Writing - review
scenarios show 14.39%, 11.49%, and 12.34% increase in the profit, & editing. Thokozani Majozi: Supervision, Software, Investigation,
respectively, compared to deterministic approach. In addition, the Data curation, Validation, Visualization, Writing - review & editing.
results have shown that the uncertainties associated with the pri- Mahmoud El-Halwagi: Supervision, Validation, Visualization,
ces of final products have a significant effect on different operations Methodology, Writing - review & editing.
16 D. Oke et al. / Energy 203 (2020) 117770

Nomenclature OC wd;nat gas Cost of withdrawing natural gas, $/m3


OC ele;trans Cost of power transmission, $/kWh
OC pro Processing plant cost, $/m3
Sets pss Expected shale gas production of wellpad s, m3
A {a | a ¼ river source} PE Processing plant efficiency
B {b | b ¼ distribution company} PF Power factor
C {c | c ¼ commercial customer} Rtrans Transmission line resistance, ohm
D {d | d ¼ injection well} rcppipe Gas pipeline reference capacity, m3
E {e | e ¼ residential customer} rcp Processing plant reference capacity, m3
I {i | i ¼ industrial customers} rchempipe Chemical Engineering plant cost index of the reference
K {k | k ¼ transmission line} year for pipeline
N {n | n ¼ time point} rchempro Processing plant reference year Chemical Engineering
P {p | p ¼ processing plant} plant cost index
S {s | s ¼ wellpad} sfp Size factor for processing plant
T {t | t ¼ impoundments} sft Size factor for gas pipeline
Tpa;t Connection between impoundment t and river source a
SP gas;ind Industrial customers gas sale price, $/m3
TPs;t Connection between impoundment t and wellpad s
SP gas;com Commercial customers gas sale price, $/m3
U {u | u ¼ underground reservoir}
SP gas;res Residential customers gas sale price to, $/m3
W {w | w ¼ power plant}
SP ngl NGL sale price, $/m3 gas
J {j | j ¼ scenario}
SP ele;com Electricity price to commercial customers, $/kWh
SP ele;ind Electricity price to industrial customers, $/kWh
Parameters
SP ele;res Electricity price to residential customers, $/kWh
AOT Annual operating time, hr
u Ratio of recycled reject to raw feed
ATs Time that wellpad s is available, day
V trans Voltage of transmission, kV
CC base;imp Base capital cost of impoundment, $
V ind Industrial customer’s voltage requirement, kV
CC pipe Capital cost of pipeline, $/km
V com Commercial customer’s Voltage requirement by, kV
CC ref ;pro Reference capital cost for processing plant, $ V res Residential customer’s Voltage requirement by, kV
CC ref ;pipe Reference pipeline capital cost, $ V ngl;max NGL Storage capacity, m3
cpmin Processing plant minimum capacity, m3
cpmax Processing plant maximum capacity, m3
CTSS

Transition time between wellpads, day Binary variables
ws;n;j Indicates if wellpad s is fractured at time point n in
chempipe Pipeline chemical engineering plant cost index
scenario j
chempro Processing plant Chemical Engineering plant cost index
ya Specifies the existence of river a
Dt Distance between the uninterruptible freshwater source
yb;j Signifies if distribution company b exists in scenario j
and impoundment t, km
Da;t Distance between the interruptible freshwater source a yp;j Signifies the existence of processing plant p in scenario j
and impoundment t, km yk;j Signifies the existence of transmission line k in scenario
Ds Distance between wellpad s and the treatment facility, j
km yu;j Signifies the existence of underground reservoir u in
Du;w Distance between underground reservoir u and power scenario j
plant w, km yw;j Signifies if power plant w exists in scenario j
EG Power produced per natural gas unit input, kWh/m3 sp
ys;p;j Signifies if a pipeline is set up between s and p in
Fapump;max Maximum quantity of water that can be withdrawn scenario j
from source a, m3
H Time horizon of interest, yr
IC max;und Underground reservoir injection capacity of, m3 Continuous variables
INDmax; Industrial customers natural gas demand ct Capacity of impoundment t, m3
cap
K min Transmission line capacity (minimum), MW cp Processing plant capacity, m3
K max Transmission line capacity (maximum), MW ct fw Fresh water cost, $
MCs Composition of methane in wellpad s ct ww Cost of flowback water management, $
NCs Composition of NGL in wellpad s pro
ctj Cost of processing plant, $
NLmax NGL Maximum demand, m3 trpt
ctj Cost of transporting shale and natural gas $
OCstruck;fw Cost of trucking freshwater, $/m3
ctjwdp Drilling and shale gas production cost, $
OC wd;fw Cost of freshwater withdrawal, $/m3
OC dis Wastewater disposal cost, $/m 3 ctjpower Power generation and transmission cost, $
OCsst;ww Cost of wastewater storage, $/m3 comb;c;n;j Quantity of natural gas delivered to commercial
OC ww;truck Cost of wastewater trucking, $/km/m3 customer c from distribution company b at time point n
OC ht Cost of heating, $/109 J in scenario j, m3
OC power;gen Cost of electricity production from natural gas, $/m3 fs;n Water requirement of wellpad s at time point n, m3
fw
OC st;ngl Cost of NGL storage, $/m 3 fs;n Freshwater requirement of wellpad s at time point n, m3
fw
OC inj;nat gas Cost of injecting natural gas, $/m3 fpa;n Freshwater pumped from source a at time point n, m3
D. Oke et al. / Energy 203 (2020) 117770 17

ww
fs;n Wastewater requirement of wellpad s at time point n, 10.1002/aic.14705.
[9] Gao J, You F. Shale gas supply chain design and operations toward better
m3 economic and life cycle environmental performance: MINLP model and global
fw optimization algorithm. J ACS Sustainable Chem Eng 2015b;3:1282e91.
it;n Amount of freshwater utilized from impoundmentt at
https://doi.org/10.1021/acssuschemeng.5b00122.
time point n, m3 [10] Guerra OJ, Calderon AJ, Papageorgiou LG, Siirola JJ, Reklaitis GV. An optimi-
trans;ind zation framework for the integration of water management and shale gas
In;j Current transmitted to the industrial customers at time
supply chain design. J Comput Chem Eng 2016;92:230e55. https://doi.org/
point n in scenario j, Amp 10.1016/j.compchemeng.2016.03.025.
[11] Arredondo-Ramirez K, Ponce-Ortega JM, El-Halwagi MM. Optimal planning
nglp;n;j NGL production at time point n in scenario j, m3
and infrastructure development for shale gas production. J Energy Convers
nlsp;n;j NGL sold at time point n in scenario j, m3 Manage 2016;119:91e100. https://doi.org/10.1016/j.enconman.2016.04.038.
rec
Pn;j Received power at substation at time point n in scenario [12] Gao J, You F. Game theory approach to optimal design of shale gas supply
chains with consideration of economics and life cycle greenhouse gas emis-
j, kWh sions. AIChE J 2017;63:2671e93. https://doi.org/10.1002/aic.15605.
res
Pe;n;j Power transmitted to residential customers e at time [13] Chen YZ, He L, Guan YL, Lu HW, Li J. Life cycle assessment of greenhouse gas
emissions and water-energy optimization for shale gas supply chain planning
point n in scenario j, kWh based on multi-level approach: case study in Barnett Marcellus, Fayetteville,
ind
Pi;n;j Power transmitted to industrial customers i at time and Haynesville Shales. Energy Convers Manag 2017;134:382e98. https://
doi.org/10.1016/j.enconman.2016.12.019.
point n in scenario j, kWh [14] Gao J, You F. Dynamic material flow analysis-based life cycle optimization
com
Pc;n;j Power sent to commercial customer c at time point n in framework and application to sustainable design of Shale gas energy systems.
ACS Sustainable Chem Eng 2018a;6:11734e52. https://doi.org/10.1021/
scenario j, kWh acssuschemeng.8b01983.
pdp;b;n;j Quantity of gas delivered to distribution company b [15] Gao J, You F. Integrated hybrid life cycle assessment and optimization of shale
gas. ACS Sustainable Chem Eng 2018b;6:1803e24. https://doi.org/10.1021/
from processing plant p at time point n in scenario j, m3 acssuschemeng.7b03198.
pup;u;n;j Quantity of gas delivered to underground reservoir u [16] Oke D, Mukherjee R, Sengupta D, Majozi T, El-Halwagi MM. Optimization of
water-energy nexus in shale gas exploration: consideration of production to
from processing plant p at time point n in scenario j, m3
transmission. Energy 2019;183:651e69. https://doi.org/10.1016/j.en-
pwp;w;n;j Quantity of gas delivered to power plant w from ergy.2019.06.104. 2019.
processing plant p at time point n in scenario j, m3 [17] Yang L, Grossmann IE, Manno J. Optimization models for shale gas water
management. AIChE J 2014;60:3490e501. https://doi.org/10.1002/aic.14526.
Q Heat requirement of the feed into MD, kJ/day
[18] Gao J, You F. Deciphering and handling uncertainty in shale gas supply chain
resb;e;n;j Quantity of natural gas sent from distribution company design and optimization: novel modeling framework and computationally
b to residential customer c at time point n in scenario j, efficient solution algorithm. AIChE J 2015c;61:3739e55. https://doi.org/
10.1002/aic.15032.
m3 [19] Lira-Barragan L, Ponce-Ortega JM, Guille n-Gos
albez G, El-Halwagi MM.
sps;p;n;j Volume of shale gas sent to processing plant p from Optimal water management under uncertainty for shale gas production. Ind
wellpad s at time point n in scenario j, m3 Eng Chem Res 2016;55:1322e35. https://doi.org/10.1021/acs.iecr.5b02748.
[20] Chebeir J, Geraili A, Romagnoli J. Development of Shale gas supply chain
ttn;j Time that correspond to time point n in scenario j, day network under market uncertainties. Energies 2017;10:246. https://doi.org/
tmp;n;j Methane production at time point n in scenario j, m3 10.3390/en10020246.
[21] Chen YZ, He L, Li J, Zhang SY. Multi-criteria design of shale-gas-water supply
tew;n;j Generated power at power plant w at time point n in chains and production systems towards optimal life cycle economics and
scenario j, kWh greenhouse gas emissions under uncertainty. Comput Chem Eng 2018;109:
tck;j Capacity of transmission line k in scenario j, MW 216e35. https://doi.org/10.1016/j.compchemeng.2017.11.014.
[22] Al-Aboosi FY, El-Halwagi MM. A Stochastic optimization approach to the
uwu;w;n;j Gas withdrawn from underground reservoir u to power design of shale gas/oil wastewater treatment systems with multiple energy
plant w at time point n in scenario j, m3 sources under uncertainty. Sustainability 2019;11:4865. https://doi.org/
10.3390/su11184865.
vngl
p;n;j
Quantity NGL stored at time point n in scenario j, m3 [23] Mukherjee R, Gonza lez-Bravo R, Napoles-Rivera F, Linke P, Ponce-Ortega JM,
El-Halwagi MM. Optimal design of water distribution networks with incor-
poration of uncertainties and energy nexus. Process Integration and Optimi-
References zation for Sustainability 2017;4:275e92.
[24] Herna ndez-Romero IM, Fuentes-Corte s LF, Mukherjee R, El-Halwagi MM,
[1] Jiang Z, Li R, Li A, Ji C. Runoff forecast uncertainty considered load adjustment Serna-Gonza lez M, Napoles-Rivera F. Multi-scenario model for optimal design
model of cascade hydropower stations and its application. Energy 2018;158: of seawater air-conditioning (SWAC) systems under demand uncertainty.
693e708. https://doi.org/10.1016/j.energy.2018.06.083. J Clean Prod 2019;238:117863. https://doi.org/10.1016/j.jclepro.2019.117863
[2] United State Environmental Protection Agency (USEPA). Energy and the [Submitted for review].
environment: electricity customers. 2018. https://www.epa.gov/energy/ [25] Lwanga SK, Lemeshow S. World Health Organization. Sample size determi-
electricity-customers. [Accessed 4 July 2018]. nation in health studies: a practical manual 1991.
[3] Hughes JD. A reality check on the shale revolution. Nature 2013;494:307e8. [26] Oke D, Majozi T, Mukherjee R, Sengupta D, El-Halwagi MM. Simultaneous
https://www.nature.com/articles/494307a. energy and water optimization in shale exploration. Processes 2018;6:86.
[4] Kerr RA. Natural gas from shale bursts onto the scene. Science https://doi.org/10.3390/pr6070086.
2010;328(5986):1624e6. https://doi.org/10.1126/science.328.5986.1624. [27] Elsayed NA, Barrufet MA, El-Halwagi MM. Integration of thermal membrane
[5] Brandt AR, Heath GA, Kort EA. Energy and environment. Methane leaks from distillation networks with processing facilities. Ind Eng Chem Res 2013;53:
North American natural gas systems. Science 2014;343:733e5. https:// 5284e98. https://doi.org/10.1021/ie402315z.
doi.org/10.1126/science.1247045. [28] McDonald J. Some generalized functions for the size distribution of income.
[6] Gao J, Ning C, You F. Data-driven distributionally robust optimization of shale Econometrica 1984;52:647e63. https://doi.org/10.2307/1913469.
gas supply chains under uncertainty. AIChE J 2019;65:947e63. https://doi.org/ [29] Yang L, Grossmann IE, Mauter MS, Dilmore RM. Investment optimization
10.1002/aic.16488. model for freshwater acquisition and wastewater handling in shale gas pro-
[7] Cafaro DC, Grossmann IE. Strategic planning, design, and development of the duction. AIChE J 2015;61:1770e82. https://doi.org/10.1002/aic.14804.
shale gas supply chain network. AIChE J 2014;60:2122e42. https://doi.org/ [30] Energy Information Administration (EIA). Natural gas. 2019. https://www.eia.
10.1002/aic.14405. gov/dnav/ng/NG_PRI_SUM_A_EPG0_PIN_DMCF_M.htm. [Accessed 20 July
[8] Gao J, You F. Optimal design and operations of supply chain networks for 2019].
water management in shale gas production: MILFP model and algorithms for
the water-energy nexus. AIChE J 2015a;61:1184e208. https://doi.org/

You might also like