Management Systems

in 2021, Volume 29, Issue 1, pp. 44-53

Production Engineering

Date of submission of the article to the Editor: 07/2020

Date of acceptance of the article by the Editor: 10/2020
DOI 10.2478/mspe-2021-0006

SUSTAINABLE COAL SUPPLY CHAIN MANAGEMENT USING

EXERGY ANALYSIS AND GENETIC ALGORITHM

Reihaneh NADERI, Mohsen SHAFIEI NIKABADI

Semnan University, Semnan

Akbar ALEM-TABRIZ
Shahid Beheshti University, Tehran

Mir Saman PISHVAEE

School of Industrial Engineering, Iran University of Science and Technology, Tehran

Abstract:
Environmental threats of coal usage in the electricity production combined with the consumption of renewable
and non-renewable resources had led to worldwide energy challenges. The cost of coal mining and economical
and environmentally sustainable usage of mined coal could be optimized by efficient management of coal supply
chain. This paper provides a mathematical model for improving coal supply chain sustainability including the cost
of exergy destruction (entropy). In the proposed method, exergy analysis is used to formulate the model consid-
ering not only economic costs but also destructed exergy cost, while genetic algorithm is applied to efficiently
solve the proposed model. In order to validate the proposed methodology, some numerical examples of coal
supply chains are presented and discussed to show the usability of the proposed exergetic coal supply chain model
and claim its benefits over the existing models. According to the results, the proposed method provides 17.6%
saving in the consumed exergy by accepting 2.7% more economic costs. The presented model can be used to
improve the sustainability of coal supply chain for either designing new projects or upgrading existing processes.

Key words: coal supply chain, exergy, genetic algorithm, sustainability

INTRODUCTION However, the amount of coal used to produce electricity

Ecosystems provide the basis for development of human is expected to rise due to growth in demands [7]. With
community and sustainable industries. Energy supply is a such an increasingly competitive coal industry, it is highly
fundamental component of industrial processes which required to develop efficient methods to minimize the
widely impacts ecosystems, degrades them, and gener- overall cost for the sustainable coal supply chain manage-
ates environmental pollutions. Therefore, minimization of ment (mining, transportation, and power plant), while
energy consumption is a main step toward environmental meeting required environmental standards and regula-
protection. Moreover, the consumed energy should be tions [8].
supplied from sustainable resources [1]. The efficiency Considering environmental risks and impacts of the coal
and quality of the energy supply can be modelled using mining, the performance of the whole process is one of
exergy analysis [2]. Exergy is a thermodynamic concept the major concerns for the involved parties [9]. In addition
used to measure the performance of processes aims at in- to the economic costs of coal supply chain (e.g., miners,
creasing the efficiency of energy utilization and quantify- washing plants, transporters, and power plants), there are
ing the magnitude of losses [3]. other costs related to environmental impacts and pollu-
In spite of recent developments to extract energy from re- tions which should be taken into account within the total
newable sources [4], coal fired power plants will continue operational costs of the system [8]. Any industrial process
to be operational for next decades to meet the worldwide including the mitigation of indirect costs such as environ-
energy demand [5]. Coal is the most important compo- mental impacts, is known as a sustainable process. Sus-
nent for electricity production. Approximately 53% of the tainability in coal supply chain is more complex as it suf-
electricity produced in USA is provided by coal, while 92% fers from other implications such as ethical and social
of extracted coal is used for electricity production [6]. risks.
R. NADERI et al. – Sustainable Coal Supply Chain Management… 45

Environmental impacts of the coal usage in power gener- method and comparison with existing techniques are pro-
ation, high production costs, and more importantly, envi- vided in Section 5. Finally, Section 6 concludes the paper
ronmentally sustainable coal usage has been discussed in following with future research directions.
many researches [2, 10, 11, 12, 13, 14, 15, 16]. Man-Zhi et
al. [11] proposed a method based on object-oriented petri LITERATURE REVIEW
net (OPN) to model coal supply chain operation flow. Yang The main waste gas emissions are CO2, SO2, NOX and
et al. [2] have extended a framework to quantify pur- smoke dust that can damage Ozon layer, make acid rain
chased resources and pollutant emissions of industrial and contribute to the global warming [17]. According to
production processes by integrating the resource, eco- the global pollution statistics published by International
nomic, and environmental factors in a coal industry. Ani- Energy Agency at 2015, over 43% of SO2, 13% of NOX, 39%
okete et al. [10] discussed the environmental impacts of of CO2 and 7% of the fine particulate matter come from
the coal fly ash, waste industrial brine, and waste cooking the coal [18]. Not only coal burning, but also coal mining
oil within coal supply chain to achieve sustainable coal results in overburden waste in air and water resources.
supply chain. Pan et al. [13] illustrated the existing chal- Therefore, studying about the environmental implications
lenges in coal supply chain. They provided some technical of coal production and consumption is an important issue
information about environmental effects and pollution [19].
from the coal mining, preparation, to the final conversion There are different steps in coal supply chain from mining
for power generation. Thomas et al. [14] examined a re- raw coal to delivering it to the real consumers. These
source constrained production planning and scheduling steps can be summarized in Fig. 1. Coal mining operations
problem within a coal supply chain. They considered mul- comprise blasting, cutting, drilling, loading, ventilation,
tiple independent producers with a single linking con- drainage, lighting, and portaging [20]. The coal prepara-
straint. They presented a mathematical model and solved tion includes coal washing and some processing steps
it using Lagrangian relaxation. such as size reducing of the mined coal and removing
rocks, ash-forming materials, and ultrafine coal. Commer-
Our contributions cial coal logistics involve sales department, transportation
Although many researches have been proposed for sus- and warehousing at customer centers. These centers are
tainable coal supply chain management, the researchers linked via logistics flow and capital flow [21].
who used the exergy concept, mostly have considered the
resource side of life cycle assessment (LCA) and paid less
attention to economic investment and environmental im-
pacts of coal supply chain. The main purpose of this paper
is to reduce the destructed exergy in the coal supply chain
model, in order to assist coal supply chain decision makers
toward more sustainable processes. To achieve this pur-
pose, a new mathematical model for sustainable coal sup-
ply chain management considering economic and ex-
ergetic costs is formulated using exergy analysis. In order
to efficiently solve the established NP-hard model, genetic
algorithm (GA) is utilized. Our contributions in this paper Fig. 1 Interactions between different steps of coal supply chain
can be summarized as follows: Source: based on [13].
− Modeling of sustainable coal supply chain manage-
ment considering the destructed exergy (entropy) to Coal cleaning process may vary due to the characteristics
remark the importance of disorders and environmen- of the raw coal, available equipment, and the required
tal aspects on achieving more sustainability. quality of the cleaned coal. In order to separate coal from
− Formulation of the sustainable coal supply chain the ash, different technologies are available, which most
model as a multi-objective optimization problem in of them are based on the gravity disparity of coal and ash.
terms of economic and exergetic cost. Some use flotation concept which is based on the dispar-
− Utilizing genetic algorithm to optimize the established ity in surface properties of coal and ash [22]. The choice
model. of cleaning technology has implications on the total per-
− Encoding of feasible solutions using a combined bi- formance of the coal supply chain, i.e., resources, eco-
nary-integer structure to simultaneously optimize dif- nomic, and environmental (exergetic) costs.
ferent size/type decision variables.
− Performance evaluation of the proposed method on RESEARCH METHODOLOGY
three numerical examples. In the majority of previous studies within the area of coal
The rest of the paper is organized as follows. Problem def- supply chain management, the intention of the cost is
initions are described in section 2. The mathematical workflow related cost elements and not the indirect costs
model and solution method are presented in Section 3 and disorders. In this paper, the destructed exergy (en-
and 4, respectively. Simulation results of the proposed tropy) cost is also included into coal supply chain cost
function using exergy analysis. The aim is to establish a
 46 Management Systems in Production Engineering 2021, Volume 29, Issue 1

mutually beneficial relation between enterprise and the first, the coal is extracted, and then, coal processing is
resource, economic and environmental performance. To done at mining sites. During the coal preparation step, the
achieve this purpose, the sustainable coal supply chain coal is processed and cleaned from unwanted materials
management is considered as a multi-objective optimiza- such as dirt, ash, sulfur, and rock. These processes in-
tion problem comprising eight objective functions for crease the heating value of the coal. The prepared coal is
minimization of economic costs as well as exergetic cost. transported to warehouses to be delivered to target cus-
The list of the indices, parameters, and decision variables tomers. Although in some cases, there are demands for
of the proposed sustainable coal supply chain model can raw coal which should be directly delivered from the
be summarized in Table 1. mine; this scenario is not considered at the proposed
model. The goal of this model is to satisfy customers’ de-
Coal supply chain model mand, balance the flow at the network facilities, fulfilling
The supply chain model used in this paper is a four-stage the capacity constraints, and meeting the logical con-
coal supply chain include mines, washing plants, ware- straints.
houses, and customers, which can be seen in Fig. 2. At
Table 1
List of indices, parameters, and decision variables
Index Definition
i Mine index, i = 1,2,3,…,I
j Coal preparation center (washing plant) index, j = 1,2,3,…, J
p Product index, p = 1,2,3,…, P
w Warehouse index, w = 1,2,3,…, W
k Customer index, k = 1,2,3,…, K

Fixed cost of coal mine

Parameter Definition

Fixed cost of coal washing plant 

fi

Fixed cost of coal warehouse 

fj

Raw coal capacity of mine
of product 

fw

Coal washing plant  capacity of product 

αpi

Coal warehouse  capacity of product 

αpj

Unit raw coal cost of product  supplied by mine

αpw

̅ pi Unit market purchase price of raw coal of product  supplied by mine

γpi

Unit coal washing cost of product  manufactured by coal washing plant 

Unit coal warehouse cost of product  at warehouse 
γpj

Unit holding cost of product  at washing plant 

γpw

Unit holding cost of product  at warehouse 

hpj

Unit freight rate from mine
to coal washing plant 

hpw

Unit freight rate from coal washing plant  to warehouse 

τij

Unit freight rate from warehouse  to customer 

τjw

Distance between mine
to coal washing plant 

τwk

Distance between coal washing plant  to warehouse 

rij

Distance between warehouse  to customer 

rjw

Demands of customer  for product 

rwk

Unit selling price of product  to customers

Dpk

Unit market price of the product 

Cp

Inventory of product  at washing plant  after product transferring to warehouses

Cp0

Inventory of product  at warehouse  after product transferring to customers

Ipj
Ipw
∂ 1 if entropy is accounted; 0 otherwise
Exdest Amount of destroyed (wasted) exergy
Sgen Entropy generated

ash of commercial coal  produced by coal washing plant 

T0 Environmental temperature

Lowest rate to meet the demand of customer  for the product 

SHpj
Epk

Amount of coal  which is sold to the customer 

Fpk Minimum requirements for coal ash
Qpk
Decision Variable Definition
Yi 1 if mine i is open; 0 otherwise
Yj 1 if washing plant j is open; 0 otherwise
Yw 1 if warehouse w is open; 0 otherwise
Xpij Amount of raw coal transported from mine i to coal washing plant j for product p
Xpjw Amount of product p transported from coal washing plant j to warehouse w
Xpwk Amount of product p transported from warehouse w to customer k
R. NADERI et al. – Sustainable Coal Supply Chain Management… 47

  "
'( #( '( *(+
#() ,(+
*
(4)
Therefore, according to the Guoy-Stodola theorem [28]

equal to &  , and can be expressed as:

and the heat pump model [4], the destroyed exergy is

- &   & %& . ", /
#( *(+
#() *(+
(5)

Multi-objective optimization function

By extending Eq. (5) considering the process model and
Fig. 2 System boundary of coal supply chain model applying the mentioned assumptions, economic cost Cos-
tEC comprising seven elements can be expressed as fixed
Exergy modelling operation costs at mines (Eq. 6), washing plants (Eq. 7),
Based on the first thermodynamic law, energy can be and warehouses (Eq. 8), inventory cost at washing plants
transformed but it is always conserved. However, exergy and warehouses (Eq. 9), and variable transportation costs
is consumed in any process as entropy is generated (sec- between mines and washing plants (Eq. 10), washing
ond thermodynamic law) [23]. Exergy quantifies the use- plants and warehouses (Eq. 11), and warehouses and cus-
fulness of energy, and can be defined as the quality of en- tomers (Eq. 12). Therefore, multi-objective function, Ob-
ergy, available energy or available work [24]. According to jFun1, considering economic factors can be formulated as

where 
is the consumed exergy (MJ),  is the temper-
[25], the consumed exergy can be expresses as Eq. (1),
0  ∑2 12 32
Eq. (13).

ature of surroundings (K), and is entropy (MJ/K).

0  ∑4 14 34
(6)


  ∑ ∆
05  ∑6 16 36
(7)
(1)

07  ∑&8∑4 ℎ&4 :&4 34 ; ∑6 ℎ&6 :&6 36 <

(8)
Based on the second thermodynamic law, all natural or

0=  ∑& ∑2 ∑48&2 ; >24 ?24 <@&24 32

man-made processes suffer from inevitable shortfalls, in- (9)

0A  ∑& ∑4 ∑68&4 ; >46 ?46 <@&46 34

efficiencies and destructed exergy accounts due to the (10)

0B  ∑& ∑6 ∑C 8&6 ; >6C ?6C <@&6C 36

gradually generated entropy [26]. The destructed exergy (11)
in a system can be calculated as the difference between (12)
the total exergy entering the system and the total exergy Min: ObjFun1 
∑2 12 32 ; ∑4 14 34 ; ∑6 16 36 ; ∑&8∑4 ℎ&4 :&4 34 ; ∑6 ℎ&6 :&6 36 <
rem [28], the exergy destruction
 can be computed O; ∑& ∑2 ∑48&2 ; >24 ?24 <@&24 32 ; ∑& ∑4 ∑68&4 ; >46 ?46 <@&46 34 P
leaving the system [27]. According to Guoy-Stodola theo-
(13)
; ∑& ∑6 ∑C8&6 ; >6C ?6C <@&6C 36
 as
from a knowledge of the entropy generated in the system
The objective function ObjFun1 of Eq. (13) considers only

   .  (2) the workflow of related costs. To be able to include the
indirect process costs including the losses due to the dis-
As mentioned in [1], the relationship between the vendor
orders in the process (entropy) to the optimization prob-
and buyer can be modelled as two series heat pumps,
lem, the destructed exergy cost CostDE of the system can
while the price is analogues to temperature. Moreover,
be expressed based on the exergy analysis as:
0Q  ∑& ∑2 ∑C R & %&C . " /  32
#( *(+
work (money) is generated by heat pump (vendor). The
#() *̄ (+
required work to operate a production process includes (14)
the cost of energy, physical equipment, labor, and trans- As a result, the multi-objective function of the sustainable
portation. As inefficiencies result in wastage of the re- coal supply chain model, ObjFun2 considering both eco-
sources, exergy analysis can be effectively used to en-
Min: ObjFun2 
nomic and exergetic factors can be expressed as:
hance the efficiency of the system, and consequently, im-
∑2 12 32 ; ∑4 14 34 ; ∑6 16 36 ; ∑&8∑4 ℎ&4 :&4 34 ; ∑6 ℎ&6 :&6 36 <
⎧ ⎫
prove both economic and environmental performances.
⎪; ∑& ∑2 ∑48&2 ; >24 ?24 <@&24 32 ; ∑& ∑4 ∑68&4 ; >46 ?46 <@&46 34 ⎪
pressed as Eq. (3), where  is the heat energy extracted ; ∑& ∑6 ∑C8&6 ; >6C ?6C <@&6C 36
In summary, entropy in a thermal system can be ex-
⎨ ⎬
(15)
from the colder environment at temperature  , and ⎪; ∑ ∑ ∑ R  % . #( " *(+/  3 ⎪
⎩ ⎭
 is the heat energy rejected at temperature  .
& 2 C & &C # *̄ (+ 2
()

  "
 #$
The boundary conditions of the optimization problem can
! !#$
(3) be represented by Eqs. (16) to (25). Equations (16) to (18)
In this paper, the same analogous as in Jawad et al. [1] is are capacity constraints that control the maximum flows.
used to present the entropy (destructed exergy) in coal Equations (19) to (21) define the logistics balance and con-
supply chain. It is assumed that the buyer in the coal sup- straints of the coal supply chain system. It is obvious that
ply chain represents the surroundings, while the market is for each product, the inbound flow should be equal to the
represented as the environment. The wasted energy, raw sum of all outbound flows at each node. Equation (19)
materials, capital, and labor efforts, that may not be re- represents the quantity balance at the production facili-

tropy due to product  can be formulated similar to the

generated, are represented as entropy. The generated en- ties and assure the flow balance between the mines and

thermal system, as Eq. (4), where %& is the total demands

the coal washing plants. Equation (20) is the quantity bal-

for product .
ance of raw coal at washing plants. Equation (21) is re-
lated to the coal sale balance. Equation (22) defines the
 48 Management Systems in Production Engineering 2021, Volume 29, Issue 1

constraints of the customer demands to ensure that all ristic algorithms cannot usually produce an optimal solu-
demands for products are completely fulfilled. Equation tion, as they don’t explore the whole search space, none-
(23) is about the customers quality requirements. Finally, theless it may yield locally optimal solutions in a reasona-
constraints (24) and (25) enforce the binary and non-neg- ble amount of time [34]. However, metaheuristic algo-

∑4
&24 \ ]&2 32 , ∀,

ativity restrictions on the decision variables, respectively. rithms can provide a more robust solution at the expense

∑6
&46 \ ]&4 34 , ∀, 
(16) of increased computational efforts for globally searching
among the search space [35]. Since genetic algorithm (GA)
∑C
&6C \ ]&6 36 , ∀, 
(17)
has proved its efficiency in solving different supply chain
∑2
&24  ∑4
&46 ; ∑4 &4 ∀, 
(18)
problems in different studies with a satisfactory perfor-
∑4
&46  ∑6
&6C ∀, 
(19)
mance [36, 37, 38, 39], we utilize GA to solve the proposed

&C  ∑6
&6C , ∀, 
(20) model for sustainable coal supply chain management in-

%&C &C \ ∑6
&6C \ &C ∀, 
(21) cluding the economic and exergetic objectives.

∑4
&46 &4 \ &C 0&C
(22) The flowchart of the proposed GA can be seen in Fig. 3.

32 , 34 , 36 ∈ a0,1c
(23)

&24 ,
&46 ,
&6C d 0
(24)
(25)
In order to solve the problem and find the minimum eco-
nomic and exergetic costs of the sustainable coal supply
chain model, ObjFun2 of Eq. (15) should be solved to-
gether with fulfilling constrains of Eqs. (16) to (25).

Model assumptions
In the case of gross coal, various types of coal can be ob-
tained, intended for different energy applications. The
selling price of various products to various customers is
different, e.g., due to bonus. Moreover, all facilities and
washing plants are under aging effects that will cause
some additional costs to the supply chain. However, to
simplify the problem, the following assumptions have
been made: Fig. 3 Overall flowchart of the proposed GA
1. There is only one specific product, i.e., prepared coal.
2. The selling price to all customers is assumed to be the The GA procedure starts by generating a random initial
same. population of chromosomes (feasible solutions) according
3. It is assumed that the washing plant is new and the to the encoding scheme (Sec. 4-1), each contains three bi-
aging effects are neglected. nary structures Yi, Yj, and Yw, and three integer structures
4. It is assumed that all washing plants are using the Xpij, Xpjw, and Xpwk. Then, multi-objective function evalua-
same coal washing technology. tion (Sec. 4-2) and population updating (Sec. 4-3) are suc-
cessively done until the stop criterion (completion of the
Solution method based on genetic algorithm maximum iterations) reaches. The detail of the proposed
The complexity of the supply chain design problem is GA is provided in the following.
known to be NP-hard [29]. Therefore, the proposed sus- A feasible solution of the problem can be represented as
tainable coal supply chain combined with the ecological a hybrid binary-integer structure comprising three binary
impact aspects is also a NP-hard problem. The available and three integer matrices. As shown in Fig. 4, decision
techniques for NP-hard problems can be classified into ex- variables of the optimization problem include
act, heuristic, and metaheuristic methods [30]. Exact algo- 1. Choosing a subset of mines to supply raw coals Yi is a
rithms (e.g., branch & bound) are guaranteed to find an binary vector of length I.
optimal solution. However, the size of search space and 2. Selecting a subset of washing plants Yj is a binary vec-
consequently required run-time increases dramatically tor of length J.
with the instance size, and thus, only small instances can 3. Selecting a subset of coal warehouses Yw is a binary
be practically solved using exact methods [31]. Therefore, vector of length W.
the only possibility for real-world applications with me- 4. Amounts of raw coals transported from mines to
dium/large instances is to utilize heuristic or metaheuris- washing plants Xpij is an integer matrix of dimension
tic algorithms. I×J.
Heuristics (e.g., greedy algorithms) are problem-depend- 5. Amounts of products transported from washing plants
ent techniques specifically developed to the problem with to warehouses Xpjw is an integer matrix of dimension
inspiration from available information in system model J×W.
[32]. Greedy algorithms built a solution piece by piece, 6. Amounts of products transported from warehouses to
making the locally optimal choice at each stage [33]. Heu- customers Xpwk is an integer matrix of dimension W×K.
R. NADERI et al. – Sustainable Coal Supply Chain Management… 49

the time complexity of the proposed GA can be simplified

as O(MaxIter×PopSize×I×J×W×K).

RESULTS OF RESEARCH
The proposed model and the solution method based on
GA have been successfully coded in MATLAB R2019b en-
vironment. The experiments were executed on a PC with
Quad Core 2.4 GHz processor and 8 GB RAM running on
windows 10. In order to performance evaluation of the
proposed method, it has been tested on three datasets
considering different complexities. Moreover, to under-
Fig. 4 Encoding of a feasible solution (a chromosome)
stand the effectiveness of the exergy analysis, the model
is solved under two times: one by considering both eco-
At every iteration, each chromosome is decoded accord- nomic and destructed exergy costs of Eq. (15), and an-
ing to Fig. 4, and then, is evaluated according to the multi-
other by considering only the economic costs of Eq. (13)
objective function ObjFun as in Eq. (15). Moreover, all
as objective function of GA.
constraints are checked to be verified or not. As a result, Parameter setting of the GA is summarized in Table 2. To
the overall cost of each chromosome can be calculated as
adjust each controllable parameter of the algorithm, dif-
COST = ObjFun×2PF, where PF is the total number of con-
ferent values were evaluated, and the best value in term
straints of Eqs. (15) to (24) which have not been fulfilled
of convergence speed and total cost was determined for
(penalty function). Population updating includes recombi- final simulations. Population size and maximum iterations
nation, crossover and mutation which produce PR, PC, and
were considered as 50 and 500, respectively. The percent-
PM percentages, respectively. These rates are considered
age of the population achieved by recombination, crosso-
as PR = 10%, PC = 50% and PM = 40%. In recombination ver, and mutation, were set as PR = 10%, PC = 50%
phase, all chromosomes are sorted from the best to the and PM = 40%. Different selection strategies, i.e., Roulette
worst according to their COST, and then, PR% of the best
Wheel Selection (RWS), Tournament Selection (TS), and
solutions are directly transferred to the next generation.
Elitism Parent Selection (EPS) were evaluated. Among
In order to generate an offspring using crossover opera- them, EPS was chosen, because of better performance in
tor, at first, two chromosomes are selected via a parent
terms of total cost and convergence speed. Moreover, bi-
selection strategy, and then, uniform crossover operator
nary swap and integer operators were chosen for the mu-
is applied for each binary or integer structure of the se-
tation in the hybrid binary-integer structure of the algo-
lected parents. In uniform crossover, each gene of the rithm.
structure is transferred from the same gene of parent 1 or
parent 2, each by a probability of 50%. In mutation phase, Table 2
a chromosome is selected by the EPS, and then, a gene is Dimension of coal supply chain datasets
randomly selected and mutated. Based on the hybrid bi- Dataset Washing
Mines Warehouses Customers
nary-integer structure of the solutions, either binary swap # Plants
mutation or integer mutation is performed on the se- 1 5 5 10 10
lected gene of the binary or integer structure. 2 10 15 20 25
3 15 25 30 40
Time complexity analysis
Typically, time complexity of any population-based me- In order to validate the proposed method, three coal sup-
taheuristic algorithm can be expressed as O(Max- ply chain datasets were used with different complexities.
Iter×(PopSize×TCObjFun+TCPopUp)), where PopSize is the pop- Details of these datasets in term of the number of coal
ulation, MaxIter is the maximum number of iterations, mines, washing plants, warehouses, and customers can be
TCObjFun is the time complexity required for the objective seen in Table 2. The model parameters were generated
function evaluation for a single solution at any iteration (a using uniform distributed random values according to
chromosome), and TCPopUp is the time complexity required valid ranges as summarized in Table 3.
for the population updating in one iteration (include re- These parameters are considered based on real ranges of
combination, crossover, and mutation). For complex data provided in Iran Ministry of Industry, Mine and Trade
models including many objectives and constraints (like [40] and Iranian Mines and Mining Industries Develop-
the present model), the time complexity required at every ment and Renovation Organization, IMIDRO [41]. It is
iteration of the algorithm for the population updating can mentioned that annual coal production capacity in Iran is
be neglected against the time complexity of the objective around 1.5 million tons (3 million tons of raw coal). How-
function evaluation. Therefore, the time complexity of the ever, because of improper situation of the market and ab-
proposed GA can be simplified as O(MaxIter×Pop- sence of investment for the purchase of equipment and
Size×TCObjFun). Based on the system model of sustainable lack of preparation of the mines, the actual coal produc-
coal supply chain in Sec. 3-4, one objective function eval- tions had been lower than its capacity in last years.
uation has a time complexity of O(I×J×W×K). Therefore,
 50 Management Systems in Production Engineering 2021, Volume 29, Issue 1

Table 3 Table 4
Details of parameters in the datasets Computational results of different costs (in millions)
Uniform Distribution and CPU time (in seconds)
Parameter
or Rate Optimization of ObjFun2
Optimization of ObjFun1
Demands 10,000-50,000 (ton) Da- (Proposed)
Market Prices 150-250 (USD/ton) taset # CPU CPU
CostEC CostDE CostEC CostDE
Unit market purchasing price Time Time
60-90 (USD/ton)
for raw coal 1 55.7 14.9 105 57.2 12.4 108
Mine capacity of raw coal 100,000-300,000 (ton) 2 105.9 27.1 149 108.6 22 154
Production capacity 50,000-200,000 (ton) 3 162.5 46.9 191 167.3 38.7 197
Capacity of other stages 100,000-300,000 (ton)
Minimum requirements %10 of production In order to capture the convergence speed of the pro-
for coal ash capacity posed GA, the diagram for dataset 2 is provided in Fig. 5.
1,000,000-2,000,000
Fixed cost of coal mines As seen in Fig. 5, the final cost of the GA is achieved as
(USD)
Fixed cost of washing plants 500,000-1,000,000 (USD)
130.6 million USDs, which can be calculated by the sum-
Fixed cost of warehouses 100,000-200,000 (USD) mation of CostEC (108.6 million) and CostDE (22 million).
Unit raw coal mine cost 30-50 (USD/ton)
Unit production cost
100-150 (USD/ton)
at washing plants
Unit variable cost at warehouses 4-7 (USD/ton)
Unit inventory holding cost
3-5 (USD/ton)
at washing plants
Unit inventory holding cost
2-3 (USD/ton)
at warehouses
Unit transportation cost between
1.5 (USD/ton/Km)
mines and washing plants
Unit transportation cost between
2 (USD/ton/Km)
other stages

But according to Deputy for Mine and Mineral Industries

Projects at IMIDRO, some plans and new projects are de-
signed in 2019 for attracting private sector investment in
the construction of coal washing factories and develop-
ment of coal mines. For example there are projects to Fig. 5 Convergence of the proposed GA under ObjFun2
for dataset 2
equip and for construction of coal washing factories in
Savadkouh (Central Alborz), Shahroud (Semnan Province)
In order to better understanding the effect of considering
of Iran with production capacity of 150 and 200 thousand
the destructed exergy into the multi-objective function of
tons [41].
the algorithm, the results of two scenarios are compared
in Table 5 and Fig. 6.
Simulation results
In order to understand the effectiveness of considering
Table 5
the exergetic cost in our model, two scenarios are simu- Effects of economic-exergetic objective function ObjFun2
lated. To achieve this purpose, the result of the economic against ObjFun1
objective function ObjFun1 in Eq. (13) is compared with Additional Eco- Saving in Exergy Additional CPU
that of the proposed economic-exergetic objective func- Dataset
nomic Cost Cost Time
#
tion ObjFun2 in Eq. (15). Because of random nature of Value % Value % Value %
search process in metaheuristic algorithms, the proposed 1 1.5 2.7 2.5 16.4 3 2.8
GA was run 10 times for each dataset under the two sce- 2 2.7 2.5 5.1 18.8 5 3.3
narios. 3 4.8 3 8.2 17.5 6 3.1
The average results (over 10 runs) of the economic cost Average 3 2.7 5.26 17.6 4.67 3.06
(CostEC) and the destructed exergy cost (CostDE) can be
summarized in Table 4. By considering the destructed exergy in the proposed
Although by considering economic-exergetic model of Ob- model, significant saving in the destructed exergy cost
jFun2, economic costs are a little bit more than those of in (CostDE) can be obtained by accepting additional economic
the economic model of ObjFun1, efficient saving in de- cost (CostEC). According to the results in Fig. 6, the pro-
structed exergy costs can be obtained. From time com- posed economic-exergetic model on average can save
plexity of view, considering the proposed economic-ex- 17.6% (5.26 million) in the destructed exergy by consum-
ergetic model increases the CPU running time for only ing 2.7% (3 million) more economic costs.
about 3 to 6 seconds for the different datasets, which can
be ignored.
R. NADERI et al. – Sustainable Coal Supply Chain Management… 51

DISCUSSION
In order to have sustainable supply chain management, it
is of utmost importance to pay more attention to the ex-
ergy removed from the nature and consider the environ-
mental impacts of any product or services especially for
those which create more pollution like coal extraction,
processing and consumption. The obtained results show
the ability of the proposed model to help the supply chain
decision makers to decide about the best choices for hav-
ing a profitable supply chain besides, less destroying the
environment and approaching to more sustainable coal
supply chain. There is an additional economic cost as
Fig. 6 Saving in exergy cost % by accepting additional economic showed above, but the benefit of the model is that it will
cost %
help decision makers to quantify this additional cost
Sensitivity analysis which is necessary for further decisions.
In this section, a sensitivity analysis is provided by chang- Table 7 provides the mass of the material resources, emis-
ing the default values of the system model. To achieve this sions and wastes to produce one tonne of raw coal [42].
purpose, 20% reduction in the variable cost of production, Therefore, by proper optimization of the coal supply
20% reduction in warehouse costs, and 20% reduction in chain, more sustainability and less environmental de-
the coal demands. The obtained results for dataset 2 can struction from renewable and non-renewable resources
be summarized in Table 6, in term of the total cost (Ob- can be obtained.
jFun2) and the eight elements of the objective function.
Table 7
All changes greater than 10% are shown in bold. Accord-
Environmental impacts to produce one tonne of raw coal
ing to the obtained results in Table 6, reduction in produc-
Item Quantity (kg)
tion costs has a great effect on reducing operation costs
Oxygen 185
at washing plants (F2), while reduction in warehouse costs Renewable Resources Freshwater 402
reduces operation costs at warehouses (F3) and inventory Salt 1.6
cost (F4). Reduction in coal demands leads to reduction in Limestone 0.51
more stages of the supply chain: operation costs at wash- Clay 0.0054
ing plants (F2), transportation costs between washing Sand 0.087
plants and warehouses (F6), transportation costs between Non-Renewable Resources Gravel 0.047
warehouses and customers (F7), and destructed exergy Natural gas 0.13
Hard coal 1230
cost (F8). As a result, by 20% reduction in coal demands,
Oil 4
the total cost is reduced from 130.6 million to 116.5 mil- Co2 0.41
lion, and consequently, 10.8% of the total cost of the sus- Emissions Nox 0.29
tainable coal supply chain can be saved. So2 2.48
Ash 0.32
Table 6
Solid Wastes Slag 0.26
Sensitivity analysis for dataset 2 for different cases
Gypsum 98.1
Sensitivity Analysis under
Default
20% 20% 20%
Values CONCLUSION
Cost Function Reduction Reduction Reduction
(in mil- To design a new coal supply chain process or optimize an
in Produc- in Ware- in Coal
lions) existing one, some decisions should be made such as
tion Costs house Costs Demands
F1 (Eq. 6) 12.35 12.2 12.6 11.6 choosing the number, capacity, and the technology used
F2 (Eq. 7) 5.12 5.27 for different network facilities. The current study has
6.11 6.07
(-16.2%) (-13.7%) proved that exergy concept can be effectively used to for-
F3 (Eq. 8) 2.21 mulate the total cost of the system including the de-
2.5 2.48 2.31
(-11.6%) structed exergy. It can provide an insight about the poten-
F4 (Eq. 9) 1.13
1.26 1.23 1.32 tial of environmental destruction saving per unit of addi-
(-10.3%)
F5 (Eq. 10) 22.4 20.8 22.72 20.7
tional economic costs. It would be extremely beneficial at
F6 (Eq. 11) 52.4 the time of business case calculation for the new projects
58.3 56.8 57.2 or modification and upgrading the current coal supply
(-10.1%)
F7 (Eq. 12) 4.66 chain, in order to minimize economic costs while protect
5.7 5.53 5.27
(-18.3%) the environment.
F8 (Eq. 14) 18.2 In this paper, it was assumed that there is only one spe-
22 22.4 21.8
(-17.3%) cific product (i.e., prepared coal). However, in the case of
Total Cost (Eq. 15) 130.6 126.5 129 116.5 gross coal, various types and assortments of coal can be
Reduction
N/A 3.1% 1.2% 10.8% obtained. Moreover, the selling price was considered to
in Total Cost (%)
be the same for all customers. It was also assumed that all
 52 Management Systems in Production Engineering 2021, Volume 29, Issue 1

facilities are new that is possible only in the case of new [14] A. Thomas, J. Venkateswaran, G. Singh and M. Krishna-
projects and large mining investments in completely new moorthy. “A resource constrained scheduling problem
coal basins. As a future work, the proposed model can be with multiple independent producers and a single linking
constraint: A coal supply chain example”. European Jour-
extended to various types of coal products with different
nal of Operational Research, vol. 236(3), pp. 946-956,
selling prices to different customers, considering aging ef-
2014.
fects of facilities. Another suggestion for further improve- [15] H. Jawad, M.Y. Jaber, M. Bonney and M.A. Rosen “Deriving
ment of this model is to investigate the impact of machin- an exergetic economic production quantity model for bet-
eries aging on pollution and the impact of different tech- ter sustainability”. Appl. Math. Model. vol. 40, pp. 6026-
nologies on the consumed exergy. Using the real data of 6039, 2016.
the coal mines in developing countries and investigate the [16] A. Baral, B.R. Bakshi, and R.L. Smith. “Assessing resource
exergy destructions would be another interesting topic intensity and renewability of cellulosic ethanol technolo-
for future study. gies using Eco-LCA”. Environmental science & technology,
vol. 46(4), pp. 2436-2444, 2012.
[17] I. Manisalidis, E. Stavropoulou, A. Stavropoulos and E. Be-
REFERENCES
zirtzoglou. “Environmental and health impacts of air pollu-
[1] H. Jawad, M.Y. Jaber and R.Y. Nuwayhid. ”Improving sup-
tion: A review”. Frontiers in public health, vol. 8, 2020.
ply chain sustainability using exergy analysis”. European
[18] International Energy Agency, Data and Statics:
Journal of Operational Research, Vol. 269(1), pp. 258-271,
https://www.iea.org/data-and-statistics. July. 01, 2020
2018.
[July. 01, 2020].
[2] S. Yang and Y. Qian. “The inclusion of economic and envi-
[19] M. Mann and P. Spath. “A life cycle assessment of biomass
ronmental factors in the ecological cumulative exergy con-
cofiring in a coal-fired power plant”. Clean Products and
sumption analysis of industrial processes”. Journal of Cle-
Processes, vol. 3(2), pp. 81-91, 2001.
aner Production, vol. 108, pp. 1019-1027, 2015.
[20] J. Bijańska and K. Wodarski. “Model of process manage-
[3] A. Muchtar. “Preliminary Analysis of Single-Flash Geother-
ment system in enterprises of the hard coal mining indus-
mal Power Plant by Using Exergy Method. Case Study:
try”. Management Systems in Production Engineering, vol.
Ulubelu Geothermal Power Plant-Indonesia”. Interna-
28(2), pp. 112-120, 2020.
tional Journal of Renewable Energy Research (IJRER), vol.
[21] C. Wang, D. Mu. “An LCA study of an electricity coal supply
8(3), pp. 1685-1696, 2018.
chain”. Journal of Industrial Engineering and Manage-
[4] A.A. Ojo, O. Awogbemi, and A.O. Ojo. “An Overview of the
ment, vol. 7(1), pp. 311-335, 2014.
Exploitation of Renewable Energy Resources in Nigeria,
[22] S. Bhagwat, X. Zhang and H. Fan. “Estimation of coal clean-
South Africa, and the United Kingdom”. International Jour-
ing costs: a spreadsheet based interactive software for use
nal of Renewable Energy Research (IJRER), vol. 10(2), pp.
in estimation of economically recoverable cost reserves”.
843-861, 2020.
US Geological Survey Professional. pp. 1-13, 2009.
[5] S. Nyquist. “Energy 2050: Insights from the ground up”.
[23] M.E. Bösch, S. Hellweg, M.A. Huijbregts and R.
McKinsey & Company, 2016.
Frischknecht. “Applying cumulative Exergy demand (CExD)
[6] F. Freme. “US Coal supply and demand: 2009 review”. Elec-
indicators to the ecoinvent database”. The International
tric Power, Vol. 922(937.8), pp. 946-8, 2009.
Journal of Life Cycle Assessment, vol. 12(3), pp. 181-190,
[7] US Energy Information Administration (Ed.). (2011). An-
2007.
nual Energy Outlook 2011: With Projections to 2035.
[24] A. Vadiee and M. Yaghoubi, “Exergy Analysis of the Solar
Government Printing Office.
Blind System integrated with a Commercial Solar Green-
[8] J. Phillips, (2008). Modeling the US Coal Supply Chain. Col-
house,” International Journal of Renewable Energy Re-
orado School of Mines. Retrieved from http://dahl. mines.
search, vol. 6, no. 3, 2016.
edu/coalphillips. pdf,(last accessed in June 2012).
[25] J. Szargut. Exergy method: technical and ecological appli-
[9] S. Mehmood, B.V. Reddy and M.A. Rosen. “Exergy analysis
cations. WIT press, vol. 18, 2005.
of a biomass co-firing based pulverized coal power gener-
[26] R. Leutz, Nonimaging Fresnel Lenses: Design and Perfor-
ation system”. International journal of green energy, vol.
mance of Solar Concentrators, Springer Berlin Heidelberg,
12(5), pp. 461-478, 2015.
2001.
[10] T. Aniokete, M. Ozonoh, and M.O. Daramola. “Synthesis of
[27] A.J. Mahmood. “An Experimental Study on Energy and Ex-
Pure and High Surface Area Sodalite Catalyst from Waste
ergy for Glazed and Unglazed Solar System with Perforated
Industrial Brine and Coal Fly Ash for Conversion of Waste
Absorber Plate and Wire Mesh Layers”. International Jour-
Cooking Oil (WCO) to Biodiesel”. International Journal of
nal of Renewable Energy Research (IJRER), vol. 9(4), pp.
Renewable Energy Research (IJRER), vol. 9(4), pp. 1924-
1901-1911, 2019.
1937, 2019.
[28] A. Bejan. Entropy Generation Minimization: The Method of
[11] L. Man-Zhi, Z. Mei-Hua, L. Xue-Qing, and Y. Ji-Xian. “The
Thermodynamic Optimization of Finite-Size Systems and
research on modeling of coal supply chain based on objec-
Finite-Time Processes, CRC Press, 1995.
toriented Petri net and optimization”. Procedia Earth and
[29] T. Loukil, J. Teghem, and D. Tuyttens. “Solving multi-objec-
Planetary Science, vol. 1(1), pp. 1608-1616, 2009.
tive production scheduling problems using metaheuris-
[12] A. Thomas, J. Venkateswaran, G. Singh and M. Krishna-
tics”. European journal of operational research, vol.
moorthy. “A resource constrained scheduling problem
161(1), pp. 42-61, 2005.
with multiple independent producers and a single linking
[30] M. Shokouhifar, A. Jalali. “Simplified symbolic transfer
constraint: A coal supply chain example”. European Jour-
function factorization using combined artificial bee colony
nal of Operational Research, vol. 236(3), pp. 946-956,
and simulated annealing”. Applied Soft Computing, vol. 55,
2014.
pp. 436-451, 2017.
[13] L. Pan, P. Liu, L. Ma and Z. Li. “A supply chain based assess-
ment of water issues in the coal industry in China”. Energy
Policy, vol. 48, pp. 93-102, 2012.
R. NADERI et al. – Sustainable Coal Supply Chain Management… 53

[31] Z.M. Zahedi, R. Akbari, M. Shokouhifar, F. Safaei and A. Ja- [37] Y.B. Woo and B.S. Kim. “A genetic algorithm-based me-
lali. “Swarm intelligence based fuzzy routing protocol for taheuristic for hydrogen supply chain network problem
clustered wireless sensor networks”. Expert Systems with with two transportation modes and replenishment cy-
Applications, vol. 55, pp. 313-328, 2016. cles”. Computers & Industrial Engineering, vol. 127, pp.
[32] K. Sorensen. “Metaheuristics-the metaphor exposed”. In- 981-997, 2019.
ternational Transactions in Operational Research, vol. [38] A. Rostami, M. M. Paydar and E. Asadi-Gangraj. “A Hybrid
22(1), pp. 3-18, 2015. Genetic Algorithm for Integrating Virtual Cellular Manufac-
[33] P. Festa. “A brief introduction to exact, approximation, and turing with Supply Chain Management Considering New
heuristic algorithms for solving hard combinatorial optimi- Product Development”. Computers & Industrial Engineer-
zation problems”. In 2014 16th International Conference on ing, 2020.
Transparent Optical Networks (ICTON) pp. 1-20, 2014. [39] H. Gholizadeh and H. Fazlollahtabar. “Robust Optimization
[34] V. Haleh and F. Imam Ibrahim. “Feature Selection Meth- and modified genetic algorithm for a closed loop green
ods: Genetic Algorithms vs. Greedy-like Search”. In Proc. supply chain under uncertainty: Case study in Melting In-
Int. Conf. Fuzzy Intell. Control Syst pp. 1-10, 2005. dustry”. Computers & Industrial Engineering, 2020.
[35] F. Fanian, V.K. Bardsiri and M. Shokouhifar. “A new task [40] Iran Ministry of Industry, Mine and Trade:
scheduling algorithm using firefly and simulated annealing https://en.mimt.gov.ir. July. 10, 2020 [July. 10, 2020].
algorithms in cloud computing”. International Journal of [41] Iranian Mines & Mining Industries Development & Reno-
Advanced Computer Science and Applications, vol. 9 (2), vation Organization: http://imidro.gov.ir/general_con-
pp. 195-202, 2018. tent/2634-coal.html, July. 10, 2020 [July. 10, 2020].
[36] A. Saghaeeian, and R. Ramezanian. “An efficient hybrid ge- [42] G. Luo, J. Zhang, Y. Rao, X. Zhu and Y. Guo. “Coal Supply
netic algorithm for multi-product competitive supply chain Chains: A Whole-Process-Based Measurement of Carbon
network design with price-dependent demand”. Applied Emissions in a Mining City of China”. Energies, vol. 10 (11),
Soft Computing, vol. 71, pp. 872-893, 2018. pp. 1855, Nov. 2017; https://doi.org/10.3390/
en10111855.

Reihaneh Naderi
Semnan University, Semnan
Faculty of Economics, Management
and Administration Sciences
Industrial Management Department, Iran
e-mail: [email protected]

Mohsen Shafiei Nikabadi*

ORCID ID: 0000-0002-9744-960X
Semnan University, Semnan
Faculty of Economics, Management
and Administration Sciences
Industrial Management Department, Iran
e-mail: [email protected]
Tel.: +989125404808, Fax: +982331532579
*corresponding autor

Akbar Alem-Tabriz
Shahid Beheshti University, Tehran
Faculty of Management and Accounting
Industrial Management Department, Iran
e-mail: [email protected]

Mir Saman Pishvaee

School of Industrial Engineering
Iran University of Science and Technology, Tehran, Iran
e-mail: [email protected]