Professional Documents
Culture Documents
A Risk Assessment Framework of PPP Waste-To-Energy Incineration Projects in China Under 2-Dimension Linguistic Environment
A Risk Assessment Framework of PPP Waste-To-Energy Incineration Projects in China Under 2-Dimension Linguistic Environment
a r t i c l e i n f o a b s t r a c t
Article history: Risk assessment is critical to insuring the success of waste-to-energy (WTE) projects under public
Received 4 April 2017 eprivate partnership (PPP) model as risks are accompanied with the life-cycle of the projects. Uncer-
Received in revised form tainty composed of fuzziness and randomness is an important issue in the risk assessment under
21 January 2018
increasingly complex environments. The 2-dimension linguistic information is a powerful tool to express
Accepted 7 February 2018
Available online 14 February 2018
the fuzziness of information, and meanwhile, the cloud model perfectly depicts the randomness of in-
formation with three numerical characteristics. Combining them together, this paper puts forward a risk
assessment framework of PPP WTE incineration projects. Firstly, an evaluation index system for risk
Keywords:
Risk assessment
assessment of China's PPP WTE incineration projects is constructed, which includes 14 criteria involved
PPP in four dimensions of construction and operation risk, macro-economic risk, legal and socio-political risk,
WTE and government risk; secondly, a conversion model between 2-dimension linguistic variables and clouds
2-dimension linguistic information is developed; thirdly, the cloud Choquet integral (CCI) operator is utilized to handle the correlations
Cloud model among risk factors; fourthly, a risk assessment framework is established to assess the overall risk level of
several China's PPP WTE incineration projects. From the empirical demonstration and comparison
analysis, the proposed framework can effectively deal with such a complicated problem and lead to an
outstanding result.
© 2018 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.jclepro.2018.02.077
0959-6526/© 2018 Elsevier Ltd. All rights reserved.
Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617 603
dimension linguistic variables, CCI operator as well as the model of incineration projects more realistic. With the rapid development
converting 2-dimension linguistic variables into clouds. Section 3 of the cloud model theory, successful applications have been car-
constructs an evaluation index system for the risk assessment of ried out in various fields, such as intelligent control (Javed et al.,
PPP WTE incineration projects in China. Section 4 establishes a risk 2017), algorithm improvement (Wu et al., 2015), edge detection
assessment framework of PPP WTE incineration projects. Section 5 (Zhao et al., 2017) and water eutrophication assessment (Yan et al.,
assesses and compares the risk of several PPP WTE incineration 2017), etc. In particular, after (Wang and Feng, 2005) introduced
projects in China. Section 6 makes a comparison between the the conversion between linguistic variables and clouds, the cloud
proposed method and the existing method. Section 7 draws the model has also been applied to the field of MCDM (Peng and
conclusions, contributions, limitations and outlooks. Wang, 2017; Wu et al., 2016a, 2016b). The concept of cloud
model is as follows:
2. Methodology Definition 2 (Li et al., 2009) Let U be the universe of discourse
and T a linguistic variable inU. If xðx2UÞ is a random instantiation
2.1. 2-Dimension linguistic variable of linguistic variable T, which satisfiesEn0 NðEn;
2 02
He Þ,x NðEx; En Þ and the certainty degree of x belonging to
2
Most factors for the risk assessment of PPP WTE incineration ðxExÞ
linguistic variable T satisfies m ¼ e 2ðEn0 Þ2 .
projects are qualitative rather than quantitative. A common prac-
Then the distribution of x in the universe U is called a normal
tice is to ask several authoritative experts to express their opinions
cloud. The cloud model can effectively integrate the randomness
based on their knowledge and experience. Compared with the
and fuzziness of concepts and describe the overall quantitative
linguistic variable, the 2-dimension linguistic variable is more
property of a concept by the three numerical characteristics as
suitable for the problem of PPP WTE incineration projects risk
follows:
assessment. The linguistic variable assumes that all the experts are
fully confident with their results. However, some experts may not
Expectation Ex is the mathematical expectation of the cloud
be very sure about the results given by themselves since PPP WTE
drops belonging to a linguistic variable in the universe and is the
incineration projects risk assessment is conducted in an uncertain
most representative and typical sample of the linguistic
environment. Clearly it will lead to information distortion and loss
variable;
if the reliability of the given evaluation result is ignored. The 2-
Entropy En represents the fuzziness measurement of a linguistic
dimension linguistic variable add an additional information,
variable, which is determined by both the randomness and the
which describes the self-assessment of the expert on the reliability
fuzziness of the linguistic variable;
of the given evaluation result, based on the linguistic variable.
Hyper entropy He is the uncertain degree of entropy En, which
There is no doubt that the application of 2-dimension linguistic
reflects the dispersion of the cloud drops.
variable can increase the reliability and accuracy of the risk
assessment on PPP WTE incineration projects. Usage of the 2-
As for aggregating the clouds, how to select a properly aggre-
dimension linguistic variable in MCDM is very popular (Liu et al.,
gation operator is a key problem. Most of the existing operators for
2016; Wu et al., 2017b). The concept of 2-dimension linguistic
aggregating the decision information of each criterion, such as
variable is as follows:
weighted averaging (WA) operator, ordered weighted averaging
Definition 1 (Liu, 2012). Let b s ¼ ðsi ; hj Þ, where si is I class lin-
(OWA) operator, and ordered weighted geometric averaging
guistic information that represents expert's judgment to an eval-
(OWGA) operator, are based on the classical additive measure of
uated object, and si is the element from pre-defined linguistic
linear aggregation operators, assuming that the decision criteria are
assessment set SI ¼ ðsl ; /; 0; /sl Þ, while hj is II class linguistic independent and uncorrelated. As a matter of fact, there are surely
information that represents the subjective evaluation on the reli- some correlations among the decision criteria involved in PPP MSW
ability of their given results, and hj is the element from pre-defined incineration projects risk assessment. For instance, mistakes in
linguistic assessment set HII ¼ ðht ; /; 0; /ht Þ,then bs is called a 2- government decision-making may lead to improper project loca-
dimension linguistic variable. tion, which in turn may worsen surrounding environment and
Example 1. The 2-dimension linguistic variable b s ¼ ðs1 ; h1 Þ is cause severe opposition from local residents (Song et al., 2013).
adopted by an expert to evaluate a PPP WTE incineration project on Thus, Choquet integral with respect to 2-order additive fuzzy
a certain criterion. The pre-defined linguistic assessment sets are measure is a powerful tool for it can effectively reduce computing
SI ¼ ðs2 ¼ very high; s1 ¼ high; s0 ¼ fair; s1 ¼ low; s2 ¼ complexity and describe the correlations among the criteria.
very lowÞands. Thus, the I class linguistic information “very high” Recently, a CCI operator, which combines the cloud model and
represents the risk level of the PPP WTE incineration project on the Choquet integral with respect to the 2-order additive fuzzy mea-
criterion, and the “partial confident” reflects the evaluation on the sure, is proposed by Wu et al. (2016a). Accordingly, the CCI operator
reliability of the expert's given risk level. is utilized to model the correlations among decision criteria
involved in PPP WTE incineration projects risk assessment, which is
2.2. Cloud model and CCI operator presented as follows:
Definition 3 (Wu et al., 2016a). Let Yi ¼ ðExi ; Eni ; Hei Þ be a
Using the 2-dimension linguistic variable to reflect the experts’ collection of clouds, and m is the fuzzy measure on Y ¼ fY1 ; Y2 ;
preferences is one way to depict the fuzziness and increase the /Yn g. The mapping CCIm : Y n /Yis defined as a CCIm operator:
evaluation quality, and the other one to improve the reasonability
of risk assessment is to characterize the randomness of uncer- n
X
tainty. As mentioned earlier, the risk assessment of PPP WTE CCIm ðYa1 ; Ya2 ; /Yan Þ ¼ Yðai Þ Yðai1 Þ m AðiÞ (1)
incineration projects is conducted under the uncertain environ- i¼1
ment. Randomness is an important property of uncertainty due to
the fact that conditions cannot determine outcomes sometimes. where ($) indicates a permutation on Y such that Yða1 Þ Yða2 Þ /
Fortunately, the cloud model can characterize the randomness Yðan Þ andAðiÞ ¼ fYðai Þ ; /; Yðan Þ g.
effectively, which makes the risk assessment of PPP WTE Example 2. Assuming that the performances of a PPP WTE
Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617 605
CCIm ðYa1 ; Ya2 ; /Yan Þ ¼ 50 1 þ ð55 50Þ 0:75 þ ð60 55Þ*0:2;
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
52 1 þ 42 þ 52 0:75 þ 32 þ 42 0:2; 0:22 1 þ 0:22 þ 0:12 0:75 þ 0:12 þ 0:12 0:2
The classical model proposed by (Wang and Feng, 2005) has kis given beforeHe0 ¼ 3k , and.Hei ¼ Hei ¼ Hei1 =0:618;
jþtþ1
made a great contribution to the application of cloud model.
1 i l:
However, it cannot deal with the linguistic variables with the form
To show the superiority of the new model, a comparison be-
of 2-dimension linguistic information. In this section, we define a
tween the proposed model and the classical model proposed by
new model for converting the 2-dimension linguistic variable set
(Wang and Feng, 2005) is made. In this comparison, three 2-
into a series of adjacent clouds. Because the Entropy (En) and Hyper
entropy (He) of a cloud are determined by both the randomness and dimension linguistic variables bs 1 ¼ ðs1 ; h1 Þ ¼ ðhigh;
the fuzziness of the linguistic variable expressed by experts, this partial confidentÞ, bs 2 ¼ ðs1 ; h0 Þ ¼ ðhigh; confidentÞ
conversion is deduced by the principle that the higher the reli- andbs 3 ¼ ðs1 ; h1 Þ ¼ ðhigh; fully confidentÞ are transformed by the
ability of the results (hj ) given by experts, the smaller the Entropy new model while a linguistic variables1 ¼ high is converted by the
(En) and Hyper entropy (He) are. classical model (Wang and Feng, 2005). The four corresponding
Generally, there are two ways in generating clouds. One is on the clouds Y1 , Y2 , Y3 and Y4 are calculated as follows:
basis of the 3s principle of the normal cloud, and the other utilizes
the golden ratio which originates from Fibonacci series (Liu and Liu, Y1 ¼ ð69:1; 19:1; 4:85Þ; Y2 ¼ ð69:1; 9:55; 2:43Þ; Y3 ¼ Y4
2016). The application of the former one is limited since the
¼ ð69:1; 6:37; 1:62Þ:
determination of parameters requires abundant experiments or
some subjective methods. By comparison, the latter method is The linguistic variable implies that the experts are fully confi-
more objective and convenient to operate. Since it plays an dent about their given results, thus the cloudY3 , transformed by 2-
important role in transformation between linguistic terms and dimension linguistic variableb s 3 ¼ ðs1 ; h1 Þ ¼ ðhigh; fully confidentÞ,
clouds, the golden ratio has been widely applied in the decision- is the same as the cloud converted by the linguistic varia-
making field (Wang et al., 2014b, 2015). Then, applying the bles1 ¼ high in the case of t ¼ 1.
golden ratio, the new model is defined as follows: A graphical representation of the comparison result is shown in
Definition 4. Let bs ¼ ðsi ; hj Þ be a 2-dimension linguistic variable, Fig. 2.
si and hj are the element from linguistic assessment set SI ¼ ðsl ; / It is clear from the graph that the coverage and discrete degree
; 0; /sl Þ and HII ¼ ðht ; /; 0; /ht Þ, respectively. The corresponding of these clouds has obvious difference. Since Ex1 ¼ Ex2 ¼ Ex3 ¼
three numerical characters of 2l þ 1 clouds are calculated as Ex4 , En1 > En2 > En3 ¼ En4 and He1 > He2 > He3 ¼ He4 , obviously,
follows: the ranking result of these clouds is Y3 ¼ Y4 > Y1 > Y2 , which is in
line with the cognition that the same results with different re-
(i) Calculate Ex: liabilities are discrepant. Also, the more confident with the results
given by experts are, the more valuable the results would be.
To illustrate the effectiveness of the proposed model concisely,
Ex0 ¼ ðXmin þ Xmax Þ=2; Exl ¼ Xmin ; Exl ¼ Xmax the advantages and disadvantages of the existing models and the
proposed model are summarized and shown in Table 1. As can be
seen from the table, the proposed model overcomes some short-
Xmax þ Xmin comings of other models, such as not corresponding to the human
Exi ¼ Ex0 þ 0:382i ðl 1Þ
2 cognitive habits, ignoring the reliability of the given evaluation
result, and failing to handle the randomness. Therefore, the quality
of risk assessment can be improved largely based on the above
Xmax þ Xmin improvements. Whereas, the disadvantages of the proposed model
Exi ¼ Ex0 0:382i ðl 1Þ; ð1 i l 1Þ are also obvious. That is, its calculation process is slightly compli-
2
cated and the decision cost is high. However, compared with the
loss of the decision-making mistake, the increasing decision cost
(ii) Calculate En: and the complex calculation process may seem fairly trifling.
606 Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617
Fig. 2. Clouds comparison. The cloudsY1 ,Y2 andY3 are transformed by the 2-dimension linguistic variablesbs 1 ¼ ðs1 ; h1 Þbs 2 ¼ ðs1 ; h0 Þ, andbs 3 ¼ ðs1 ; h1 Þ respectively, while the
cloudY4 is converted by the linguistic variables1 .
Table 1
The advantages and disadvantages of the existing models and the proposed model.
Description Linguistic variable (Zadeh, 2-Dimension linguistic variables Cloud model (Li et al., Model converting linguistic The proposed model
1975) (Zhu et al., 2009) 1995) variable to cloud model
(Wang and Feng, 2005)
Advantages It accords with the human It accords with the human It can depict the It accords with the human It accords with the human
cognitive habits, and can cognitive habits, and can depictfuzziness and cognitive habits, and can cognitive habits, at the same time
depict the fuzziness. The the fuzziness and considers the randomness depict the fuzziness and considers the reliability of the
calculation process is simultaneously. The
reliability of the given evaluation randomness simultaneously. given evaluation result. And it can
simple and the decision calculation process is
result. The calculation process is Moreover, the decision cost depict the fuzziness and
cost is low. simple simple and the decision is low. randomness simultaneously.
cost is low.
Disadvantages It ignores the reliability of It fails to handle the randomness, It is not in accordance It ignores the reliability of Its calculation process is slightly
the given evaluation result and the decision cost is high. with the human the given evaluation result. complicated and the decision cost
and fails to handle the cognitive habits, and And its calculation process is is high.
randomness, ignores the reliability of slightly complicated
the given evaluation
result
3. Evaluation index system of China's PPP WTE incineration projects are identified from the analyzed literature, as Table 2
projects risk assessment shows.
Phase 2. Identifying CRFs of China's PPP WTE projects by case
PPP projects involve a large number of uncertain risks (Xu et al., study analysis.
2010), and CRFs identification is an essential phase in the project The case study analysis draws on documented lessons and ex-
risk management process (Hwang et al., 2013). In this section, the periences from projects that were stalled or terminated, completed
CRFs for the risk assessment of PPP WTE incineration projects are or in operation, which provides an effective means to identify and
identified and classified, and then the classified CRFs are integrated understand factors that contribute to failure of the studied projects
into an evaluation index system. Given that there is little study on (Ameyaw and Chan, 2015). In this phase, a thorough analysis of 12
the risk management of PPP WTE incineration projects, a four- PPP WTE incineration projects in China is conducted, and 11 risk
phase procedure is adopted to establish the evaluation index sys- factors are identified from the analyzed cases, as shown in Table 3.
tem for PPP WTE incineration projects in this paper. Phase 3. Selecting the CRFs by questionnaire survey.
Phase 1. Identifying risk factors of various PPP infrastructure After conducting the comprehensive literature review and case
projects by literature review. study analysis, a total of 39 (9 risk factors are duplicate in the above
Considering that some risk factors are shared among various two steps) risk factors are gathered. Risk is generally characterized
countries and PPP infrastructure projects, literature concerning by the severity of an adverse consequence that can result from an
with risk identification of international PPP infrastructure projects action and the occurrence probability of the given adverse conse-
is reviewed in this phase. A large number of studies have attempted quence (Liu et al., 2010). A questionnaire survey method is adopted
to identify risk factors in various PPP infrastructure projects, such as to select the CRFs among the 39 identified risk factors, which has
water supply projects (Ameyaw and Chan, 2015), transport projects been used frequently in the PPP risk management studies (Ameyaw
(Garrido et al., 2017), green building projects (Zhao et al., 2016) and and Chan, 2015). The 39 identified risk factors are formulated into a
WTE incineration projects (Xu et al., 2015). Risks identified in 15 questionnaire for a survey. The objective of the questionnaire is to
literature about PPP infrastructure projects between 2007 and 2016 require respondents to estimate the severity and the occurrence
are first marked down, and then a total of 37 risk factors for PPP
Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617 607
Table 2
Risk factors associated with various PPP infrastructure projects.
a
No. Risk factors Reference Total
1 2 3 4 5 6 7 8 9 10
1 Government intervention ✓ ✓ ✓ ✓ ✓ ✓ 6
2 Force majeure ✓ ✓ ✓ ✓ ✓ ✓ 6
3 Change in tax ✓ ✓ ✓ ✓ ✓ ✓ 6
4 Financing and refinancing risk ✓ ✓ ✓ ✓ ✓ ✓ 6
5 Change in interest rate ✓ ✓ ✓ ✓ ✓ 5
6 Nationalization or expropriation ✓ ✓ ✓ ✓ ✓ 5
7 Legal and policy risk ✓ ✓ ✓ ✓ ✓ 5
8 Delay in approval and permits ✓ ✓ ✓ ✓ ✓ 5
9 Organization and coordination risk ✓ ✓ ✓ ✓ ✓ 5
10 Government corruption ✓ ✓ ✓ ✓ ✓ 5
11 Environment risk ✓ ✓ ✓ ✓ ✓ 5
12 Construction cost overrun ✓ ✓ ✓ ✓ 4
13 Inflation risk ✓ ✓ ✓ ✓ 4
14 Technical risk ✓ ✓ ✓ ✓ 4
15 Design or construction deficiencies ✓ ✓ ✓ ✓ 4
16 Unforeseen weather or geotechnical conditions ✓ ✓ ✓ ✓ 4
17 Land acquisition risk ✓ ✓ ✓ ✓ 4
18 Change in foreign exchange ✓ ✓ ✓ ✓ 4
19 Public opposition risk ✓ ✓ ✓ ✓ 4
20 Weak regulation system ✓ ✓ ✓ ✓ 4
21 Residual value risk ✓ ✓ ✓ ✓ 4
22 Low operation productivity ✓ ✓ ✓ ✓ 4
23 Prolonged construction period ✓ ✓ ✓ ✓ 4
24 Improper distribution of responsibilities and rights ✓ ✓ ✓ 3
25 Faulty demand forecasting ✓ ✓ ✓ 3
26 Improper operation and management ✓ ✓ ✓ 3
27 Payment risk ✓ ✓ ✓ 3
28 Infrastructures support lack risk ✓ ✓ ✓ 3
29 Government credit risk ✓ ✓ ✓ 3
30 Government decision-making risk ✓ ✓ ✓ 3
31 Slumping financial market ✓ ✓ ✓ 3
32 Industrial regulatory change ✓ ✓ 2
33 Early termination ✓ ✓ 2
34 Contract change risk ✓ ✓ 2
35 Poor contract design ✓ ✓ 2
36 Unstable government ✓ ✓ 2
37 Lack of PPP experience ✓ 1
a
References: 1 ¼ Medda (2007); 2 ¼ Maslyukivska (2007); 3 ¼ Xu et al. (2010); 4 ¼ Ke et al. (2010); 5 ¼ Hwang et al. (2013); 6 ¼ Ameyaw and Chan (2015); 7 ¼ Chou and
Pramudawardhani (2015); 8 ¼ Xu et al. (2015); 9 ¼ Zhao et al. (2016); 10 ¼ Sastoque et al. (2016).
Table 3
PPP WTE incineration cases and risk factors.
a
No. Risk factors Projects Total
probability of each risk factor. In order to get a high-quality result of Workers who have at least three years working experience in
questionnaire, it is necessary to select potential respondents. the PPP WTE incineration plants or thermal power plants
Considering that the number of professionals involved in the construction.
implementation of MSW incineration projects in China is relatively Project managers who have managed at least one PPP WTE
small, and the similarity between thermal power plant and WTE incineration project or thermal power plants construction.
incineration plant, the following three conditions are designed to Scholars who have published at least one paper in international
identify qualified respondents for this questionnaire: journals related to the PPP projects risk management.
608 Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617
The opinions of these qualified respondents provide an insight in Fig. 4. A risk assessment committee is set up consisting of deci-
to risks in PPP WTE incineration projects. Moreover, a 7-point scale sion makers, risk managers of different level, inquiry panel, risk
is defined for this evaluation (where 1 ¼ extremely low; 2 ¼ very evaluation panel and calculation panel. It is inevitable that the
low; 3 ¼ low; 4 ¼ fair; 5 ¼ high; 6 ¼ very high and 7 ¼ extremely views of these members of panel are different. So, the Delphi
high). Through visiting some completed MSW incineration plants technique, developed by the RAND Corporation during the 1950s
and thermal power projects as well as sending E-mail to scholars, a and 60s, is adopted since a consensus among them is needed. The
total of 200 questionnaires are sent out, and feedback from 31 key elements are: 1) anonymous responses by experts to multiple
workers, 5 project managers and 17 scholars are received, yielding a rounds of formal questionnaires; 2) an exercise incorporating
response rate of 26.5%, which meets the norm of 20e30% of most iterative, controlled feedback with respect to the information pro-
questionnaire surveys in the construction industry (Akintoye, vided at each round; and 3) summary of the group's responses. The
2000). Then, a mean score ranking technique, which is a normal technique is designed to minimize the influence of dominant in-
technique used to analyze the results obtained by questionnaire dividuals, group pressure, and irrelevant communication and to
surveys, is employed to select the CRFs. reduce noise.
Firstly, the mean score (MS) for each risk factor with respect to The corresponding tasks of different roles have been stated
their occurrence probability and severity degree is calculated by Eq. explicitly during the three phases, which will greatly improve
(2) (Chan and Kumaraswamy, 1996) management efficiency and quality in practice.
P
ðf sÞ Stage 1. Determining several feasible projects.
MS ¼ ; ð1 MS 7Þ (2)
N
First of all, the decision makers screen out a number of projects
where s is the score given to each risk factor by the respondents,
in the most promising provinces based on the information of the
ranging from 1 to 7; f is the frequency of each rating for each risk MSW amount and electricity demand in various provinces of China.
factor; andN is the total number of responses concerning a partic- Then, according to feasible reports and combining with the budget
ular risk factor and expected return of the firm, they further choose the projects
Secondly, after calculating the probability and severity mean which are economically feasible. The aim of this stage is to elimi-
score, the ‘Impact’ of a risk factor can be calculated by Eq. (3) (Chan nate inferior projects and keep a number of “qualified” sites. The
et al., 2014) benefits would be twofold. On the one hand, the complexity and
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi difficulty of decision-making can be reduced partly; on the other
Impact ¼ probability severity (3) hand, nonsense and useless work can be avoided.
Thirdly, the risk factors can be ranked based on their ‘Impact’
values. The ‘Impact’ values and ranking results are shown in Table 4. Stage 2. Constructing the evaluation index system.
The risk factors with normalized impact values 0:5are deemed as
‘CRF’. In the second stage, two tasks need to be completed: 1) gath-
Phase 4. Constructing the evaluation index system. ering the risk factors for PPP WTE incineration projects through
In this phase, the 14 selected CRFs are divided into four cate- literature review and case study analysis; 2) selecting CRFs from the
gories according to their relation, including the construction and gathered risk factors via questionnaire survey. The questionnaire
operation risk, the macro-economic risk, the legal and socio- designed by the inquiry panel aims to collect the experience of
political risk and the government risk. Then, the evaluation index project managers and workers as well as the knowledge of scholar.
system of China's PPP WTE incineration projects risk assessment is
constructed and shown in Fig. 3. Stage 3. Ranking the projects based on the proposed method.
Table 4
Risk factor with normalized impact values. 0:5
a
CRFs Probability Severity Impact Ranking Normalization
Macro-economic risk
risk assessment
Fig. 3. Evaluation index system for risk assessment of China's PPP WTE incineration projects.
Experts from the evaluation panel are responsible to assess the (Wang et al., 2014a) is utilized to this aggregation process.
performance of each project on each risk criterion. Before the ex- Let Yi ðExi ; Eni ; Hei Þ ði ¼ 1; 2; :::; nÞ be a collection of
perts assign the 2-dimension linguistic variable, generally it is clouds,u ¼ ðu1 ; u2 ; :::; uk Þ be the weight vector of experts and
Pk
i¼1 ui ¼ 1. Then their collective values with the use of the CWAA
necessary to set an appropriate linguistic assessment set in advance
(Liu and Yu, 2014). is shown as follows (Wang et al., 2014a).
The calculation tasks in following steps are undertaken by the
calculation panel.
X rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k Xn 2 Xn 2
CWAAu ðY1 ; Y2 ; :::; Yn Þ ¼ u Ex ;
i¼1 i i i¼1 i
u Eni ; i¼1 i
u Hei (4)
Step 2. Converting the 2-dimension linguistic variables into Step 4. Calculating performance scores of each project on each
clouds Yi ðExi ; Eni ; Hei Þ by using the new conversion model. criterion.
Using the new conversion model as stated in definition 4 to The Technique for Order Preference by Similarity to Ideal Solu-
convert the decision-making information in the form of 2- tion (TOPSIS) method is employed to calculate the performance
dimension linguistic variables b
s ¼ ðsi ; hj Þ, into clouds, denoted scores. The basic principle of TOPSIS is that the chosen alternative
asYi ðExi ; Eni ; Hei Þ. should have the shortest distance from positive-ideal solution and
farthest from the negative-ideal solution. Firstly, let Yi ðExi ; Eni ; Hei Þ
Step 3. Aggregating the evaluation values of each expert into a ði ¼ 1; 2; :::; nÞ be a collection of clouds, and the positive and
collective value. negative ideal clouds can be obtained by Eq. (5) as follows:
After the 2-dimension linguistic variables are converted to yþ ¼ ðmaxExi ; minEni ; minHei Þ; y ¼ ðminExi ; maxEni ; maxHei Þ
clouds, an aggregation operator must be performed for a collective
(5)
value. The cloud weighted arithmetic averaging (CWAA) operator
610 Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617
Stage 1
Determining several
feasible projects Identifying prospective WTE Feasibility report
incineration projects Firm strategy
Stage 2
Constructing a evaluation Literature review
index system Identifying the risk factors
Case study analysis
Questionnaire
Inquiry Selecting the CRFs
panel survey
Stage 3
Ranking the projects
based on the proposed Assigning the 2-dimension linguistic
Evaluation
method variables of projects on each criteria panel
Then, using the normal cloud generator (Li et al., 2009) to Step 5. Aggregating the collective values of each criterion into
calculate the size of clouds. Next, the performance score of each one result.
project can be determined by Ri, which is defined as follows:
Because there are correlations among the risk criteria, the CCI
operator is utilized in this aggregation process. First, the low-level
risk managers determine the weights of the criteria and the cor-
si s
R*i ¼ (6) relation coefficient among the criteria. Then, the fuzzy densities are
sþ s
deduced by the calculation panel. Finally, the collective values of
where si is the size of each cloud, while sþ and s are the sizes of each criterion are aggregated into one result by utilizing the CCI
positive and negative ideal clouds respectively. The larger R*i is, the operator.
better the project will be.
Step 6. Ranking the projects.
The objects of this step are to: 1) help decision makers to further
understand the performance of several feasible projects on each
Using Eqs. (5) and (6), overall performance scores of each
risk factor, and in which way decision makers can make a more
project is calculated based on the idea of the TOPSIS, and then the
rational selection; and 2) provide guidance for policy makers to
projects are ranked according to the overall performance scores.
take corresponding measures to eliminate or mitigate the identi-
fied high risks (as far as is reasonable) in an effort to draw more
Step 7. Selecting the optimal one.
potential investors.
Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617 611
Fig. 5. (a) MSW collected amount of various provinces; (b) Electricity demand quantity of various provinces (Data source: Web of Nation Bureau of Statistics of China: http://www.
stats.gov.cn/).
The high-level risk managers analyze the advantages and dis- Table 6
advantages of projects, and feedback the result to decision makers. 2-Dimension linguistic information given by the expertE1 .
The decision makers then determine the optimal PPP WTE incin- Criteria A1 A2 A3 A4
eration project through analyzing and discussing the evaluation C11 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h0 Þ
result. C12 ðs1 ; h1 Þ ðs2 ; h0 Þ ðs0 ; h0 Þ ðs1 ; h1 Þ
C13 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ
C14 ðs0 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ
5. Empirical demonstration C15 ðs2 ; h0 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ
C16 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ ðs0 ; h1 Þ
C21 ðs1 ; h1 Þ ðs2 ; h0 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ
With the Chinese government actively supporting the WTE
C22 ðs2 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h0 Þ
incineration industry, a risk investment firm located in Beijing, C31 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ
China is going to invest in WTE incineration projects. The MSW C32 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs0 ; h1 Þ ðs1 ; h1 Þ
amount and electricity demand quantity of ten Chinese represen- C41 ðs0 ; h0 Þ ðs1 ; h0 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ
tative provinces are reviewed (Fig. 5a and b). Considering that there C42 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h0 Þ ðs1 ; h1 Þ
are large amounts of MSW and high electricity demand in Guang- C43 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ
C44 ðs2 ; h1 Þ ðs1 ; h0 Þ ðs2 ; h1 Þ ðs0 ; h0 Þ
dong and Jiangsu province, the decision makers of this firm hold
that the two provinces are ideal for WTE business.
The risk evaluation panel includes three experts Ek ðk ¼ 1; 2; 3Þ
whose academic backgrounds are project management and ther- weighting (SAW) technique is employed to obtain the overall score
moelectricity fields. After consulting the decision makers, three of the potential projects. Finally, a ranking result is got according to
economic indicators are selected to screen these projects that are their overall scores.
total investment, internal rate of return and payback period. After After the project selection, the top four projects are screened
that, the weights of the three indicators are calculated as (0.3, 0.3, out, and labeled as A1 ; A2 ; A3 ; A4 . The general engineering charac-
0.4) using analytic hierarchy process (AHP) technique. Then, the teristics of the four projects are given in Table 5.
corresponding values of the potential projects are collected from Firstly, as the criteria involved are all the cost type (a lower value
the approved feasibility study reports. Then, simple additive is better), the I class linguistic set is defined in descending order to
Table 5
General engineering characteristics of the four projects.
A1 A2 A3 A4
Location Huishan district, Wuxi city, Huqiu district, Suzhou city, Jinwan district, Zhuhai city, Longhu district, Shantou city,
Jiangsu province Jiangsu province Guangdong province Guangdong province
Land area (square meters) 235600 187000 346000 209000
Waste disposal 400 233 346 292
capacity(thousand tons/
year)
Installed capacity (MW) 2*12 2*9 1*20 1*15
Total investment (billion RMB) 0.363 0.277 0.428 0.315
Internal rate of return (%) 9.61 8.29 8.62 8.47
Payback period (year) 9.15 8.32 9.27 8.46
612 Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617
simplify the calculation. The I class linguistic set and II class lin-
Table 7
guistic set are shown as follows. 2-Dimension linguistic information given by the expertE2 .
Criteria A1 A2 A3 A4
SI ¼ ðs2 ¼ very high; s1 ¼ high; s0 ¼ fair; s1 ¼ low; s2
C11 ðs1 ; h1 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ
¼ very lowÞ; C12 ðs1 ; h1 Þ ðs2 ; h0 Þ ðs0 ; h0 Þ ðs1 ; h1 Þ
C13 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ
C14 ðs2 ; h1 Þ ðs0 ; h1 Þ ðs1 ; h0 Þ ðs2 ; h1 Þ
HII ¼ ðh1 ¼ partial confident; h0 ¼ familiar confident; h1 C15 ðs2 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h0 Þ
C16 ðs2 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h0 Þ
¼ fully confidentÞ
C21 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h0 Þ ðs2 ; h0 Þ
C22 ðs2 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h0 Þ
The performances of projects on each criterion are assessed by
C31 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ
the three experts with 2-dimension linguistic variables, as shown C32 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ
in Tables 6e8 respectively. C41 ðs1 ; h1 Þ ðs1 ; h0 Þ ðs2 ; h0 Þ ðs1 ; h1 Þ
Secondly, in this case U ¼ ½0; 100, k ¼ 1. The 2-dimension lin- C42 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ
guistic variables given above are all transformed into clouds, as C43 ðs2 ; h1 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ ðs1 ; h0 Þ
C44 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ
shown in Tables 9e11 respectively.
Thirdly, according to the knowledge background of the three
experts, their weight vector is determined as ð0:35; 0:3;0:35Þ. Then Table 8
the evaluation information of each expert is aggregated into a 2-Dimension linguistic information given by the expertE3 .
collective value, as shown in Table 12. Criteria A1 A2 A3 A4
Fourthly, the performance of the four projects on each criterion
C11 ðs2 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ
is calculated by utilizing Eqs. (5) and (6), which is shown in Fig. 6. C12 ðs1 ; h1 Þ ðs1 ; h0 Þ ðs0 ; h1 Þ ðs1 ; h0 Þ
Fifthly, the weights of the criteria are obtained by the AHP C13 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h0 Þ ðs2 ; h0 Þ
method and showed in Table 13. C14 ðs2 ; h1 Þ ðs0 ; h1 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ
The correlation coefficient matrixes are constructed as follows. C15 ðs2 ; h0 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ
C16 ðs2 ; h0 Þ ðs1 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ
C21 ðs2 ; h1 Þ ðs2 ; h0 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ
C22 ðs2 ; h0 Þ ðs2 ; h1 Þ ðs0 ; h1 Þ ðs2 ; h0 Þ
C31 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h0 Þ ðs1 ; h1 Þ
C32 ðs1 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h0 Þ ðs2 ; h1 Þ
C41 ðs1 ; h1 Þ ðs1 ; h0 Þ ðs0 ; h0 Þ ðs1 ; h1 Þ
C42 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs2 ; h0 Þ ðs1 ; h1 Þ
C43 ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ ðs2 ; h0 Þ
C44 ðs2 ; h0 Þ ðs1 ; h1 Þ ðs2 ; h1 Þ ðs1 ; h1 Þ
Then, according to the CCI operator, the C41; C42; C43 and C44
are aggregated intoC4. The corresponding cloud variables of A1
with respect to C4 are shown as follows.
By the normal cloud generator, the sizes of four clouds are listed
in an effective order as below:
Y41 ð25:8302Þ > Y44 ð21:9155Þ > Y42 ð20:2492Þ > Y43 ð6:6126Þ,
then:
Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617 613
ExA1 ðC4Þ ¼ mfC41; C42; C43; C44g Ex43 þ mfC41; C42; C44g ðEx42 Ex43 Þ þ mfC41; C44g ðEx44 Ex43 Þ þ mfC41g ðEx41 Ex44 Þ
¼ 1 21:63 þ 0:7643 ð44:92 21:63Þ þ 0:5429 ð55:73 44:92Þ þ 0:2587 50:96
¼ 44:0632
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
u
u mfC41; C42; C43; C44g En2 þ mfC41; C42; C44g En2 þ En2
u 43
43
42
EnA1 ðC4Þ ¼ t
þmfC41; C44g En244 þ En243 þ mfC41g En241 þ En244
¼ 30:9031
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u mfC41; C42; C43; C44g He2 þ mfC41; C42; C44g He2 þ He2
u 43
43
42
HeA1 ðC4Þ ¼ t
þmfC41; C44g He244 þ He243 þ mfC41g He241 þ He244
¼ 7:7284
Table 9
Clouds transformed by the 2-dimension linguistic information given by the expertE1 .
Criteria A1 A2 A3 A4
C11 ð30:9; 9:55; 2:43Þ ð100; 10:3; 2:62Þ ð30:9; 6:37; 1:62Þ ð69:1; 9:55; 2:43Þ
C12 ð30:9; 6:37; 1:62Þ ð0; 15:45; 3:93Þ ð50; 5:9; 1:5Þ ð30:9; 6:37; 1:62Þ
C13 ð69:1; 6:37; 1:62Þ ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ ð0; 10:3; 2:62Þ
C14 ð50; 3:93; 1Þ ð30:9; 6:37; 1:62Þ ð30:9; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ
C15 ð100; 15:45; 3:93Þ ð30:9; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ
C16 ð69:1; 6:37; 1:62Þ ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ ð50; 3:93; 1Þ
C21 ð69:1; 19:1; 4:85Þ ð100; 15:45; 3:93Þ ð100; 10:3; 2:62Þ ð69:1; 19:1; 4:85Þ
C22 ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ ð30:9; 6:37; 1:62Þ ð100; 15:45; 3:93Þ
C31 ð30:9; 6:37; 1:62Þ ð0; 10:3; 2:62Þ ð69:1; 9:55; 2:43Þ ð69:1; 6:37; 1:62Þ
C32 ð30:9; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ ð50; 3:93; 1Þ ð69:1; 6:37; 1:62Þ
C41 ð50; 5:9; 1:5Þ ð69:1; 9:55; 2:43Þ ð69:1; 9:55; 2:43Þ ð69:1; 6:37; 1:62Þ
C42 ð69:1; 6:37; 1:62Þ ð69:1; 19:1; 4:85Þ ð0; 15:45; 3:93Þ ð30:9; 6:37; 1:62Þ
C43 ð30:9; 6:37; 1:62Þ ð100; 10:3; 2:62Þ ð30:9; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ
C44 ð100; 10:3; 2:62Þ ð69:1; 9:55; 2:43Þ ð100; 10:3; 2:62Þ ð50; 5:9; 1:5Þ
614 Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617
Table 10
Clouds transformed by the 2-dimension linguistic information given by the expertE2 .
Criteria A1 A2 A3 A4
C11 ð30:9; 6:37; 1:62Þ ð69:1; 9:55; 2:43Þ ð69:1; 6:37; 1:62Þ ð100; 10:3; 2:62Þ
C12 ð30:9; 6:37; 1:62Þ ð0; 15:45; 3:93Þ ð50; 5:9; 1:5Þ ð30:9; 6:37; 1:62Þ
C13 ð30:9; 6:37; 1:62Þ ð0; 10:3; 2:62Þ ð69:1; 9:55; 2:43Þ ð69:1; 6:37; 1:62Þ
C14 ð100; 10:3; 2:62Þ ð50; 3:93; 1Þ ð30:9; 9:55; 2:43Þ ð100; 10:3; 2:62Þ
C15 ð0; 10:3; 2:62Þ ð30:9; 6:37; 1:62Þ ð100; 10:3; 2:62Þ ð69:1; 9:55; 2:43Þ
C16 ð100; 10:3; 2:62Þ ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ ð100; 15:45; 3:93Þ
C21 ð69:1; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ ð30:9; 9:55; 2:43Þ ð100; 15:45; 3:93Þ
C22 ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ ð30:9; 6:37; 1:62Þ ð100; 15:45; 3:93Þ
C31 ð30:9; 6:37; 1:62Þ ð100; 10:3; 2:62Þ ð30:9; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ
C32 ð69:1; 6:37; 1:62Þ ð69:1; 19:1; 4:85Þ ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ
C41 ð30:9; 6:37; 1:62Þ ð30:9; 9:55; 2:43Þ ð100; 15:45; 3:93Þ ð30:9; 6:37; 1:62Þ
C42 ð69:1; 6:37; 1:62Þ ð69:1; 19:1; 4:85Þ ð30:9; 6:37; 1:62Þ ð30:9; 6:37; 1:62Þ
C43 ð0; 30:91; 7:85Þ ð69:1; 9:55; 2:43Þ ð69:1; 19:1; 4:85Þ ð69:1; 9:55; 2:43Þ
C44 ð69:1; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ ð0; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ
Table 11
Clouds transformed by the 2-dimension linguistic information given by the expertE3 .
Criteria A1 A2 A3 A4
C11 ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ ð0; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ
C12 ð69:1; 19:1; 4:85Þ ð69:1; 9:55; 2:43Þ ð50; 11:8; 3Þ ð69:1; 9:55; 2:43Þ
C13 ð69:1; 6:37; 1:62Þ ð30:9; 6:37; 1:62Þ ð30:9; 9:55; 2:43Þ ð100; 15:45; 3:93Þ
C14 ð100; 10:3; 2:62Þ ð50; 3:93; 1Þ ð30:9; 9:55; 2:43Þ ð69:1; 6:37; 1:62Þ
C15 ð100; 15:45; 3:93Þ ð30:9; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ ð69:1; 6:37; 1:62Þ
C16 ð100; 15:45; 3:93Þ ð69:1; 6:37; 1:62Þ ð30:9; 6:37; 1:62Þ ð100; 10:3; 2:62Þ
C21 ð100; 30:91; 7:85Þ ð100; 15:45; 3:93Þ ð30:9; 9:55; 2:43Þ ð69:1; 19:1; 4:85Þ
C22 ð0; 15:45; 3:93Þ ð100; 10:3; 2:62Þ ð50; 3:93; 1Þ ð100; 15:45; 3:93Þ
C31 ð69:1; 19:1; 4:85Þ ð0; 10:3; 2:62Þ ð69:1; 9:55; 2:43Þ ð69:1; 6:37; 1:62Þ
C32 ð69:1; 6:37; 1:62Þ ð30:9; 6:37; 1:62Þ ð0; 15:45; 3:93Þ ð0; 10:3; 2:62Þ
C41 ð69:1; 6:37; 1:62Þ ð69:1; 9:55; 2:43Þ ð50; 5:9; 1:5Þ ð69:1; 6:37; 1:62Þ
C42 ð69:1; 6:37; 1:62Þ ð0; 30:91; 7:85Þ ð0; 15:45; 3:93Þ ð69:1; 6:37; 1:62Þ
C43 ð30:9; 19:1; 4:85Þ ð0; 30:91; 7:85Þ ð30:9; 19:1; 4:85Þ ð100; 15:45; 3:93Þ
C44 ð0; 15:45; 3:93Þ ð30:9; 6:37; 1:62Þ ð100; 10:3; 2:62Þ ð69:1; 6:37; 1:62Þ
Table 12
Expert group aggregation.
Criteria A1 A2 A3 A4
C11 ð68:46; 7:97; 2:29Þ ð79:92; 8:87; 2:26Þ ð31:55; 7:97; 2:03Þ ð78:37; 8:83; 2:25Þ
C12 ð44:27; 12:41; 3:15Þ ð24:19; 13:68; 3:48Þ ð50; 8:22; 2:15Þ ð44:27; 7:64; 1:94Þ
C13 ð57:64; 6:37; 1:62Þ ð45:82; 9:12; 2:32Þ ð55:73; 8:57; 2:18Þ ð55:73; 11:53; 2:93Þ
C14 ð82:5; 8:62; 2:19Þ ð43:32; 4:92; 1:25Þ ð30:9; 8:57; 2:18Þ ð78:37; 7:76; 1:97Þ
C15 ð70; 14:1; 3:59Þ ð30:9; 6:37; 1:62Þ ð78:37; 7:76; 1:97Þ ð69:1; 7:47; 1:9Þ
C16 ð89:19; 11:38; 2:9Þ ð89:19; 9:12; 2:32Þ ð55:73; 6:37; 1:62Þ ð82:5; 7:47; 1:9Þ
C21 ð79:92; 21:78; 5:53Þ ð90:73; 13:39; 3:41Þ ð55:09; 9:01; 2:5Þ ð78:37; 18:08; 4:59Þ
C22 ð65; 12:35; 3:14Þ ð79:92; 7:97; 2:03Þ ð37:59; 5:64; 1:43Þ ð100; 15:45; 3:93Þ
C31 ð44:27; 12:41; 3:15Þ ð30; 10:3; 2:62Þ ð57:64; 8:72; 2:22Þ ð69:1; 6:37; 1:62Þ
C32 ð55:73; 6:37; 1:62Þ ð55:73; 11:74; 2:98Þ ð47:5; 10:99; 2:8Þ ð44:92; 7:97; 2:03Þ
C41 ð50:96; 6:05; 1:58Þ ð57:64; 9:55; 2:43Þ ð71:69; 10:66; 2:74Þ ð57:64; 6:37; 1:62Þ
C42 ð44:92; 7:97; 1:62Þ ð44:92; 23:91; 6:07Þ ð9:27; 13:39; 3:41Þ ð44:27; 6:37; 1:62Þ
C43 ð21:63; 20:7; 5:26Þ ð55:73; 19:97; 5:07Þ ð42:36; 15:85; 4:03Þ ð79:92; 11:19; 2:85Þ
C44 ð55:73; 11:53; 2:93Þ ð55:73; 7:64; 1:94Þ ð70; 10:3; 2:62Þ ð62:42; 6:21; 1:58Þ
for this comparison, which is elaborated as follows. The ranking result of the four projects isA4 > A1 > A2 > A3 , which
Step 1. Aggregating the evaluation valuesrijk of each expert by the is the same as the one gained by the proposed method. The cor-
2DLPGWA operator and obtaining the overall values, as shown in rectness of the proposed method in the paper can be proved.
Table 15. Actually, a significant shortcoming of Liu and Yu's method (2014) is
Step 2. Aggregating the overall values of each criterion by the that the 2-dimension linguistic computational model based on the
2DLPGWA operator and obtaining the comprehensive evaluation 2DLPGWA operator utilizes the minimizing operation to aggregate
values, as shown in Table 16. the self-assessment of the experts. In others words, once a certain
Step 3. Calculating the expectations of the comprehensive expert is “partial confident” on his/her own evaluation result, the
evaluation values. result of group aggregation is “partial confident”, as shown in
Table 15. Obviously, it will lead to information distortion and loss.
By contrast, the main advantages of the proposed method mainly
EðA1 Þ ¼ 0:046; EðA2 Þ ¼ 0:022; EðA3 Þ ¼ 0:001; EðA4 Þ ¼ 0:055:
lie in the following:
Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617 615
Table 15
Overall values of the four projects with respect to each criterion.
Criteria A1 A2 A3 A4
Acknowledgements
above three points. Fifthly, a case study in China is carried out to
illustrate the applicability of the proposed framework. The result
Project supported by the Fundamental Research Funds for the
shows the project A4 is the optimal one with minimum risk. Last
Central Universities (NO. 2017XS099, NO. 2016MS70, NO.
but not least, a comparison analysis with an existing method is also
2018ZD14), the 2017 Special Project of Cultivation and Develop-
conducted to prove the correctness and effectiveness of the pro-
ment of Innovation Base (NO. Z171100002217024), the NCEPU
posed method.
“Double First-Class” Graduate Talent Cultivation Program and the
Research Funds of Beijing Social Science (16GLC069).
7.2. Contributions
References
The contributions of this study are multifaceted. Therefore, the
target audiences of this study are extensive. Firstly, this study Akintoye, A., 2000. Analysis of factors influencing project cost estimating practice.
measures the impact of every risk factor quantitatively, and es- Construct. Manag. Econ. 18, 77e89.
Al, S.A., Harik, G., Alameddine, I., Bruschi, D., Garcia, D.A., El-Fadel, M., 2017. Risk
tablishes an evaluation index system for PPP WTE incineration assessment of oil spills along the Mediterranean coast: a sensitivity analysis of
projects which decomposes the objective into several operable the choice of hazard quantification. Sci. Total Environ. 574, 234e245.
evaluation indexes. This contribution helps the risk managers Ameyaw, E.E., Chan, A.P., 2015. Evaluation and ranking of risk factors in
publiceprivate partnership water supply projects in developing countries using
better understand each risk factor, and to improve quality in fuzzy synthetic evaluation approach. Expert Syst. Appl. 42, 5102e5116.
practice. Secondly, this study proposes a conversion model be- Arbulú, I., Lozano, J., Rey-Maquieira, J., 2017. The challenges of tourism to waste-to-
tween the 2-dimension linguistic variables and clouds that com- energy public-private partnerships. Renew. Sustain. Energy Rev. 72, 916e921.
Ayodele, T.R., Ogunjuyigbe, A.S.O., Alao, M.A., 2017. Life cycle assessment of waste-
bines the advantages of both. This contribution will attract
to-energy (WtE) technologies for electricity generation using municipal solid
attention from researchers who study the 2-dimension linguistic waste in Nigeria. Appl. Energy 201, 200e218.
variable or cloud model. Thirdly, the study specifies the tasks of Chan, A.P., Lam, P.T., Wen, Y., Ameyaw, E.E., Wang, S., Ke, Y., 2014. Cross-sectional
analysis of critical risk factors for PPP water projects in China. J. Infrastruct. Syst.
decision makers, risk managers and various expert panels to reduce
21, 1e10.
the duplication of effort and improve efficiency. This will help de- Chan, D.W., Kumaraswamy, M.M., 1996. An evaluation of construction time per-
cision makers of investment firms with the improvement of man- formance in the building industry. Build. Environ. 31, 569e578.
agement efficiency. Fourthly, the study provides not only the Chou, J.-S., Pramudawardhani, D., 2015. Cross-country comparisons of key drivers,
critical success factors and risk allocation for public-private partnership pro-
performance scores of projects under the over goal, but also that jects. Int. J. Proj. Manag. 33, 1136e1150.
under the criteria. This merit can help the decision makers to Cucchiella, F., D'Adamo, I., Gastaldi, M., 2017. Sustainable waste management: waste
analyze the merits and demerits of various projects clearly, so as to to energy plant as an alternative to landfill. Energy Convers. Manag. 131, 18e31.
Dong, J., Chi, Y., Zou, D., Fu, C., Huang, Q., Ni, M., 2014. Ener-
make more reasonable decisions. It also provides enough infor- gyeenvironmenteeconomy assessment of waste management systems from a
mation for local government to raise awareness and prompt cor- life cycle perspective: model development and case study. Appl. Energy 114,
responding action to guard against the potential risks, and thus to 400e408.
Fern andezgonza lez, J.M., Grindlay, A.L., Serranobernardo, F., Rodríguezrojas, M.I.,
attract investors. Zamorano, M., 2017. Economic and environmental review of Waste-to-Energy
systems for municipal solid waste management in medium and small munic-
ipalities. Waste Manag. 67, 360e374.
7.3. Limitations and outlooks Vassallo, J.M., 2017. Is EU financial support
Garrido, L., Gomez, J., Baeza, M.D.L.A.,
enhancing the economic performance of PPP projects? An empirical analysis on
Despite some contributions have been made, there are still some the case of Spanish road infrastructure. Transport Pol. 56, 19e28.
Gierczak, M., 2014. The quantitative risk assessment of MINI, MIDI and MAXI hor-
limitations in this study. Firstly, the risk preferences of decision
izontal directional drilling projects applying fuzzy fault tree analysis. Tunn.
makers are not considered. Since decision makers’ risk preferences Undergr. Space Technol. 43, 67e77.
have a great influence on the decision results, the accuracy of Guo, Y., Meng, X., Wang, D., Meng, T., Liu, S., He, R., 2016. Comprehensive risk
describing the risk preferences is a long-term goal to pursue. Sec- evaluation of long-distance oil and gas transportation pipelines using a fuzzy
Petri net model. J. Nat. Gas Sci. Eng. 33, 18e29.
ondly, it lacks the big data processing capability. With the Hwang, B.-G., Zhao, X., Gay, M.J.S., 2013. Public private partnership projects in
increasing number of projects, the decision information to be Singapore: factors, critical risks and preferred risk allocation from the
processed will increases exponentially. Thus, enhancing the perspective of contractors. Int. J. Proj. Manag. 31, 424e433.
Javed, A., Larijani, H., Ahmadinia, A., Emmanuel, R., Mannion, M., Gibson, D., 2017.
computing ability of the proposed model is admirable. In future Design and implementation of a cloud enabled random neural network-based
research, we are going to introduce some psychological behavior decentralized smart controller with intelligent sensor nodes for HVAC. IEEE
theories, such as prospect theory, is going be introduced to take into Internet of Things Journal 4, 393e403.
Jiang, Y., Nan, Z., Yang, S., 2013. Risk assessment of water quality using Monte Carlo
account the risk preferences of decision makers. Moreover, intelli- simulation and artificial neural network method. J. Environ. Manag. 122,
gent algorithms such as artificial intelligence and deep learning will 130e136.
also be combined to assess risks of thousands of projects. Jin, X.H., Doloi, H., 2008. Interpreting risk allocation mechanism in publiceprivate
partnership projects: an empirical study in a transaction cost economics
The study has a wide range of applicability. It is not limited to perspective. Construct. Manag. Econ. 26, 707e721.
the regions in China, but other areas worldwide are also applicable. Ke, Y., Wang, S., Chan, A.P., Lam, P.T., 2010. Preferred risk allocation in China's
Because the proposed risk assessment framework for PPP WTE publiceprivate partnership (PPP) projects. Int. J. Proj. Manag. 28, 482e492.
Khakzad, N., Khan, F., Amyotte, P., 2011. Safety analysis in process facilities: com-
incineration projects and the proposed transformation model is
parison of fault tree and Bayesian network approaches. Reliab. Eng. Syst. Saf. 96,
universally applicable. The potential inapplicability may lie in the 925e932.
selected critical risk factors since they are largely influenced by the Khan, M.D., Khan, N., Sultana, S., Joshi, R., Ahmed, S., Yu, E., Scott, K., Ahmad, A.,
national conditions and vary from country to country. Therefore, in Khan, M.Z., 2017. Bioelectrochemical conversion of waste to energy using mi-
crobial fuel cell technology. Process Biochem. 57, 141e158.
the practical management problems, the managers should pay Li, D., 2004. Artificial intelligence with uncertainty. In: Computer and Information
more attention to the suitability of the evaluation index system. Technology, 2004. CIT'04. The Fourth International Conference on. IEEE, pp.
Y. Wu et al. / Journal of Cleaner Production 183 (2018) 602e617 617
2e2. 679e681þ685.
Li, D., Haijun, M., Xuemei, S., 1995. Membership cloud and membership cloud Wang, J.-q., Peng, L., Zhang, H.-y., Chen, X.-h., 2014a. Method of multi-criteria group
generators. J. Comput. Res. Dev. 32, 15e20. decision-making based on cloud aggregation operators with linguistic infor-
Li, D., Liu, C., Gan, W., 2009. A new cognitive model: cloud model. Int. J. Intell. Syst. mation. Inf. Sci. 274, 177e191.
24, 357e375. Wang, J.Q., Peng, J.J., Zhang, H.Y., Liu, T., Chen, X.H., 2015. An uncertain linguistic
Liu, J., Martínez, L., Wang, H., Rodríguez, R.M., Novozhilov, V., 2010. Computing with multi-criteria group decision-making method based on a cloud model. Group
words in risk assessment. Int. J. Comput. Intell. Syst. 3, 396e419. Decis. Negot. 24, 171e192.
Liu, P., 2012. An approach to group decision making based on 2-dimension uncer- Wang, J.Q., Peng, L., Zhang, H.Y., Chen, X.H., 2014b. Method of multi-criteria group
tain linguistic information. Technol. Econ. Dev. Econ. 18, 424e437. decision-making based on cloud aggregation operators with linguistic infor-
Liu, P., He, L., Yu, X., 2016. Generalized hybrid aggregation operators based on the 2- mation. Inf. Sci. 274, 177e191.
dimension uncertain linguistic information for multiple attribute group deci- Wu, L., Zuo, C., Zhang, H., 2015. A cloud model based fruit fly optimization algo-
sion making. Group Decis. Negot. 25, 103e126. rithm. Knowl. Base Syst. 89, 603e617.
Liu, P., Liu, X., 2016. Multi-attribute group decision-making method based on cloud Wu, Y., Chen, K., Zeng, B., Yang, M., Geng, S., 2016a. Cloud-based decision framework
distance operators with linguistic information. Int. J. Fuzzy Syst. 19, 1e14. for waste-to-energy plant site selectioneA case study from China. Waste
Liu, P., Yu, X., 2014. 2-Dimension uncertain linguistic power generalized weighted Manag. 48, 593e603.
aggregation operator and its application in multiple attribute group decision Wu, Y., Li, L., Xu, R., Chen, K., Hu, Y., Lin, X., 2017a. Risk assessment in straw-based
making. Knowl. Base Syst. 57, 69e80. power generation public-private partnership projects in China: a Fuzzy Syn-
Maslyukivska, O., 2007. European infrastructure procurement through PPP. Man- thetic Evaluation analysis. J. Clean. Prod. 161, 977e990.
agement Procurement & Law 160, 159e167. Wu, Y., Liu, L., Gao, J., Chu, H., Xu, C., 2017b. An extended VIKOR-based approach for
Medda, F., 2007. A game theory approach for the allocation of risks in transport pumped hydro energy storage plant site selection with heterogeneous infor-
public private partnerships. Int. J. Proj. Manag. 25, 213e218. mation. Information 8, 106.
Nie, Y., 2008. Development and prospects of municipal solid waste (MSW) incin- Wu, Y., Yang, M., Zhang, H., Chen, K., Wang, Y., 2016b. Optimal site selection of
eration in China. Front. Environ. Sci. Eng. China 2, 1e7. electric vehicle charging stations based on a cloud model and the PROMETHEE.
Okoro, U., Kolios, A., Lin, C., 2017. Multi-criteria risk assessment approach for Method. Energies 9, 157.
components risk rankinge the case study of an offshore Wave Energy Con- Xu, Q., Ge, J., 2011. Reduction of CO2 emission using bioreactor technology for waste
verter. International Journal of Marine Energy 17, 21e39. management in China. Energy Procedia 5, 1026e1031.
Papapostolou, A., Karakosta, C., Nikas, A., Psarras, J., 2017. Exploring opportunities Xu, Y., Chan, A.P., Xia, B., Qian, Q.K., Liu, Y., Peng, Y., 2015. Critical risk factors
and risks for RES-E deployment under Cooperation Mechanisms between EU affecting the implementation of PPP waste-to-energy projects in China. Appl.
and Western Balkans: a multi-criteria assessment. Renew. Sustain. Energy Rev. Energy 158, 403e411.
80, 519e530. Xu, Y., Yeung, J.F., Chan, A.P., Chan, D.W., Wang, S.Q., Ke, Y., 2010. Developing a risk
Peng, H.G., Wang, J.Q., 2017. Cloud decision model for selecting sustainable energy assessment model for PPP projects in Chinada fuzzy synthetic evaluation
crop based on linguistic intuitionistic information. Int. J. Syst. Sci. 48, 1e18. approach. Autom. ConStruct. 19, 929e943.
Sastoque, L.M., Arboleda, C.A., Ponz, J.L., 2016. A proposal for risk allocation in social Yan, H., Wu, D., Huang, Y., Wang, G., Shang, M., Xu, J., Shi, X., Shan, K., Zhou, B.,
infrastructure projects applying PPP in Colombia. Procedia Engineering 145, Zhao, Y., 2017. Water eutrophication assessment based on rough set and
1354e1361. multidimensional cloud model. Chemometr. Intell. Lab. Syst. 164, 103e112.
Song, J., Song, D., Zhang, X., Sun, Y., 2013. Risk identification for PPP waste-to- Zadeh, L.A., 1975. The concept of a linguistic variable and its application to
energy incineration projects in China. Energy Pol. 61, 953e962. approximate reasoning. Inf. Sci. 8, 199e249.
Song, J., Sun, Y., Jin, L., 2017. PESTEL analysis of the development of the waste-to- Zhang, D.Q., Tan, S.K., Gersberg, R.M., 2010. Municipal solid waste management in
energy incineration industry in China. Renew. Sustain. Energy Rev. 80, China: status, problems and challenges. J. Environ. Manag. 91, 1623e1633.
276e289. Zhao, L., Dong, X., Chen, W., Jiang, L., Dong, X., 2017. The combined cloud model for
Wan, Z., Chen, J., Craig, B., 2015. Lessons learned from Huizhou, China's unsuccessful edge detection. Multimed. Tool. Appl. 76, 15007e15026.
waste-to-energy incinerator project: assessment and policy recommendations. Zhao, X., Hwang, B.-G., Gao, Y., 2016. A fuzzy synthetic evaluation approach for risk
Util. Pol. 33, 63e68. assessment: a case of Singapore's green projects. J. Clean. Prod. 115, 203e213.
Wang, H.L., Feng, Y.Q., 2005. On multiple attribute group decision making with Zhu, W.D., Zhou, G.Z., Yang, S.L., 2009. An approach to group decision making based
linguistic assessment information based on cloud model. Control Decis. 20, on 2-dimension linguistic assessment information. Syst. Eng. 27, 113e118.