1 s2.0 S095183202100048X Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Reliability Engineering and System Safety 209 (2021) 107482

Contents lists available at ScienceDirect

Reliability Engineering and System Safety


journal homepage: www.elsevier.com/locate/ress

Supply reliability assessment of a gas pipeline network under


stochastic demands
Qian Chen a, Lili Zuo a, *, Changchun Wu a, *, Yankai Cao b, Yaran Bu a, Feng Chen c, Rehan Sadiq d
a
National Engineering Laboratory for Pipeline Safety/Beijing Key Laboratory of Urban Oil and Gas Distribution Technology, China University of Petroleum (Beijing),
Changping, Beijing, 102249, China
b
Department of Chemical and Biological Engineering, University of British Columbia, Vancouver, BC, V6T1Z3, Canada
c
National Academy of Innovation Strategy, China Association of Science and Technology. No.3 Fuxing Rd., Beijing, 100863, China
d
School of Engineering, University of British Columbia, Kelowna, BC, V1V1V7, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: An integrated methodology to assess the gas supply reliability of a gas pipeline network considering stochastic
Gas supply reliability demands is proposed in this study. Typical scenarios are selected based on the structural reliability calculated by
Structural reliability probability theory and stochastic process, including the normal scenario and some failure scenarios with a high
Gas pipeline network
probability. For each specific scenario, the gas supply condition is assessed based on the Latin hypercube sam­
Stochastic demand
pling with the Cholesky decomposition method under stochastic demands. The maximum flow method based on
Optimal gas supply scheme
the Dijkstra algorithm is adopted to determine whether the gas demand of customers can be fully covered and
optimize the supply scheme under shortages. Finally, the assessment results are demonstrated from the following
four aspects: the probability distribution of gas shortages under the normal scenario, identification of units with a
high failure probability and vulnerable units, the reasons of gas supply shortages and corresponding probabili­
ties, and the probability distribution of supply reliability for a gas pipeline network and each customer. The
methodology is applied to a large-scale gas pipeline network in China. The results of the supply reliability
assessment are analyzed in detail, and the sensitivity analysis of the gas demand uncertainty level on gas supply
reliability is conducted.

1. Introduction to satisfy customers’ demand. The pipeline system is considered to be


reliable as long as the transmission task is fulfilled. The supply reliability
Fulfilling the task of gas transmission and satisfying the demand for of a gas pipeline network might be equal to one under some failure
natural gas customers as much as possible are the primary duties for scenarios. Also, due to the restrictions of the supply quantity of gas
pipeline companies. In recent years, with the implementation of some sources and the transmission capacity of pipeline networks, the gas
environmental policies, the consumption of natural gas increased supply reliability might be less than one even if all units operate nor­
dramatically in China, and it caused gas supply shortages in some areas mally [3].
during some heating seasons. Therefore, it is necessary to assess the For the gas supply reliability analysis of a pipeline system, the first
supply reliability of the natural gas pipeline network and take some step is to analyze the structural reliability of each unit and the system.
measures to relieve the tension of gas supply during peak periods of gas Many approaches have been proposed for the structural reliability
consumption. evaluation of pipeline segments. The primary method is the probability
The reliability study of a gas pipeline system involves its structural statistics [4] based on historical operating data. Based on the failure
reliability and gas supply reliability. For the former topic, more effort frequency under different causes (external interference, corrosion,
has been devoted to the structure and integrity of gas pipelines. construction defect, hot tap made by error, ground movement) provided
Avoiding or reducing operational safety accidents is the goal of struc­ by the 8th EGIG report, the risk of the Italian gas pipeline network is
tural reliability researches [1,2]. However, the study of supply reli­ assessed in the reference [5], including the probability and frequency of
ability focuses more on evaluating the ability of the gas pipeline system top events in event tree analysis which include vapor cloud explosions,

* Corresponding author.
E-mail addresses: zuolili@cup.edu.cn (L. Zuo), wucc@cup.edu.cn (C. Wu).

https://doi.org/10.1016/j.ress.2021.107482
Received 6 May 2020; Received in revised form 25 September 2020; Accepted 15 January 2021
Available online 19 January 2021
0951-8320/© 2021 Elsevier Ltd. All rights reserved.
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Fig. 1. The framework of the gas supply reliability assessment of a gas pipeline network.

jet fire, flash fire. In reference [6], the failure rate of a pipeline segment and price effects on gas supply security, the TIGER model [35] is
is calculated after classifying operating data by failure causes (inter­ adopted to optimize the supply scheme of the European gas networks.
ference, corrosion, defect, movement, and other unknown reasons). Considering the behavior of gas traders and gas system operators, a
Also, the statistical model based on the nonlinear quantile regression is simulation approach [36] is developed to analyze the impact of the Nord
adopted for structural reliability analysis of pipeline units in reference Stream pipeline on the gas supply security of the European gas network
[7]. Besides the classical statistical analysis, the Bayesian networks in 2030. However, in the above researches about the gas supply security,
[8–11] are also employed to analyze the structural reliability of pipeline demand uncertainty and the stochastic failure of components in the
segments. In order to reduce the spatial dimensions, a Bayesian pipeline system are ignored, and these neglected factors might lead to
approach based on the systemic decomposition and reconstruction the overestimate of the gas supply condition.
approach [12] is developed to estimate the failure rates of pipeline The stochastic state transitions of units in the gas pipeline system are
segments. Besides, the limit state function is also often used in the considered subsequently in some literature about the supply reliability
structural reliability analysis. The limit state function returns a negative assessment of gas pipelines. Based on 1 million Monte-Carlo simulations
value under failure conditions and a positive value when the system is of unit operations, the ProGasNet model [37,38] is established to
reliable. The establishment of the limit state function of pipeline seg­ analyze the supply security of gas pipeline networks. The stochastic state
ments should consider the operating pressure, ultimate stress, yield transitions of gas pipeline networks are simulated by 1 million Monte
stress, pipeline diameter, pipeline wall thickness, and corrosion of Carlo simulations based on the Markov process in [39,40]. The sto­
pipeline segments [13–15]. Some limit state function models of pipe­ chastic state transition process considering the repair of units is simu­
lines include the DNV RP-F101 [16], ASME B31G [17], RAM PIPE lated by the Monte Carlo approach that includes samplings of states and
REQUAL [18], Kale et al. [19] model and Netto et al. [20] model. This durations for each transition in [41,42]. A sequential Monte Carlo
method usually needs large-scale samplings due to the nonlinearity of simulation method in [43] is used to analyze the reliability of a com­
the limit state function and multiple variables. And it is hard to obtain bined gas and electricity network, and this reference assumes that the
the limit state function of compressors. For structural reliability study of network is under the normal state at the beginning, the chronological
large systems, Monte Carlo simulations [21–24] are usually adopted to load curve and the failed unit need to be sampled sequentially for each
simulate random state behaviors of the system. Some variance reduction Monte Carlo sampling.
techniques [25,26] are also proposed aimed to reduce the number of Although Monte Carlo simulations can address the stochastic simu­
samples to obtain reliability estimations without affecting their expected lations of large systems, the enormous computational burden will make
values. In order to reduce the computational effort for the convergence the supply reliability assessment model inefficient when the uncertainty
of reliability indices, the state space decomposition based on the of the operational state and gas demand in pipeline networks are
population-based metaheuristic method [27,28] is proposed to prune considered at the same time. To avoid large scale samplings of the sto­
the state space during Monte Carlo simulations. chastic state transitions of a large system, the analytical method needs to
Gas supply security became a very concerning topic since the conflict be adopted to calculate the structural reliability of a large system. And it
between Russia and Ukraine due to the disagreement about the transit is not necessary to put too much emphasis on structural reliability
fee of natural gas [29–31]. In the MC-GENERCIS model [32,33], the analysis in supply reliability assessment. Instead, we should focus on
quantity of natural gas provided to each country via pipelines is those typical scenarios with a high probability.
considered as random variables, and the gas supply security for each In most previous studies, the reason for the unreliable gas supply is
country in the European Union is estimated by Monte Carlo simulations. only attributed to the failure of units, and gas supply is regarded as
To analyze the relations among demands, supplies, and investments, the totally reliable as long as all units in the system operate normally.
GASTALE model [34] is established to study the supply security of the However, even if all units in the system operate normally, gas supply
European gas networks. For the investigation of flow diversion effects shortages may also occur due to the restrictions of the supply quantity of

2
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

the gas sources and the transportation capacity of the gas pipeline Table 1
network. To reflect the gas supply condition comprehensively, the gas The state transition rate matrix of a pipeline segment.
supply reliability should also be assessed under the normal scenario (no [i, j] j =1 j =2
unit failure). And reasons for the unreliable gas supply and the corre­ i =1 − λP ⋅L λP ⋅L
sponding probability of each reason should also be analyzed in detail, i =2 μP − μP
which can help the gas pipeline company to make urgent decisions to
relieve some tight situations. Also, the demand uncertainty should be
considered for the consequence evaluation of gas supply under each ⎤ ⎡
P11 P12 ⋯ P1n
typical scenario. Under gas shortage situations, the optimal supply ⎢ P21 P22 ⋯ P2n ⎥
scheme should be formulated in order to satisfy the demand of the gas [P1 (t+dt),P2 (t+dt),⋯,Pn (t+dt)]=[P1 (t),P2 (t),⋯,Pn (t)]⎢
⎣ ⋮ ⋮ ⋱ ⋮ ⎦

pipeline network to the greatest extent. Pn1 Pn2 ⋯ Pnn
To overcome these deficiencies, an integrated methodology to assess
(1)
the long-term gas supply reliability of gas pipeline networks is devel­
oped in this study. Firstly, to avoid large-scale samplings of the sto­ Where Pi (t) and Pi (t+dt) are the probability that the unit is under
chastic state transition process of the large gas pipeline network, typical state i at time t and t+dt respectively; Pij denotes the probability of a unit
scenarios are selected based on the probability calculated by the Markov transferring to state j from state i during the time interval [t, t + dt].
stochastic process. Secondly, the proposed methodology considers the According to the reference [44], the definition of the transition rate
demand uncertainty in supply reliability assessment. Gas supply ade­ aij is the conditional probability density of a unit transitioning to state j
quacy under each typical scenario is assessed based on the Latin hy­ from state i during the time interval [t, t + dt], and the physical meaning
percube sampling (LHS) method. The sampling efficiency is compared of the transition rate is the number of transitions from state i to state j for
between the Monte Carlo method and the LHS method. Thirdly, a unit during the time interval [t, t + dt]. Based on this definition, the
considering the operational cost, the maximum flow method based on relation between the conditional probability Pij and the state transition
the Dijkstra algorithm is adopted to optimize the supply scheme under rate aij can be determined by Eqs. (2) and (3). The transition rate of a
gas supply shortages. Fourthly, the gas supply reliability is assessed unit (compressor or pipeline segment) can be obtained through statis­
based on the following four aspects: the probability distribution of gas tical analysis of historical operating data.
shortages under the normal scenario, analysis of high failure probability
units and vulnerable units, the reasons of unreliable gas supply and
Pij = dt⋅aij (i ∕
= j) (2)
corresponding probabilities, and the probability distribution of the

n
supply reliability for a gas pipeline network and each customer. Finally, Pii = 1 − dt⋅ aij (3)
a case study to illustrate the feasibility of this methodology is presented, j=1
the assessment results are analyzed in detail, and sensitivity analysis of j∕
=i
the uncertainty level of gas demand on supply reliability is conducted.
The proposed methodology is demonstrated in Section 2. In Section Where aij is the transition rate of a unit from state i to state j.
3, this methodology is applied to a large-scale gas pipeline network, and By substituting Eqs. (2-3) into Eq. (1), the first-order differential
the results are analyzed in detail. In Section 4, conclusions and future equation can be obtained, as shown in Eq. (4).
work are illustrated. This study can help to improve the management ⎡ ∑ n ⎤
level of gas pipeline systems. Also, it can lay a solid foundation in the − a1j a12 ⋯ a1n

⎢ j=2
transportation reliability research of other pipeline networks. ⎢



⎢ a21 − ∑
n

[ ] ⎢ a2j ⋯ a2n ⎥
dP1 dP2 dPn ⎥
2. Methodology , ,⋯,

=[P1 (t),P2 (t),⋯,Pn (t)]⎢ j=1 ⎥ (4)
dt dt dt ⎢ ⎥
⎢ j∕
=2 ⎥
⎢ ⎥
2.1. Proposed framework ⎢ ⋮ ⋮ ⋱ ⋮ ⎥

⎣ n∑
− 1 ⎦
an1 an2 ⋯− anj
The developed methodology is divided into four parts: structural j=1
reliability calculation, Latin hypercube sampling under stochastic de­
The matrix on the right side of the Eq. (4) is only related to the
mand, gas supply scheme optimization under shortages, and supply
transition rate of a unit, so this matrix is also named as the transition rate
reliability assessment. The framework of this methodology is shown in
matrix TR. For a pipeline segment, there are two states in the state space
Fig. 1. The structural reliability of each unit (pipe units and compressor
(state 1 for normal and state 2 for failure). The transition rate from state
station units) and the system is calculated based on the stochastic pro­
1 to state 2 is equal to the failure rate λP ⋅L, which is equal to the
cess in Section 2.1, and typical scenarios are selected based on the
reciprocal of the mean time between two successive failures of a pipeline
calculated probability. To evaluate the supply adequacy under sto­
segment. The transition rate from state 2 to state 1 is equal to the repair
chastic demands for each typical scenario, the Latin hypercube sampling
rate μP , which is equal to the reciprocal of the mean time to repair. The
with Cholesky decomposition method (Section 3.2) is adopted to sample
failure rate and repair rate can be obtained through the statistical
demand cases based on historical demand data of each customer. In
analysis of the historical data about the time interval between two
Section 3.3, the maximum flow method based on the Dijkstra algorithm
successive failures and the average repair time of a pipeline segment.
is adopted to verify whether the gas demand of the system can be fully
The transition rate matrix TRpipe is shown in Table 1, where λp is the
covered and optimize the supply scheme under gas shortages. Finally,
failure rate of a pipeline segment, times/(km⋅a), L is the length of a
the supply reliability assessment of a gas pipeline network is conducted
pipeline segment, km; μp is the repair rate of a pipeline segment, times/a.
from four aspects (Section 3.4).
Based on the probability distribution at the initial time [PP1 (t0 ),
2.2. Structural reliability calculation of gas pipeline networks PP2 (t0 )], the probability that a pipeline segment is under the normal state
or the failure state can be calculated. The results are shown as Eqs. (5)
Assuming that there is one state transition at most for a unit during a and (6).
time step dt, then Eq. (1) can be obtained based on probability theory. μP [ ]
PP1 (t) = 1 − e− (λP ⋅L+μP )(t− t0 )
+ PP1 (t0 )e− (λP ⋅L+μP )(t− t0 )
(5)
λP ⋅L + μP

3
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Table 2
The possible operating states for a station with two operating compressors and
one standby compressor.
Operating Description
states

State 1 Three compressors are under the normal operating state.


State 2 Two compressors are under the normal operating state, and one
compressor is under the failure state.
State 3 One compressor is under the normal operating state, and two
compressors are under the failure state.
State 4 Three compressors are under the failure state.

Table 3
The state transition rate matrix for a compressor station with two operating
compressors and one standby compressor.
[i, j] j =1 j =2 j =3 j =4

i =1 − 3λc 3λc 0 0
i =2 μc − 2λc − μc 2λc 0
i =3 0 2μc − λc − 2μc λc
i =4 0 0 3μc − 3μc Fig. 2. Diagram of the Latin Hypercube sampling.

under this degeneration state is equal to that when all compressors in


PP2 (t) =
λP ⋅L [
1 − e− (λP ⋅L+μP )(t− t0 )
]
+ PP2 (t0 )e − (λP ⋅L+μP )(t− t0 )
(6) this station are under the failure state. In this paper, the worst case is
λP ⋅L + μP considered for failure scenarios of compressor stations, that is, the fail­
Where PP1 denotes the normal probability of a pipeline segment at ure of two compressor units in a station will cause the failure of this
time t and PP2 denotes the failure probability of a pipeline segment at compressor station.
time t. It is not appropriate to put too much emphasis on the structural
Assuming that all compressor units in a compressor station are reliability in supply reliability assessment of a gas pipeline network,
identical, then the number of possible operating states for a compressor because it may cause the assessment model unpractical and inefficient.
station is x + y + 1 when the number of operating compressors and Instead, we need to focus on typical scenarios, including the normal
standby compressors is x and y, respectively. Similar to the transition scenario and failure scenarios with a high probability. It is assumed that
rate matrix of a pipeline segment, the transition rate matrix of a the operating state of each unit is independent of each other, then the
compressor station can also be expressed by the failure rate and repair probability that the system is under the normal scenario or other failure
rate of compressors. For example, there are four possible operating scenarios can be calculated based on the probability of the normal state
states for a compressor station with two operating compressors and one or the failure state of each unit. It is found that the probability of the
standby compressor (Table 2). scenario that more than two failures during the assessment period (one
If any one of the three compressors fail in this compressor station, the month) is lower than 10− 6, so these failure scenarios are neglected in
state of this station will be transferred from state 1 to state 2, so the supply reliability assessment.
transition rate of this compressor station from state 1 to state 2 is equal
to 3λc . The transition rate matrix TRcom of this compressor station is 2.3. Latin hypercube sampling with Cholesky decomposition under
shown in Table 3. Where λc is the failure rate of a compressor unit, times stochastic demand
/a, μc is the repair rate of a compressor unit, times/a.
Based on the initial probability distribution PC (t0 ), the probability At the beginning of each month, gas customers need to report their
under each state of this compressor station PC (t) can be obtained as Eq monthly nominations to the gas pipeline company. Since it is difficult to
(7). precisely predict the gas demand and demand uncertainty is the main
PC (t) = A⋅B− 1 ⋅PC (t0 ) (7) factor that influences the supply reliability, the gas demand of customers
should be regarded as stochastic variables in the reliability assessment.
Where A = [w1 , e− (λC +μC )⋅t ⋅w2 , e− 2(λC +μC )⋅t ⋅w3 , e− 3(λC +μC )⋅t ⋅w4 ]; Assuming that the gas demand of each customer follows the normal
B = [w1 , e− (λC +μC )⋅t0 ⋅w2 , e− 2(λC +μC )⋅t0 ⋅w3 , e− 3(λC +μC )⋅t0 ⋅w4 ]; distribution, and the supply reliability is assessed based on one month in
⎡ ⎤ ⎡ ⎤ ⎡ / ⎤ this paper.
1 − λC /μC λC 2 μC 2/
⎢ ⎥ The Monte Carlo method is commonly used to address uncertainty
⎢1⎥ ⎢ (μ − 2λC )/3μC ⎥ λ (λ − 2μC ) 3μC 2
w1 = ⎣ ⎦; w2 = ⎣ C
1
;w = ⎢ C C
(2μC − λC )/3μC ⎦ 3 ⎣ (μC − 2λC )/3μC
⎥; w4
⎦ issues. However, it has the defects of poor repeatability and low effi­
1 1 1 ciency. To reduce sampling times and guarantee that all the sampling
⎡ / ⎤ intervals can be covered, the Latin hypercube sampling (LHS) method is
− λC 3/ μC 3
⎢ λC 2 μ 2 ⎥ adopted in this paper. To reflect the demand correlations among cus­
=⎢⎣ − λC /μ ⎦
C ⎥ tomers, the Cholesky decomposition (CD) is integrated into the LHS
method.
C
1
The first step of this method is the Latin hypercube sampling.
Under this configuration (two working compressors and one standby Supposing that the sampling size is N, then divide the [0,1] interval of
compressor), the inlet flow rate will be doubled if two compressors are the cumulative probability distribution function into N nonoverlapping
under the failure state at the same time. In this situation, the only intervals with equivalent lengths for a random variable. Randomly
working compressor is likely to shut down automatically due to the sample an interval without replacement by N times, and take the median
stagnation state, and the transmission capacity of a gas pipeline network of each interval as the sample value. If the number of customers with
random demand is K, then the LHS sampling matrix S (N × K) can be

4
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

The LHS sampling matrix S needs to be reordered, sequenced, and


decomposed to reflect gas demand correlations between customers. The
correlation between any two customers can be calculated by Eq. (8).
[ ]
(X − E(X))(Y − E(Y))
ρXY = E √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (8)
D(X)⋅D(Y)

Where X and Y denote historical demand data of the customer #1


and #2, respectively; E(X) and E(Y) denote the expected demand of the
customer #1 and #2, respectively; D(X) and D(Y) denote the variance of
gas demand of the customer #1 and #2, respectively.
The detailed procedures to get the LHS-CD sampling matrix are as
follows:

Fig. 3. Diagram of nodes and edges in the maximum flow method. (a) For each column in the matrix S, rearrange this column by the
ascending order to get the ordering matrix S1 .
obtained by repeating the above procedures K times. As can be seen in (b) For each element in the matrix S1 (i, j), search the same number
Fig. 2, fewer samples are taken from the tail and end due to the lower from the matrix S in column j, and assign the corresponding row
probability, and more samples are taken from the middle with higher number in column j of matrix S to L(i,j), then the sequence matrix
probabilities. L can be obtained.

Fig. 4. Gas supply adequacy assessment under no failure scenario and each failure scenario.

Fig. 5. The detailed procedures of supply reliability assessment.

5
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

(c) Suppose that the correlation matrix of the sequence matrix L is contains information about the gas transmission capacity of each edge
C1 , then decompose the correlation matrix C1 and the input de­ (pipeline segment). If the pipeline segment exists and does not fail, the
mand correlation matrix C respectively by Cholesky decomposi­ gas transmission capacity can be calculated by Eq. (9). Otherwise, the
tion as C1 = Q1 ⋅Q1 and C = Q2 ⋅Q2 , where Q1 and Q2 are lower
′ ′
gas transmission capacity is equal to zero. For each element CM(i, j) in
triangular matrixes. the matrix CM, it is equal to the length of the edge (i, j) multiplies the
(d) Get the decomposed matrix G by G = Q2 ⋅Q1 − 1 ⋅L. corresponding cost (dollars/(km⋅m3 )) if the pipeline segment exists and
(e) According to step (b), get the sequence matrix T of the decom­ does not fail. Otherwise, it can be seen as a big number M. The output
posed matrix G. Then according to step (a), rearrange the LHS parameters are the maximum gas transmission capacity of a pipeline
sampling matrix S based on the sequence matrix T to get the LHS- network and the actual gas delivery flow rate to each customer. The
CD sampling matrix SC. mathematical optimization model of this problem is as follows:

max f (i, j) (10)
2.4. Supply scheme optimization (i,j)∈E
i=S

Under the normal scenario and some typical failure scenarios, it is


Subject to:
necessary to check whether customers’ demand can be fully covered. If
customers’ demand cannot be fully covered, the optimal gas supply (1)0 ≤ f (i, j) ≤ c(i, j) (11)
scheme should be formulated. In this section, the maximum flow method ∑ ∑
is adopted to check whether the gas demand in the system can be fully (2) f (i, j) − f (j, i) = 0 (12)
covered and formulate the optimal supply scheme under shortages. (i,j)∈E
i∕
=S,j∕=T
(j,i)∈E
i∕
=S,j∕=T
There are many gas sources and customers in a real gas pipeline
network, the virtual super source node and the virtual super sink node ∑ ∑
(3) f (i, j) − f (i, j) = 0 (13)
are added as shown in Fig. 3, then this problem becomes a regular
(i,j)∈E
maximum flow problem. The transmission capacity of the edge from the
(i,j)∈E
i=S j=T

super source node to each general source node is equal to the supply
capacity of each gas source, the transmission capacity of the edge from Where f(i, j) means the actual flow rate of the edge (i,j); c(i, j) means
each sink node to the super sink node is equal to the gas demand of each the transmission capacity of the edge (i, j); S means the super source
customer. node; T means the super sink node, E implies the set of all edges in a gas
Under the normal scenario, the transmission capacity of each edge pipeline network.
(pipeline segment) can be calculated by the maximum allowed pressure The basic procedures of the maximum flow method are as follows:
at the inlet, and the minimum allowed pressure at the outlet, the length, (1) Set the initial value of the maximum flow MF as zero and the
and the diameter of the pipeline segment (Eq. (9)). This equation is optimal flow rate matrix OC as zero matrix. Input the transmission ca­
suitable for calculating the gas transmission capacity of long-distance pacity matrix TCM and the cost matrix CM.
horizontal gas pipeline segments, and the detailed derivations can be (2) For the current matrix TCM and CM, based on the Dijkstra al­
seen in [45]. Under the scenario of pipeline segment failures, the gorithm (see Appendix A), find a path P from the super source node to
transmission capacity of the failed pipeline segment is equal to zero. the super sink node with the minimum operating cost. If there is no
Under the scenario of compressor station failures, the transmission ca­ feasible path between the super source node and the super sink node, go
pacity of the two pipeline segments adjacent to the failed compressor to step (4); otherwise, go to step (3).
station will decrease, which can also be calculated by Eq. (9). Under (3) For all edges (i,j) ∈ P, update OC(i,j) = OC(i,j) + min(TCM(i,j)),
these situations, the two pipeline segments adjacent to the failed and MF = MF + min(TCM(i, j)).
compressor station can be regarded as a pipeline section, the maximum For all edges that satisfy (i, j) ∈ P and TCM(i,j) = min(TCM(i,j)), the
allowed inlet pressure of the pipeline section depends on the maximum cost of these edges (i, j) in the matrix CM is updated to a big number M,
allowed pressure of the upstream pipeline segment, and the minimum and the transmission capacity of the edge (i, j) and (j, i) in the matrix
allowed outlet pressure of the pipeline section depends on the minimum TCM is updated to zero and 2min(TCM(j, i)), respectively.
allowed pressure of the downstream pipeline segment. For edges that satisfy (i, j) ∈ P but do not satisfy TCM(i, j) =
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ min(TCM(i,j)), the transmission capacity of the edge (i, j) and (j, i) in the
( ) ̅ matrix TCM is updated to TCM(i, j) − min(TCM(i, j)) and TCM(j, i) +
Pmax in 2 − Pmin out 2 ⋅d5
Qmax = C1 ⋅ (9) min(TCM(i, j)), respectively.
λ⋅Z⋅T⋅L⋅Δ∗
Then go back to step (2).
Where Qmax is the maximum transmission capacity of the edge, m3 (4) Stop iterations. The optimal gas supply quantity to each customer
/s; C1 is a constant number, 0.03848(m2 ⋅s⋅K0.5 ⋅kg− 1 ); Pmax in is the can be obtained based on the optimal flow rate matrix OC.
maximum allowed pressure at the inlet of the pipeline segment, which is
usually equal to the designed pressure of the pipeline system or the 2.5. Supply reliability assessment
supply pressure of the gas source, Pa; Pmin_out is the minimum allowed
pressure at the outlet of the pipeline segment, which is usually equal to The main concerns about the supply reliability of a gas pipeline
the minimum allowed suction pressure of the compressor stations or the network include but are not limited to the following four aspects:
contract pressure of the gas customers, Pa; d is the inner diameter of a
pipeline segment, m; λ is the friction coefficient; Z is the compressibility (1) The normal scenario that no unit fails in the system during the
factor of the natural gas; ∆* is the relative density of the natural gas; T is assessment period is most likely to happen. In this particular
the average temperature of the natural gas, K; L is the length of a scenario, which customers are likely to experience gas supply
pipeline segment, m. shortages, and to what extent? And what is the supply adequacy
As shown in Fig. 3, the maximum flow problem of a gas pipeline of the gas pipeline network under this scenario?
network is to calculate the maximum flow rate from the super source (2) Which unit failure scenarios are most likely to happen, and which
node to the super sink node based on the transmission capacity and the failure scenario has the most significant impact on the gas supply
length of each pipeline segment. The input parameters of the maximum adequacy?
flow method are the transmission capacity matrix TCM and the cost (3) What are the reasons for unreliable gas supply, and what is the
matrix CM of a gas pipeline network. The transmission capacity matrix probability for each reason?

6
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Fig. 6. The topology of the tested gas pipeline network.

(4) From the perspective of a gas pipeline network system and each
Table 4
customer, what is the probability distribution of the gas supply
The supply capacity of gas sources.
reliability?
Gas sources node Capacity (104m3/ Gas sources node Capacity (104m3/
d) d)
Procedures about gas supply adequacy assessment under each typical
scenario are shown in Fig. 4. The gas supply adequacy is equal to the S1 4000 S6 2500
actual supply volume divided by the demand volume during the S2 7000 S7 700
S3 800 S8 800
assessment period. For a specific scenario, the probability distribution of S4 800 S9 700
the supply adequacy can be obtained by Eq. (14). Based on the proba­ S5 1200
bility of each typical scenario and corresponding supply adequacy, the
supply reliability for each customer and a gas network system can be
obtained by Eq. (15). Detailed procedures to assess the gas supply reli­
Table 5
ability of a gas network system are shown in Fig. 5. The average demand for each customer.
⃒ ⃒ NC(S ≥ α⋅D) ⃒⃒ Customers Demand Customers Demand Customers Demand
A(α)⃒sck = Prob(S ≥ α⋅D)⃒sck = sck (14) (104m3/d) (104m3/d) (104m3/d)
N
D1–1 430.0 D4–5 946.3 D8–2 329.4
ns (
∑ )
D2–1 962.5 D4–6 394.5 D8–3 91.9
R(α) = Prob(sck )⋅A(α)|sck (15)
D2–2 701.8 D4–7 439.8 D9–1 1039.0
k=1
D2–3 855.7 D4–8 751.5 D9–2 924.4
Where S and D denote the gas supply quantity and the gas demand D3–1 654.8 D4–9 1409.2 D9–3 78.3
quantity respectively; α denotes the supply adequacy level or the supply D3–2 765.2 D4–10 469.6 D9–4 148.3
D4–1 388.8 D5–1 2870.0 D10–1 300.0
reliability level; A(α)|sck denotes the probability that the supply ade­ D4–2 435.9 D6–1 200.0 D11–1 23.6
quacy is higher than α under the scenario sck , R(α) indicates the prob­ D4–3 613.3 D7–1 460.0 D11–2 236.4
ability that the supply reliability is higher than α, Prob(sck ) means the D4–4 291.1 D8–1 128.7
probability of the scenario sck ; N denotes the sampling times by LHS-CD, Notes: The acronym “Di-j” for each customer means the customer j of the
NC means the number of times a certain condition is satisfied, ns is the province i.
number of typical scenarios.
gas pipeline network is 8649 km. There are 9 gas sources, 29 customers
3. Case study from 11 different provinces, and 21 compressor stations in the system.
The supply capacity of each gas source is shown in Table 4. The average
3.1. Description of the case-study gas pipeline network demand of each customer during the assessment period is shown in
Table 5. The configuration of compressor stations is shown in Table 6.
The developed methodology is applied to a large-scale gas pipeline Based on the statistical analysis of the historical operating data, the
network system in China. The assessment period is one month. The to­ failure rate of pipeline segments and compressor units are taken as
pology of the gas pipeline network is shown in Fig. 6. The length of the 1.279 × 10− 4 times/(km⋅year) and 2.92 times/year, respectively, and

7
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Table 6
The configuration of each compressor station.
Compressor Station Configurations Compressor Station Configurations Compressor Station Configurations

C1 1+ 1 C8 1+1 C15 3 +1
C2 2+ 1 C9 3+1 C16 2 +1
C3 1+ 1 C10 3+1 C17 3 +1
C4 1+ 1 C11 3+1 C18 2 +1
C5 1+ 1 C12 3+1 C19 2 +1
C6 1+ 1 C13 3+1 C20 2 +1
C7 1+ 1 C14 2+1 C21 2 +1

Notes: The “x + y” in configurations means the number of operating compressors and standby compressors is x and y, respectively.

3.2. Supply adequacy under the normal scenario


Table 7
The mean and standard deviation of gas demand for each province.
The scenario that no unit fails in the pipeline network during the
Province Mean Standard Province Mean Standard assessment period is very likely to happen, so additional attention
(104m3/d) deviation (104m3/d) deviation
should be paid to this scenario. In this condition, the reason for inade­
1 430 25 7 460 35 quate gas supply is the restriction of the supply capacity of gas sources or
2 2520 200 8 550 60 the transmission capacity of the gas pipeline network.
3 1420 65 9 2190 120
4 6140 300 10 300 30
In this paper, the supply reliability assessments are conducted in
5 2870 150 11 260 30 terms of provinces. Based on the historical data, the mean and standard
6 200 20 deviation of gas demand for each province and the correlation co­
efficients between provinces are calculated, as shown in Tables 7 and 8.
In Fig. 7(a), the relative errors of the system supply adequacy under
the average maintenance time after failures for pipeline segments and
different sampling sizes are compared between the LHS-CD method and
compressor units are both taken as 72 h. In this paper, the gas supply
the Monte Carlo method (MC–CD). The supply adequacy based on 5000
reliability is assessed based on the forecast of the average monthly gas
sampling times of the LHS-CD method is regarded as the reference value.
demand during a heating season. The gas demand is relatively large in
The relative error of assessment results can be converged within 0.001
this situation.
by 1000 sampling times based on the LHS-CD method, whereas 3000
sampling times are needed for the MC–CD method. The sampling

Table 8
The correlation coefficients between any two provinces.
Province 1 2 3 4 5 6 7 8 9 10 11

1 1.000 0.311 0.514 0.435 − 0.206 − 0.159 0.548 0.532 0.515 0.076 0.380
2 0.311 1.000 0.606 0.454 0.209 0.551 0.546 0.568 0.459 0.684 0.528
3 0.514 0.606 1.000 0.413 0.276 0.375 0.803 0.580 0.241 0.433 0.376
4 0.435 0.454 0.413 1.000 0.113 0.063 0.450 0.465 0.520 0.443 0.470
5 − 0.206 0.209 0.276 0.113 1.000 0.449 0.053 0.083 − 0.362 0.223 − 0.002
6 − 0.159 0.551 0.375 0.063 0.449 1.000 0.228 − 0.039 − 0.271 0.592 − 0.041
7 0.548 0.546 0.803 0.450 0.053 0.228 1.000 0.731 0.351 0.391 0.345
8 0.532 0.568 0.580 0.465 0.083 − 0.039 0.731 1.000 0.522 0.337 0.520
9 0.515 0.459 0.241 0.520 − 0.362 − 0.271 0.351 0.522 1.000 0.102 0.566
10 0.076 0.684 0.433 0.443 0.223 0.592 0.391 0.337 0.102 1.000 0.249
11 0.380 0.528 0.376 0.470 − 0.002 − 0.041 0.345 0.520 0.566 0.249 1.000

Fig. 7. Relative error and repeatability test of the assessment results.

8
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Table 9
The probability of different gas supply adequacy levels for each province and the pipeline system under the normal scenario.
Probability System Pro4 Pro5 Pro9 Pro10 Pro1/Pro2/Pro3/Pro6/Pro7/Pro8/Pro11

P (adequacy==1) 0.9430 0.9580 0.9980 0.9900 0.9550 1.0000


P (adequacy>=0.98) 0.9840 0.9770 0.9980 0.9920 0.9550 1.0000
P (adequacy>=0.96) 0.9960 0.9940 0.9990 0.9950 0.9550 1.0000
P (adequacy>=0.94) 0.9990 0.9970 1.0000 0.9960 0.9560 1.0000
P (adequacy>=0.92) 1.0000 0.9980 1.0000 0.9960 0.9570 1.0000
P (adequacy>=0.9) 1.0000 1.0000 1.0000 0.9960 0.9580 1.0000

Table 10
Top 10 failure scenarios with the most severe consequence.
Rank Failure unit Failure Average supply adequacy
probability during failures

1 The pipeline segment 0.00012 0.64618


from S2 to C9
2 The pipeline segment 0.00013 0.64618
from C9 to C10
3 The pipeline segment 0.00015 0.64618
from C10 to C11
4 The pipeline segment 0.00017 0.64618
from C11 to C12
5 The pipeline segment 0.00014 0.64618
from C12 to C13
6 The pipeline segment 0.00015 0.65772
from C13 to C14
7 The pipeline segment 0.00017 0.65772
from C14 to C15
8 The pipeline segment 0.00013 0.71311
from C15 to D17
9 The pipeline segment 0.00031 0.76212
from C7 to C8
10 The pipeline segment 0.00014 0.76247
Fig. 8. Total supply and demand under 57 insufficient gas supply scenarios. from D17 to C16

efficiency is greatly improved by the LHS-CD method. The assessments 3.3. Supply reliability assessment of the gas pipeline network
are also conducted by five repetitive trials based on the LHS-CD method
under 1000 times sampling, and the results are shown in Fig. 7(b). It can Based on the proposed method in Section 3.1, the probability that the
be seen that the assessment results of the LHS-CD method have good gas network system is under the normal operating state during one
repeatability. Therefore, 1000 samplings of customers’ demands based month can be calculated, which is equal to 0.9564. In the supply reli­
on the LHS-CD method are conducted in this paper when assessing the ability assessment, 74 scenarios are selected as typical scenarios,
gas supply adequacy under each typical scenario. including the normal scenario and 73 failure scenarios, the total prob­
The adequacy assessment results of the system and each province are ability of typical scenarios accounts for 0.9991. For each scenario, gas
shown in Table 9. Under this normal scenario, the probability that the supply adequacy is assessed based on 1000 different demand scenarios,
demand of all customers can be fully covered is 0.9430. There will be no so 74,000 scenarios need to be simulated based on the maximum flow
gas supply shortages for seven provinces except province #4, #5, #9, method. The time taken for all simulations is approximately 3.2 h
and #10. The reason for the insufficient gas supply in province #4 is its (Machine specifications: 4.0 GHz, 64 GB RAM).
large demand. The lack of sufficient gas supply in province #5 is due to The pipeline segment with the highest failure probability (0.00051)
the large demand from the upstream customers and the limited gas is the pipeline unit from the customer D8–3 to the customer D9–1. The
transmission capacity of gas pipelines connecting province #5. While compressor stations with the highest failure probability (0.00307) are
the lack of sufficient gas supply in provinces #9 and #10 is due to the the compressor stations configured with three working compressors and
weak interconnections of gas pipelines around these two provinces and one standby compressor. The top 10 failure scenarios with the most
the limited gas supply capacities of the surrounding gas sources. severe consequence are listed in Table 10. It can be deduced that pipe­
As can be seen in Fig. 8, through 1000 times sampling based on LHS- line segments from the gas source S2 to compressor station C15 are the
CD, the number of inadequate gas supply scenarios is 57, and 45 cases of most vulnerable units in the system. The reason is that the diameter of
which are caused by the restriction of the supply capacity of the gas the pipeline segments from the compressor station C9 to compressor
sources (18,500×104m3/d), and other 12 cases are caused by the re­ station C15 (1219 mm) is larger than that from the compressor station
striction of the transmission capacity of the gas pipeline network. The C1 to compressor station C6 (1016 mm), the former regional gas pipe­
probability of inadequate gas supply due to the restriction of supply line needs to undertake more transportation tasks than the latter.
capacities of the gas sources and the transmission capacity of the gas Additional attention should be paid to the pipe segments from the
networks is 0.0450 and 0.0120, respectively. It can be seen that the compressor station C9 to compressor station C15 in the routine man­
transmission capacity of the gas network under the above 12 gas agement of gas pipelines. Also, from the analysis of vulnerable units, it
shortage scenarios is less than 18,500×104m3/d. The reason is that can be seen that the consequences of compressor station failures are less
when the gas demands of some customers decrease, the gas transmission severe than pipeline segment failures normally.
flow rate of the pipe segments around these customers might decrease, The supply reliability assessment results are shown in Fig. 9. It can be
and the transmission flow rate of other pipe segments could not be seen that the probability that the gas pipeline network can fully cover
further increased. Thus the transmission capacity of the gas pipeline the demand of all the customers is 0.9258, which is slightly lower than
network might be different under different demand conditions.

9
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Fig. 9. The supply reliability assessment results of the gas pipeline network.

Table 11
Supply reliability distribution of different provinces.
Probability Pro1 Pro2 Pro3 Pro4 Pro5 Pro6 Pro7 Pro8 Pro9 Pro10 Pro11

P (reliability==1) 0.9989 0.9985 0.9985 0.9507 0.9926 0.9984 0.9991 0.9972 0.9790 0.9407 0.9969
P (reliability >=0.99) 0.9989 0.9987 0.9985 0.9646 0.9938 0.9984 0.9991 0.9972 0.9827 0.9412 0.9969
P (reliability >=0.98) 0.9989 0.9987 0.9985 0.9744 0.9947 0.9984 0.9991 0.9972 0.9852 0.9414 0.9969
P (reliability >=0.97) 0.9989 0.9987 0.9985 0.9871 0.9959 0.9984 0.9991 0.9972 0.9887 0.9418 0.9969
P (reliability >=0.96) 0.9989 0.9991 0.9985 0.9926 0.9961 0.9984 0.9991 0.9973 0.9939 0.9422 0.9969
P (reliability >=0.95) 0.9989 0.9991 0.9986 0.9940 0.9973 0.9984 0.9991 0.9973 0.9950 0.9425 0.9969
P (reliability >=0.94) 0.9989 0.9991 0.9988 0.9960 0.9977 0.9984 0.9991 0.9975 0.9950 0.9439 0.9969
P (reliability >=0.93) 0.9989 0.9991 0.9988 0.9971 0.9978 0.9984 0.9991 0.9979 0.9951 0.9443 0.9969
P (reliability >=0.92) 0.9989 0.9991 0.9988 0.9971 0.9980 0.9984 0.9991 0.9984 0.9951 0.9457 0.9969
P (reliability >=0.91) 0.9989 0.9991 0.9988 0.9991 0.9980 0.9984 0.9991 0.9985 0.9951 0.9461 0.9969
P (reliability >=0.90) 0.9991 0.9991 0.9991 0.9991 0.9991 0.9991 0.9991 0.9991 0.9951 0.9570 0.9991

the probability that the pipeline network is under the normal state
throughout the assessment period (because the probability of sufficient Table 12
Four standard deviation levels of gas demand for each province.
gas supply of the gas pipeline network is not equal to one under the
normal scenario). Although the probability that the gas supply reliability Province Standard Standard Standard Standard
deviation deviation deviation with deviation with
equals to one is not that high, the probability that 95 percent of total gas
under the without low demand high demand
demand can be covered is larger than 0.9970. The major reason that benchmark demand uncertainty uncertainty
causes the inadequate gas supply of the pipeline network is the restric­ uncertainty
tion of the total supply capacity of the gas sources. If the supply reli­ 1 25 0 20 30
ability of the gas pipeline network wants to be improved, the production 2 200 0 160 240
quantity of gas sources should be increased, or new sources such as LNG 3 65 0 52 78
sources and gas fields should be developed. Inadequate gas supply due to 4 300 0 240 360
5 150 0 120 180
pipeline failures account for only 0.0060 because the probabilities of
6 20 0 16 24
pipe failures usually are very low. Also, the pipeline network is highly 7 35 0 28 42
interconnected, and the remained gas pipeline may also perform most of 8 60 0 48 72
the transportation tasks when a single pipeline segment fails. 9 120 0 96 144
The supply reliability assessment results of different provinces are 10 30 0 24 36
11 30 0 24 36
shown in Table 11. The supply reliability of province #4, #9, and #10 is
relatively low. The reason for the low supply reliability for province #4
is the significant gas demand. This tight gas supply situation can be It can be seen from Fig. 10 that the gas supply reliability is very high
relieved by introducing some gas sources downstream of the compressor without considering demand uncertainty, the probability that the total
station C8. The reason for the low supply reliability of province #9 and demands of the pipeline system can be fully satisfied is 0.9769. Under
#10 is that the interconnections of pipelines are not that strong. It will this special case, there will be no gas supply shortages in the gas pipeline
cause severe damages to province # 9 and # 10 once the connecting system under the normal scenario, and gas shortages can only be caused
pipeline segments around these two provinces fail. This situation can be by the failures of pipe units and compressor stations. Therefore, the
improved by constructing more gas pipelines around these two estimated supply reliability level will be higher if the uncertainty of gas
provinces. demand is not considered in assessing the supply reliability of gas
pipeline systems.
As shown in Fig. 10, compared with the assessment results under the
3.4. Sensitivity analysis of gas demand uncertainty benchmark case, the probability that the gas pipeline network can fully
meet the demand of customers decreases by 0.0321 (increases by
To study the influence of gas demand uncertainty level on the gas 0.0358) when the variance of gas demand increases by 20% (decreases
supply reliability, the supply reliability is assessed based on the by 20%). It shows that the gas supply reliability will be lower when the
following four different cases shown in Table 12.

10
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Fig. 10. Sensitivity analysis of gas demand uncertainty on gas supply reliability.

gas demand uncertainty level gets higher. The reason is that the sample total gas demand can be covered in the system is larger than
intervals of the gas demand through the LHS method will be wider when 0.9970. The major reason that causes the inadequate gas supply
the uncertainty level is higher. The probability of being sampled to a of the pipeline network is the restriction of the total supply ca­
large gas demand case will increase, which leads to a decrease in the pacity of gas sources (the probability is 0.0435). The probability
supply reliability of the gas pipeline network. From the above analysis, of gas supply shortages in the system due to compressor failures
we can infer that the supply reliability might be relatively lower in some and pipe failures is 0.0123 and 0.0060, respectively.
extremely cold days due to the higher uncertainty of gas demands. Some (5) The probability that the total demands of the pipeline system can
measures can be taken to relive the tension situation of gas supply, such be fully satisfied is 0.9769 if the uncertainty of gas demand is not
as increasing the peak shaving capacity of the gas storages or finding considered. Compared with the assessment results under the
other alternatives. benchmark case, when the variance of gas demand increases by
20% (decreases by 20%), the probability that the gas pipeline
4. Conclusions and future work network can fully meet the demand of customers decreases by
0.0321 (increases by 0.0358).
An integrated methodology to assess the gas supply reliability of the
gas pipeline network considering the demand uncertainty and the un­ The developed methodology also needs to be improved in the future.
certainty of the operating state of the system is developed in this study. The interest and politic relationships between gas importing countries
In the case study, a large-scale gas pipeline network is applied to confirm and gas exporting countries, and the supply capacity of gas storages
the feasibility of the proposed methodology. The main conclusions during heating seasons can sometimes cause uncertainty on supply sides,
drawn from the case study are: and this uncertainty needs to be considered in the gas supply reliability
assessment. Besides, the detailed operations of gas pipeline networks
(1) The probability that the gas pipeline network is under the normal will also be considered in the evaluation of gas supply consequences.
operating scenario during the assessment period is equal to
0.9564. The probability of the scenario that more than two units Declaration of Competing Interest
fail during the assessment period (one month) is lower than 10− 6.
(2) The relative error of assessment results can be converged within The authors declare that they have no known competing financial
0.001 by 1000 sampling times based on the LHS-CD method, interests or personal relationships that could have appeared to influence
whereas 3000 sampling times are needed for the MC–CD the work reported in this paper.
method. Besides, the assessment results of the LHS-CD method
have good repeatability. Acknowledgements
(3) Under the normal scenario, the probability that the demand of
customers can be fully covered is 0.943, and the probability of The authors acknowledge the financial support from the China
inadequate gas supply due to the restriction of the total supply Scholarship Council for a one-year study at the University of British
capacity of gas sources and the transmission capacity of the gas Columbia [201906440048]. The authors are also thankful to all the
network is 0.045 and 0.012, respectively. reviewers for their insights and constructive comments.
(4) The probability that the gas demand of the pipeline system can be
fully satisfied is 0.9258, and the probability that 95 percent of

11
Q. Chen et al. Reliability Engineering and System Safety 209 (2021) 107482

Appendix A [16] Veritas Norske. Corroded pipelines. Det NorskeVeritas; 1999.


[17] Anon ASME. B31G–Manual for determining the remaining strength of corroded
pipelines–a supplement to ansi/asme B31 code for pressure piping. American
Table A1 Society of Mechanical Engineers; 1991.
[18] Bea R, Xu T. Corrosion effects on burst pressures ram pipe requal. University of
California at Berkeley; 1999. p. 103e104.
[19] Kale Amit, Thacker Ben H, Sridhar Narasi, Waldhart Chris J. A probabilistic model
for internal corrosion of gas pipelines. Int Pipeline Conf Vol 2004;41766.
Table A.1
[20] Netto TA, Ferraz US, Estefen SF. The effect of corrosion defects on the burst
Basic structure of the Dijkstra algorithm. pressure of pipelines. J Constr Steel Res 2005;61:1185–204.
1. Input the cost matrix CM. Create a set of unvisited nodes UN and a set of visited [21] Zio E. The monte carlo simulation method for system reliability and risk analysis.
nodes VN, and mark all nodes unvisited firstly. DE: Springer Verlag London Limited; 2013.
[22] Celli G, Ghiani E, Pilo F, Soma GG. Reliability assessment in smart distribution
2. Assign a cost value for each node (zero for the super source node and infinity for all
networks. Electr Pow Syst Res 2013;104:164–75.
other nodes).
[23] Ghaderi A, Haghifam MR, Abedi SM. Application of monte carlo simulation in
3. While UN ∕ =∅
markov process for reliability analysis. IEEE; 2010. p. 293–8.
# Based on CM, select the node u from the set UN with the minimum cost. [24] Wang H, Yajima A, Castaneda H. A stochastic defect growth model for reliability
u = minimum_cost (UN, CM) assessment of corroded underground pipelines. Process Saf Environ 2019;123:
179–89.
# add u to the list of visited vertices
[25] Escalera A, Hayes B, Prodanović M. A survey of reliability assessment techniques
UN = CUN ({u}); VN = VN ∪ {u} for modern distribution networks. Renew Sustain Energy Rev 2018;91:344–57.
For all v ∈ neighbors[u] [26] Ali Kadhem A, Abdul Wahab NI, Aris I, Jasni J, Abdalla AN. Computational
techniques for assessing the reliability and sustainability of electrical power
If cost[v] > cost[u] + CM (u, v)
systems: a review. Renew Sustain Energy Rev 2017;80:1175–86.
Then cost[v] = cost[u] + CM (u, v) and store the traceback node [27] Singh C, Mitra J. Composite system reliability evaluation using state space pruning.
End IEEE T Power Syst 2002;12:471–9.
End[1] [28] Green RC, Wang L, Alam M, Singh C. Intelligent state space pruning for Monte
End Carlo simulation with applications in composite power system reliability. Eng Appl
Artif Intel 2013;26:1707–24.
4. Return minimum cost path P from the super source node to the super sink node.
[29] Finon D, Locatelli C. Russian and European gas interdependence: could contractual
trade channel geopolitics? Energy Policy 2008;36:423–42.
[30] Weisser H. The security of gas supply—A critical issue for Europe? Energy Policy
References 2007;35:1–5.
[31] Le Coq C, Paltseva E. Assessing gas transit risks: russia vs. the EU. Energy Policy
[1] Yu X, Liang W, Zhang L, Reniers G, Lu L. Risk assessment of the maintenance 2012;42:642–50.
process for onshore oil and gas transmission pipelines under uncertainty. Reliab [32] Flouri M, Karakosta C, Kladouchou C, Psarras J. How does a natural gas supply
Eng Syst Safe 2018;177:50–67. interruption affect the EU gas security? a Monte Carlo simulation. Renew Sustain
[2] Dundulis G, Žutautaitė I, Janulionis R, Ušpuras E, Rimkevičius S, Eid M. Integrated Energy Rev 2015;44:785–96.
failure probability estimation based on structural integrity analysis and failure [33] Monforti F, Szikszai A. A MonteCarlo approach for assessing the adequacy of the
data: natural gas pipeline case. Reliab Eng Syst Safe 2016;156:195–202. European gas transmission system under supply crisis conditions. Energy Policy
[3] Chen Q, Zuo L, Wu C, Bu Y, Huang Y, Chen F, et al. Supply adequacy assessment of 2010;38:2486–98.
the gas pipeline system based on the Latin hypercube sampling method under [34] Lise W, Hobbs BF, van Oostvoorn F. Natural gas corridors between the EU and its
random demand. J Nat Gas Sci Eng 2019;71:102965. main suppliers: simulation results with the dynamic GASTALE model. Energy
[4] Han ZY, Weng WG. An integrated quantitative risk analysis method for natural gas Policy 2008;36:1890–906.
pipeline network. J Loss Prevent Proc 2010;23:428–36. [35] Lochner S. Modeling the European natural gas market during the 2009
[5] Vianello C, Maschio G. Quantitative risk assessment of the Italian gas distribution Russian–Ukrainian gas conflict: ex-post simulation and analysis. J Nat Gas Sci Eng
network. J Loss Prevent Proc 2014;32:5–17. 2011;3:341–8.
[6] Li J, Liang B, Li C, Yan M, Yu J. Calculation methods for the gas pipeline failure [36] Eser P, Chokani N, Abhari R. Impact of Nord Stream 2 and LNG on gas trade and
rate. J Petrol Sci Eng 2019;174:229–34. security of supply in the European gas network of 2030. Appl Energy 2019;238:
[7] Pesinis K, Tee KF. Statistical model and structural reliability analysis for onshore 816–30.
gas transmission pipelines. Eng Fail Anal 2017;82:1–15. [37] Praks P, Kopustinskas V, Masera M. Probabilistic modelling of security of supply in
[8] Wu J, Wu Z, Zhou R, Xu S. Probabilistic analysis of natural gas pipeline network gas networks and evaluation of new infrastructure. Reliab Eng Syst Safe 2015;144:
accident based on Bayesian network. J Loss Prevent Proc 2017;46:126–36. 254–64.
[9] Zarei E, Azadeh A, Khakzad N, Aliabadi MM, Mohammadfam I. Dynamic safety [38] Praks P, Kopustinskas V, Masera M. Monte-Carlo-based reliability and vulnerability
assessment of natural gas stations using Bayesian network. J Hazard Mater 2017; assessment of a natural gas transmission system due to random network
321:830–40. component failures. Sustain Resilient Infrastruct 2017;2:97–107.
[10] Fakhravar D, Khakzad N, Reniers G, Cozzani V. Security vulnerability assessment [39] Su H, Zio E, Zhang J, Li X. A systematic framework of vulnerability analysis of a
of gas pipelines using Discrete-time Bayesian network. Process Saf Environ 2017; natural gas pipeline network. Reliab Eng Syst Safe 2018;175:79–91.
111:714–25. [40] Su H, Zhang J, Zio E, Yang N, Li X, Zhang Z. An integrated systemic method for
[11] Wu W, Yang C, Chang J, Château P, Chang Y. Risk assessment by integrating supply reliability assessment of natural gas pipeline networks. Appl Energy 2018;
interpretive structural modeling and Bayesian network, case of offshore pipeline 209:489–501.
project. Reliab Eng Syst Safe 2015;142:515–24. [41] Yu W, Song S, Li Y, Min Y, Huang W, Wen K, et al. Gas supply reliability assessment
[12] Younsi K, Smati A. Intrinsic availability assessment of aged gas transmission of natural gas transmission pipeline systems. Energy 2018;162:853–70.
pipeline using bayesian update and stochastic process modeling. J Nat Gas Sci Eng [42] Yu W, Wen K, Min Y, He L, Huang W, Gong J. A methodology to quantify the gas
2017;45:659–69. supply capacity of natural gas transmission pipeline system using reliability theory.
[13] Hasan Sikder, Khan Faisal, Kenny Shawn. Probability assessment of burst limit Reliab Eng System Safety 2018;175:128–41.
state due to internal corrosion. Int J Pres Ves Pip 2012;89:48–58. [43] Chaudry M, Wu J, Jenkins N. A sequential Monte Carlo model of the combined GB
[14] Zhang P, Su L, Qin G, Kong X, Peng Y. Failure probability of corroded pipeline gas and electricity network. Energy Policy 2013;62:473–83.
considering the correlation of random variables. Eng Fail Anal 2019;99:34–45. [44] Rausand M, Høyland A. System reliability theory: models, statistical methods, and
[15] Seghier Mohamed El Amine Ben, Keshtegar Behrooz, Correia José AFO, applications. John Wiley & Sons; 2003.
Lesiuk Grzegorz, Jesus Abílio MP De. Reliability analysis based on hybrid [45] Li Y, Yao G. The design and management of gas transmission pipeline. Dongying:
algorithm of M5 model tree and Monte Carlo simulation for corroded pipelines: China University of Petroleum Press; 2009. in Chinese.
case of study X60 Steel grade pipes. Eng Fail Anal 2019;97:793–803.

12

You might also like