You are on page 1of 16

Journal of Natural Gas Science and Engineering 98 (2022) 104345

Contents lists available at ScienceDirect

Journal of Natural Gas Science and Engineering


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

Quantitative assessment of hydrate blockage risk in pipelines based on


reliability theory
Xu Duan a, Bohui Shi a, *, Chaoyu Ruan b, Junpeng Jia a, Shangfei Song a, Qingyun Liao a,
Yuchuan Chen c, Sihang Chen a, Junao Wang d, Haiyuan Yao d, Jing Gong a, **
a
State Key Laboratory of Natural Gas Hydrate, National Engineering Laboratory for Pipeline Safety, MOE Key Laboratory of Petroleum Engineering, Beijing Key
Laboratory of Urban Oil and Gas Distribution Technology, China University of Petroleum-Beijing, Fuxue Road No. 18, Changping District, Beijing, 102249, PR China
b
China Waterborne Transport Research Institute, Beijing, China
c
China’s Oil and Gas Exploration and Development Company (CNODC), China
d
Key Lab of Deepwater Engineering, CNOOC Research Institute, China

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

Keywords: Hydrate blockage significantly challenges flow assurance during subsea oil and gas operations. This paper pre­
Natural gas hydrate sented a novel probabilistic method to quantitatively evaluate hydrate blockage risk based on reliability theory.
Blockage risk The limit state equation of hydrate blockage involving hydrate volume fraction was established, after which Latin
Flow assurance
hypercube sampling (LHS) was used to determine the uncertain parameter samples. Moreover, the calculation
Reliability
Quantitative
efficiency was enhanced using the Proper Orthogonal Decomposition (POD) method, involving physical field
Probability reconstruction based on OLGA-CSMHyK simulation. Furthermore, this paper considered the uncertainty of
pipeline operating parameters, such as the pipeline inlet temperature (T), outlet pressure (P), mass flow rate
production reduction (Q), heat transfer coefficient (K), and initial droplet diameter (Dp). This method was used to
predict the probability of hydrate blockage at any time and at any position of the multiphase pipeline. Three risk
levels were defined, including hydrate formation risk, low risk of hydrate blockage, and high risk of hydrate
blockage. The calculation results indicated that Q displayed a significant impact compared with the other four
uncertain parameters, while the influence of topography also warranted attention. This work may provide a
reference for optimizing hydrate management strategies and ensuring safe pipeline operation.

industry, which has attracted substantial research attention (Ham­


merschmidt, 1934; Sloan et al., 2011).
1. Introduction Hydrate prevention methods mainly include traditional hydrate
avoidance and hydrate management techniques. The traditional method
Gas hydrate is a solid crystalline compound similar to ice, consisting involves heating the pipeline, setting insulation layers, and injecting
of host molecules (water molecules) and guest molecules (small gaseous large doses of thermodynamic hydrate inhibitors (THIs) to avoid hydrate
molecules) in a non-stoichiometric ratio at high pressure and low tem­ formation. However, these techniques present significant economic and
perature, but its crystal structure differs from ice (Chen et al., 2008; Shi environmental challenges (Frostman et al., 2003). Therefore, hydrate
et al., 2018; Sloan and Koh, 2007). Hydrate formation is undesirable in risk management has attracted substantial attention from the oil and gas
the low-temperature and high-pressure environments of deep-water oil industry as a more economical way of preventing the formation of hy­
and gas operations. The agglomeration, adhesion, and deposition of drate (Sloan, 2005). Hydrate risk management technology reduces the
hydrates in multiphase transportation pipelines may lead to pipeline safety margin by injecting low-dosage hydrate inhibitors (LDHIs) to
blockage, equipment failure, and even personal safety hazards. There­ prevent hydrate blockage rather than completely avoiding hydrate
fore, hydrate blockage presents a significant challenge in the oil and gas

; CSMHyK, The Colorado School of Mines Hydrate Kinetics model; LHS, Latin hypercube sampling; POD, Proper orthogonal decomposition; THIs, Thermodynamic
hydrate inhibitors; LDHIs, Low dosage hydrate inhibitors; KHIs, Kinetic hydrate inhibitors; AAs, Anti-agglomerants; HBFW, Hydrate blockage free window; HRE,
Hydrate risk evaluator; FAMUS, Flow Assurance by Management of Uncertainties and Simulation; PoF, Probability of Failure; GOR, Gas oil ratio; MC, Monte Carlo.
* Corresponding author.
** Corresponding author.
E-mail addresses: bh.shi@cup.edu.cn (B. Shi), ydgj@cup.edu.cn (J. Gong).

https://doi.org/10.1016/j.jngse.2021.104345
Received 12 May 2021; Received in revised form 10 November 2021; Accepted 22 November 2021
Available online 10 December 2021
1875-5100/© 2021 Elsevier B.V. All rights reserved.
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Nomenclature t The time variable, s


m The number of basic functions
φtransition The critical hydrate concentration as the onset of hydrate Ω The space domain
blockage formation according to the change of pressure λ The eigenvalue
drop B The two-point correlation matrix of realizations of the
φHyd Hydrate volume fraction in the pipeline random field
φTran The critical hydrate volume fraction in the pipeline z The total number of discrete points
ΔP Pressure drop W The data vector
WC Water cut C The kernel matrix
N The equal intervals in the first implementation step of LHS v The eigenvectors
algorithm E The matrix
Nr A uniformly distributed random number between 0 and 1 ξ The energy contribution rate
sam The sample value η Cumulative energy contribution rate
Probj The cumulative probability for the jth interval T Inlet temperature, ◦ C
ai The spectral coefficient P Outlet pressure, Pa
ψi The basis function Q Mass flow rate after production reduction, kg/s
n The total number of basis functions K Heat transfer coefficient, W/(m2⋅◦ C)
x The spatial variable, m Dp Initial droplet diameter, m

formation. LDHIs are divided into two main types: kinetic hydrate in­ hydrates (φtransition) as the onset of hydrate blockage formation according
hibitors (KHIs), which delay the nucleation and growth of hydrate to the ΔP changes. Chaudhari et al. (2018) defined a mathematical
crystals and prolong the induction time (Chen et al., 2018; Kelland, correlation known as the Hydrate Risk Evaluator (HRE) to indicate the
2018; Lou et al., 2012; Singh and Suri, 2020; Wang et al., 2019a), and hydrate risk level associated with oil-dominated systems based on
anti-agglomerants (AAs), which prevent hydrate particle aggregation transportability parameters, which could quantitatively define the hy­
(Huo et al., 2001; Kelland, 2006; Perrin et al., 2013; Sa et al., 2019; drate risk in certain conditions. However, since these deterministic an­
Wang et al., 2019b; Webber and Nagappayya, 2015). However, the risk alyses and predictions of hydrate blockage risk are performed in single
of hydrate formation remains when the inhibition performance of KHIs working conditions, it is difficult to quantitatively assess hydrate risk
is destroyed due to the high undercooling. Moreover, when the according to multiple parameters in the variable working conditions of
anti-aggregation effect of AAs is nullified due to free water bridge oil and gas gathering and transportation systems. The parameters
adhesion, there is a risk of hydrate blockage. Hydrate risk assessment, involved during the normal operational process, production reduction,
which includes qualitative and quantitative analysis, is critical for and multiphase mixed pipeline restart are uncertain and typically
optimizing hydrate risk management measures. include operational parameters (temperature, pressure, and flow rate),
Traditional qualitative analysis methods mainly involve factors fluid parameters (initial droplet diameter), and pipeline insulation
based on experimental data obtained using the high-pressure reactor or performance (heat transfer coefficient). The impact of uncertainty of
flow loop with macroscopic and microscopic measurement capability. these parameters on the hydrate risk should be considered.
Extensive research has been conducted regarding the qualitative anal­ Dejean et al. (2007) combined reliability ideas (unit failure and
ysis of hydrate blockage risk in oil-, gas-, and water-dominated systems. repair) and flow assurance issues (hydrate formation and wax deposi­
These studies assessed the various factors affecting hydrate blockage, tion) in the Flow Assurance by Management of Uncertainties and
including internal factors, such as liquid bridge force (Camargo and Simulation (FAMUS) project in 2007. Kim et al. (2016) used the Monte
Palermo, 2002; Liu et al., 2020), and external factors, such as mass flow, Carlo (MC) method based on offshore reliability data to predict the
liquid load, water cut (WC), pressure, AAs, and wax in the oil-dominated failure frequency and downtime of a subsea system. However, few
system (Boxall et al., 2008; Dellecase et al., 2008; Li et al., 2013; Liu studies are available regarding the uncertain variables affecting hydrate
et al., 2018). A series of experiments have been performed on the hy­ risk.
drate formation and blockage in gas-dominated systems, indicating the Most current work focuses on the conventional qualitative analysis of
presence of hydrate deposition and sloughing (Aman et al., 2016; Di factors impacting hydrate blockage using experimental research. Higher
Lorenzo et al., 2014a, 2014b). The WC significantly impacts hydrate or lower hydrate blockage risk in different conditions is usually
blockage in water-dominated systems (Sohn et al., 2015). Many factors described as a conclusion. Few studies have successfully predicted the
play a role in increasing the degree of hydrate formation in a system, hydrate blockage probability in multiphase pipelines at a specific time
significantly elevating hydrate particle adhesion, aggregation, and for a particular position, while the design specification reference is likely
deposition. Consequently, this causes a decline in the pipeline flow area, to produce an unduly conservative assessment (Li et al., 2009; Melchers
increases friction loss, and pressure drop (ΔP) fluctuation (Chen et al., and Jeffrey, 2008). Therefore, a probabilistic description based on
2015; Song et al., 2017). However, quantitatively evaluating hydrate reliability may be the most adequate method to analyze hydrate
blockage risk based on specific work conditions remains challenging. blockage risk in flow assurance issues (Dejean et al., 2005). Conse­
Quantitative hydrate blockage risk analysis is effective in formu­ quently, this paper establishes a probabilistic analysis based on reli­
lating safety strategies for abnormal accident prevention, allowing safer, ability to quantitively evaluate the risk of hydrate blockage in pipelines
more economical hydrate risk management. Regarding gas-dominated using the limit state equation of hydrate formation volume. Moreover,
systems, Wang et al. (2018a) and (2018b) proposed the hydrate LHS is applied for sampling and coupled with the fast and accurate POD
blockage free window (HBFW) to characterize the hydrate blockage method to improve the calculation efficiency.
degree of the annular flow in deep-water wellbores. Furthermore, the The key to assessing hydrate blockage risk is to establish the limit
hydrate blockage risk was evaluated by calculating the hydrate layer state equation, which can refer to HBFW, φtransition, and HRE. However,
thickness and determining the wellhead pressure (Zhang et al., 2020a, HBFW ignores the deposition of hydrates formed from liquid droplets,
2020b). Joshi et al. (2013) proposed the hydrate formation and plugging while the HRE formula is complex and only suitable for specific working
process for high WC systems and defined a particular concentration of conditions. The hydrate volume fraction is the most important and

2
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

intuitive factor affecting hydrate blockage. As a theoretical basis for could be defined as the hydrate volume fraction value when the slope of
hydrate risk analysis, the hydrate blockage mechanism in oil- and water- the relative ΔP changed sharply based on 125 sets of ExxonMobil flow
dominated systems indicates that the hydrate volume fraction is an loop datasets using methane, brine, and a sample of Conroe crude oil as
important factor affecting hydrate blockage. It may be more appropriate the experimental materials in various operating conditions (Fig. 1). This
to establish the limit state equation by combining the hydrate volume was reasonable since a significant increase in ΔP indicated hydrate
fraction with reliability theory. blockage in the pipeline.
The hydrate volume fraction, used as a criterion for judging hydrate As shown in Fig. 1, in these working conditions, the critical hydrate
blockage risk, can be calculated using different models based on intrinsic volume fractions were 10%, 14%, and 20% at mixture velocities of 0.73
kinetics (Boxall et al., 2008; Lekvam and Ruoff, 1993; Vysniauskas and m/s, 1.13 m/s, and 1.74 m/s, respectively, while no sudden ΔP changes
Bishnoi, 1983, 1985; Yang et al., 2011), mass transfer (Skovborg and were evident when the hydrate volume fraction reached 35% at a
Rasmussen, 1994; Turner et al., 2009), heat transfer (Mochizuki and mixture velocity of 2.87 m/s. This phenomenon indicated that the
Mori, 2006; Mori, 2001; Peng et al., 2007), or combined methods (Freer critical volume fraction was related to the flow velocity and operating
et al., 2001; Ke et al., 2019; Mu et al., 2014; Shindo et al., 1993; Yin conditions of the system. Therefore, it was unreasonable to use a fixed
et al., 2018). The hydrate volume fraction value becomes more consis­ critical hydrate volume fraction value, and this value had to be deter­
tent with practical engineering only when the prediction of the hydrate mined according to the specific working conditions. This paper proposed
volume fraction is combined with the hydraulic flow characteristics a quantitative method where the critical parameter was temporarily
calculation. The Colorado School of Mines Hydrate Kinetics model determined based on the data in the literature. However, the hydrate
(CSMHyK) was embedded into the transient multiphase flow simulator, blockage limit state equation involving multi-factors requires further
OLGA (Davies et al., 2009; Turner et al., 2005; Zerpa et al., 2012), which examination to be coupled and updated into this evaluation method in
can predict hydrate formation volume in different flow patterns (Zerpa the future.
et al., 2013). The quantitative blockage risk assessment of the pipeline is
more reliable and reasonable when using the hydrate volume fraction 2.2. The random sampling of the uncertain parameters
values calculated via OLGA-CSMHyK. In this paper, OLGA-CSMHyK is
used to verify the POD method and obtain the POD sampling matrix, The uncertainty of pipeline operating parameters should be consid­
obtaining the hydrate volume fraction at any time along the pipeline. ered during the evaluation of hydrate risk based on reliability. The LHS
Therefore, to systematically discuss and examine the relationship algorithm, originally developed for MC simulation, is used for the
and interaction between reliability and flow assurance during practical sampling calculation of uncertain parameters.
production, this paper presents a novel method to quantitatively assess LHS is a multi-dimensional stratified sampling method widely used
the hydrate blockage risk by establishing the limit state equation in practical engineering, in which the distribution of sampling points is
involving the hydrate volume fraction. Then, using a quick and accurate uniform, and convergence can be achieved quickly, allowing effective
POD method for a large sample trial, considering the uncertainty of the parameter selection for computer models (Iman and Conover, 1980;
parameters via LHS, the hydrate blockage probability at any time and Mckay et al., 1979; Minasny and McBratney, 2006). LHS involved
any position of the pipeline can be calculated. This study provides stratifying the probability distribution of input and dividing the cumu­
further insight into hydrate blockage risk during multiphase pipe flow, lative curve into N contour intervals on the cumulative probability scale
which is essential for safe and economic hydrate risk management. [0,1], followed by random sampling in the divided interval. The number
of the intervals was equal to the number of sample points, ensuring that
2. Methods the probability of each value was 1/N. An example where N is 5 is shown
in Fig. 2. The specific implementation steps of the LHS algorithm were as
The first step of this hydrate blockage risk assessment model was to follows (Minasny and McBratney, 2006; Xu et al., 2005):
establish the limit state equation based on the hydrate volume fraction.
Then, considering the uncertainty of the parameters, LHS was applied to (1) The probability space of the variable was divided into N equal
produce massive samples according to the variable distribution. More intervals.
than 10,000 trials were involved in calculating the hydrate risk proba­ (2) For the jth interval, the cumulative probability could be written
bility, which was rapidly achieved using the POD algorithm. Finally, the as (Minasny and McBratney, 2006):
hydrate blockage risk was assessed quantitatively. The equations, ( )
1 (j − 1)
methods, and solutions are described below. Probj = Nr + (2)
N N

2.1. Establishing the limit state equation of hydrate blockage where Nr is a uniformly distributed random number between
0 and 1.
The risk was characterized quantitatively by calculating the failure (3) The sample value, sam, was obtained according to the probabil­
probability of the system in extreme conditions. For the hydrate ity, using the inverse function of the distribution function, F− 1
blockage risk in flow assurance, the pipeline blockage was considered a (Minasny and McBratney, 2006).
failed state. The hydrate blockage risk limit state equation was devel­
sam = F− 1 (Prob) (3)
oped based on the identified failure mode (Hasan et al., 2012), expressed
as Eq. (1). The hydrate volume fraction, as the key parameter that causes
hydrate blockage, whether in an oil-, water- or gas-dominated system,
(4) Random values were obtained.
was selected to establish the limit state equation. Here, W ≤ 0 referred to
a blocked pipeline, while W > 0 indicated no hydrate blockage risk in
Fig. 3 shows an example of the LHS and MC sampling results during
the pipeline.
normal distribution with a mean value of 14 and a standard deviation of
W = φTran − φHyd (1) 1, generating 2000 and 10,000 random numbers, respectively. LHS was
distinctly more consistent with the distribution and exhibited a better
where φTran is the critical hydrate volume fraction in the pipeline, and effect. Therefore, the LHS method was selected for sampling the un­
φHyd is the hydrate volume fraction in the pipeline. certain parameters during pipeline operation. More samples increased
The key to this limit state equation was to determine the appropriate the parameter distribution, while the consumed simulation workload
φTran. Srivastava (2018) and Qin et al. (2019) indicated that the φTran was also higher. A sample number of 10,000 not only ensured the

3
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Fig. 1. Illustration of the critical hydrate volume fraction at which the relative ΔP slope changed during tests at different mixture velocities. The relative ΔP denotes
the ratio of pressure drop with hydrates to the pressure drop at hydrate formation onset. PoF signifies the probability of failure, and WC represents the water cut.
Reproduced from Qin et al. (2019) and Srivastava (2018).

it can be expressed as shown in Eq. (4) (Berkooz et al., 1993).



n
f (x, t) = ai (t)ψ i (x), n ≤ ∞ (4)
i=1

where f(x, t) represents the physical field, ai(t) is the spectral coefficient,
ψ i(x) is the basis function, and n is the total number of basis functions,
while x and t are the spatial variable and time variable, respectively.
The most prominent characteristic of the POD method is that any
physical system can be expressed in the form of linear series expansion,
while the basis functions can be used to maximize the approximation of
the reconstructed physical quantity to the actual physical field using
least square. Therefore, it is necessary to find a set of basis functions
satisfying Eq. (5) to achieve optimal reconstruction (And and Lee, 1998;
Cizmas et al., 2003).
(⃦ ⃦2 )
⃦ ∑m ⃦
⃦ ⃦
ψ ∈ argmin 〈⃦f (x, t) − ai (t)ψ i (x)⃦ 〉 , m ≤ n (5)
⃦ i=1

Fig. 2. Example of LHS displaying normal distribution where the cumulative
probability is divided into five equal layers, and a random number is drawn where ||⋅|| is the L2 norm, ||f ||2=(f, f), (⋅,⋅) denotes an inner product,
from each layer. ∫T
and〈⋅〉denotes an ensemble average (such as: 〈f〉 = T1 0 f(x, t)dt), while
m is the number of basic functions that approximately describe f(x, t).
uniformity and rationality of normal distribution but also displayed an Eq. (5) has been proven mathematically equivalent to Eq. (6)
acceptable consumed simulation workload. (Lumley, 1981).

〈(f , ψ )2 〉
2.3. Fast numerical simulation of large samples ψ ∈ argmax (6)
(ψ , ψ )
Predicting the hydrate volume fraction is essential since it is used to The search for the extreme value of Eq. (6) could be transformed into
estimate the limit state of hydrate blockage. The OLGA-CSMHyK soft­ an eigenvalue problem. The necessary condition demanded that ψ was
ware is recognized as a useful tool to predict the hydrate formation an eigenfunction of the two-point correlation tensor (Berkooz et al.,
volume in transition conditions for a multiphase pipeline flow. During 1993).
reliability analysis, at least 10,000 or more predictions of the hydrate ∫
volume fraction should be performed to calculate the failure probability 〈f (x)f * (x )〉ψ (x )dx = λψ (x) (7)
′ ′ ′

based on 10,000 samples of a parameter satisfying a certain distribution. Ω

However, it is unrealistic to predict 10,000 hydrate volume fractions in a


where Ω is the space domain, * denotes a complex conjugate, and λ is the
pipeline using OLGA-CSMHyK due to extensive consumption time and
eigenvalue.
high data storage requirements. Therefore, this study adopted the POD
algorithm to reduce the computational complexity. POD, also known as B(x, x ) = 〈f (x)f * (x )〉 (8)
′ ′

principal component analysis, represents a statistical method that ex­


tracts the optimal basis from the observation set to achieve dimension­ where B(x, x′ ) is defined as the two-point correlation matrix of the
ality reduction and data compression, simplifying the physical model realization of the random field, which forms the kernel of the eigenvalue
and saving computational load (Berkooz et al., 1993). The basic POD problem.
principle involves the following (Berkooz et al., 1993; Rathinam and Eq. (7) could then be written as Eq (9), representing a second type of
Petzold, 2003): Fredholm integral equation. The optimal basis function of POD could be
For any physical field (such as velocity, temperature, and pressure), obtained when the function f(x) was known.

4
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Fig. 3. The frequency histograms of the 2000 and 10,000 random numbers are extracted respectively using LHS and MC sampling during normal distribution with a
mean value of 14 and a standard deviation of 1.


rate ξ and the cumulative energy contribution rate η, to quantitatively
(9)
′ ′ ′
B(x, x )ψ (x )dx = λψ (x)
Ω express the cumulative contribution of the basis function to the total
energy:
The given function to be decomposed is usually discrete, while the
/
observations forming the data set are vectors rather than continuous ∑
n

functions. For example, Ω= (x1, x2, …, xz), where xz is the zth grid point, ξi = λi λi (14)
and z is the total number of discrete points. Additionally, f(x, t) is the
i=1

data vector W = [f(x1, tm), f(x2, tm), ⋅⋅⋅, f(xz, tm)]T, written as matrix W. /

m ∑
n
⎡ ⎤ ηm = λi λi , m ≤ n (15)
f (x1 , t1 ) f (x1 , t2 ) ⋯ f (x1 , tm )
i=1 i=1
⎢ f (x2 , t1 ) f (x2 , t2 ) ⋯ f (x2 , tm ) ⎥
W =⎢ ⎥ (10)
⎣⋮ ⋮ ⋱ ⋮ ⎦ According to the basis function energy from large to small, the first m
f (xz , t1 ) f (xz , t2 ) ⋯ f (xz , tm ) group accounted for most of the energy. Therefore, the function could be
expressed as Eq. (16) with high precision, denoting the reconstructed
The data set was obtained via numerical simulation and experi­
physical field.
mental research. In the discrete case, the tensor product matrix, C, was
used as the kernel (Cizmas et al., 2003; Sirovich, 1987). ∑
m
f (x, t) ≈ ai (t)ψ i (x) (16)
1
(11)
i=1
C= WW T
m For univariate problems, the spectral coefficients ai were obtained
The snapshot method proposed by Sirovich (1987) and Holmes using the POD-interpolation method.
(1996) was used to obtain the eigenvectors v and corresponding ei­
t − ti
genvalues λ, using Eq. (12) a(t) = a(ti ) + [a(ti+1 ) − a(ti )] (17)
ti+1 − tt
Cv = λv (12)
The obtained spectral coefficients were substituted into Eq. (16)
The matrix, E, could then be calculated. Each column of E was reconstruction to acquire the low-order POD model results.
√̅̅̅̅
divided to be orthonormal via λi ,(i = 1, …,N), obtaining the basis The POD premise for reconstructing the physical field involved
functions [ψ 1, ψ 2, …, ψ z] (Li et al., 2020). selecting a small number of basic samples for the numerical calculation
to obtain the sampling matrix required for POD. This sampling matrix
E = W⋅v (13) consisted of hydrate volume fractions at any simulation time and posi­
The eigenvalue λi represents the energy contained in its basis func­ tion along the multiphase pipelines. Based on the distribution charac­
tion. This study defined two parameters, namely the energy contribution teristics of the parameters, LHS was performed to generate plenty of
random numbers. About ten samples were selected for OLGA-CSMHyK

5
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

calculation to obtain the POD sampling matrix. The basis functions and hydrate volume fractions could be obtained via POD accelerated
spectral coefficients were obtained via POD to construct the physical solution of reconstructed physical field.
field. The desired results for the remaining samples were obtained (5) The POD sampling matrix was obtained after 10,000 trials with T
algebraically. as a variable.
(6) The hydrate volume fractions of five trials were randomly
selected from the calculation results of 10,000 trials and
2.4. Calculating the hydrate blockage probability
compared with the OLGA-CSMHyK results. The next step
commenced when the error was within the allowable deviation
Establishing the limit state equation using reliability theory was
range. Otherwise, the second step was repeated with an increased
combined with random sampling and fast numerical simulation to
number of basic samples for POD calculation.
calculate and quantitatively characterize the probability of hydrate
(7) The probability of hydrate blockage was calculated using the
blockage in pipelines.
limit state equation.
All the parameters, including the pipeline operation parameters,
fluid parameters, and pipeline insulation performance, provided
3. Results and discussion
fundamental data to calculate the hydrate blockage probability. In this
study, this information is derived from the uncertainty of individual
During oil and gas production, hydrate formation and plugging are
variable parameters, such as T, P, Q, K, and Dp. When one parameter was
more prone to occur in transient operating conditions, such as shut-in or
selected as uncertain during calculation, the others represented deter­
production loss, since the fluid temperature in the pipeline may drop
ministic parameters. The lognormal distribution fit the distribution
below that of the hydrate formation. This study analyzes the risk of
characteristics of the Dp (Chen et al., 2020). Since the normal distribu­
hydrate blockage due to production loss. Fig. 5 shows the geometry of
tion was generally universal, the other variables were considered as
the simulated pipeline in terms of elevation, as well as the horizontal
following the normal distribution. For example, with T as a variable, the
length, which was 5780 m, while the inner diameter was 282.7 mm.
calculation process of the hydrate blockage probability is summarized
Table 1 displays the critical parameters of the simulated pipeline. The oil
below (Fig. 4).
and gas components in the pipeline are summarized in Table 2 and
Table 3, respectively, displaying a molecular gas-to-oil ratio of 9:1 and a
(1) LHS was used to generate 10,000 random samples according to
WC of 30%.
the distribution characteristics.
In addition to the specific data, such as the pipeline data (Table 1)
(2) The basic samples were selected at equal intervals, covering the
and the fluid data (Tables 2 and 3) employed during the calculation of
range involved in parameter distribution. The number of basic
the hydrate blockage risk probability, one of uncertain T, P, Q, K, and Dp
samples was as small as possible to reduce the computational
parameters should be selected, while the unselected parameters should
workload and achieve fast calculation.
be deterministic. To better understand and analyze the hydrate forma­
(3) The basic samples were used as the OLGA-CSMHyK input pa­
tion and flow characteristics, an OLGA-CSMHyK simulation without
rameters to calculate the hydrate volume fraction at any time and
uncertain parameters with a 400-min transition operating condition was
any position along the pipeline, as the POD sampling matrix. The
performed. Q decreased linearly from 56 kg/s at 50 min to 14 kg/s
basis function and spectral coefficient were then calculated via
within 15 min from normal operation. The T, P, K, Dp values were 58 ◦ C,
POD decomposition to reconstruct the physical field.
13 MPa, 30 W/(m2⋅◦ C), and 40 μm, respectively. The final steady-state
(4) When the random samples were equal to the basic samples, the
operation condition is shown in Fig. 6, while the transition result of
hydrate volume fraction could be obtained directly via the pre­
the hydrate volume fraction during the 400-min simulation is displayed
vious step using OLGA-CSMHyK. When the random samples were
in Fig. 7.
not the basic samples, namely the remaining samples, their

Fig. 4. The flow chart of the hydrate risk probability calculation. Here, T is considered an uncertain parameter, while other parameters are deterministic values. The
calculation processes for the other parameters are similar.

6
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Fig. 6. The pressure-temperature diagram with the hydrate equilibrium curve


Fig. 5. The geometry of a multiphase transportation pipeline. and the normal and loss operation conditions at the 400 min simulation time.

Table 1
Key parameters of the pipeline in the case study.
Parameter Value Unit

Inner diameter 282.7 mm


Roughness 0.15 mm
Line length 5780 m
Ambient temperature 6.5 ◦
C

Table 2
Gas composition (mol%).
N2 CO CO2 C1 C2 C3 iC4 iC5 nC6+

1.53 2.05 0.89 89.02 3.07 3.06 0.33 0.04 0.01

Table 3
–20 # Diesel oil composition (mol%).
C7 C8 C9 C10 C11 C12 C13 C14

1.05 0.92 4.60 11.39 11.72 12.23 9.93 6.90


C15 C16 C17 C18 C20 C24 C28+ –
4.86 4.37 4.64 5.62 10.73 9.76 1.28 – Fig. 7. The hydrate volume fractions along the pipeline during the 400 min
transient simulation.

As shown in Fig. 6, hydrate formation was evident at a lower tem­ 3.1. Random variable generation
perature and higher pressure on the left of the hydrate equilibrium
curve, presenting a hydrate blockage risk. In normal production condi­ As mentioned above, the T, P, Q, K, and Dp represented vital pa­
tions, the pipeline was in a safe area without hydrate formation. How­ rameters affecting hydrate formation and were selected for the uncer­
ever, when Q was reduced to 14 kg/s, the section from the position of tainty analysis to calculate the hydrate blockage probability. Their
2100 m to the pipeline outlet fell in the hydrate formation zone with a distribution characteristics are shown in Table 4. When one of the pa­
risk of hydrate blockage. rameters was uncertain, the rest were fixed values. For example, when T
As shown in Fig. 7, the hydrate volume fraction values at the end of was a random variable, P, Q, K, and Dp were certain as the mean value
pipeline exceeded zero at around 80 min, indicating hydrate formation. shown in Table 4. LHS was used to obtain 10,000 samples of each var­
From 80 min to 160 min, these values increased rapidly and substan­ iable, as shown in Fig. 8.
tially due to the high subcooling. After 200 min, the hydrate volume
fraction decreased and finally stabilized at about 350 min. Furthermore,
the hydrate formation was affected by the intrinsic driving force, as well Table 4
as the heat and mass transfer. The entire system reached a steady state at Distribution types and parameters of the stochastic variables.
around 400 min, while the hydrate volume fractions along the pipeline
Parameters Distribution type Mean Standard deviation
during the 80 min–160 min period were distinctly higher than other
T (◦ C) Normal distribution 58 2
time periods. Therefore, the 80 min–160 min period was selected to
P (MPa) Normal distribution 13 1
analyze the hydrate risk probability in the subsequent sections. Q (kg/s) Normal distribution 14 1
K (W/(m2⋅◦ C)) Normal distribution 30 1
Dp (μm) Lognormal distribution 40 10

7
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Fig. 8. The random parameter sampling results using LHS.

3.2. Obtaining the POD sampling matrix each random variable are shown in Table 5.
Here, T was used as an example. Since the sample range was between
The basic samples were selected based on the parameter range, that 50 ◦ C and 66 ◦ C (Fig. 8a), the interval was 2 ◦ C to obtain uniform basic
is, the range of random numbers generated via LHS. These samples samples, as shown in Table 5. When T was a random variable, P, Q, K,
needed to conform to the characteristics of uniformly including the and Dp represented the determined mean value, as shown in Table 4. The
range of parameters and covering the physical field. The basic samples of normal mass flow rate from 56 kg/s to a Q value of 14 kg/s over a 400-
min period was simulated based on the basic T samples using OLGA-
CSMHyK to obtain the hydrate volume fractions at any time along the
Table 5 pipeline. The simulation results represented the POD sampling matrix.
Basic samples of the parameters. Fig. 9a presents the results of the hydrate volume fractions along the
Parameters Basic samples
pipeline at the basic sample condition of T at 120 min. Similarly, the P,
Q, K, and Dp results are shown in Fig. 9b–e. The basic samples of these
T (◦ C) 50 52 54 56 58 60 62 64 66
five parameters exhibited different effects on the hydrate volume frac­
P (MPa) 9 10 11 12 13 14 15 16 17
Q (kg/s) 9 10 11 12 13 14 15 16 17 18 tions, as shown in Fig. 9. The similarities illustrated in Fig. 9a–e shows
K (W/(m2⋅◦ C)) 26 27 28 29 30 31 32 33 34 no hydrate formation in the 0–2000 m range since the temperature and
Dp (μm) 10 20 30 40 50 60 70 80 90 100 pressure in this section of pipe do not enter the hydrate formation zone.

8
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Fig. 9. The hydrate volume fractions along the pipeline under the basic sample condition after running for 120 min with different variables.

The hydrate volume fraction shown in Fig. 9a initially increased first, longer than at 15–18 kg/s. At 18 kg/s, the highest hydrate volume
followed by a decline along the pipeline. The higher hydrate volume fraction was about 17%, significantly lower than the highest value of
fraction between 3000 m and 4000 m was mainly attributed to the fact 64% at 9 kg/s. Since Q affected the flow and heat transfer of the fluid,
that this was a low-lying area where hydrate accumulation occurred. the hydrate volume fraction changed more significantly in the 2000
Distinct hydrate volume fraction distribution was evident in the 4000 m–4000 m pipe section in conjunction with an increase in Q. At a higher
m–6000 m pipeline section at different T values (Fig. 9a). The basic Q, more heat was generated by flow friction, increasing the fluid tem­
sample value range of T exceeded that of the hydrate formation tem­ perature and reducing the driving force behind hydrate formation.
perature, displaying little influence on the hydrate volume fractions K characterized the heat dissipation ability of the fluid in the pipeline
within this range. The influence of P was mainly concentrated at around to the surrounding environment, while its influence on the hydrate
3000 m, while the position of the hydrate formation region shifted volume fraction was not obvious in this range (Fig. 9d). Fig. 9d indicates
slightly due to pressure change (Fig. 9b). that more hydrates accumulate in low-lying sections. For the different Dp
A change in Q modified the hydrate volume fraction and altered the values, the hydrate volume fraction curve only differed in the 3000
range of the hydrate formation region (Fig. 9c), evidencing significant m–4000 m section of the pipeline while remaining unaffected in other
hydrate volume fraction variation in the relevant range. When Q was positions (Fig. 9e). The Dp primarily impacted the hydrate formation
9–14 kg/s, the section of pipeline displaying hydrate formation was rate by determining the interface area of the water and oil phases, while

9
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

minimally affected the hydrate volume fraction in the 10 μm–100 μm Dp 2.14 GB of data storage was required, significantly improving the
range. In the specific basic sample range listed in Table 5 and consid­ computation efficiency. This was because only the basis function and
ering the fluctuation range, the hydrate volume fraction impact in corresponding spectral coefficient were stored during the calculation
descending order was Q, P, K, T, and Dp. process, reducing the necessity for storing data.
In addition, the high hydrate volume fraction in the low-lying posi­
tion at about 3500 m indicates that the influence of topography cannot 3.4. Hydrate blockage risk probability
be ignored. The influence of gravity facilitated high-velocity hydrate
particles in the downhill section, while those in the uphill section dis­ The hydrate blockage risk quantification was realized via probabi­
played low velocity, allowing easy particle accumulation in low-lying listic calculations based on the hydrate volume fraction of a large sam­
sections and leading to a larger hydrate volume fraction (Shi et al., ple. After completing the hydrate formation calculation of 10,000
2021). Therefore, the low-lying pipeline sections are more prone to samples for each variable, the hydrate blockage risk probability, using
hydrate blockage risk. five individual parameters, was calculated according to the limit state
equation.
3.3. Verification of the POD calculation In this work, the critical hydrate volume fraction was mainly ob­
tained according to the statistical data in the available literature (Qin
After obtaining the POD basic sample matrix, the physical field was et al., 2019). Here, three levels of critical volume fraction values were
reconstructed by solving the basis function and spectral coefficient, defined, namely 0%, 10%, and 30%. The corresponding limit state
followed by the hydrate volume fraction calculation after 10,000 trials. equation results included hydrate formation risk, low risk of hydrate
The calculation results of five trials were randomly selected for each blockage, and high risk of hydrate blockage. However, hydrate blockage
variable to be compared with the OLGA-CSMHyK results. Some of the is not only related to the hydrate volume fraction in practical engi­
verification results are shown in Fig. 10. The calculated results using this neering but also factors, such as viscosity, WC, liquid holdup, density,
method were mostly consistent with those obtained using OLGA- and mixture velocity. Follow-up work should consider establishing a
CSMHyK. The maximum relative error was 3.32% (Fig. 10b), indi­ suitable multi-factor limit state equation combined with experimental
cating the reliability of the constructed POD physical field. and simulation data to better define the blockage state.
The time and the amount of data storage required for the calculation Fig. 11 shows the probability of hydrate formation risk along the
involving 10,000 samples using OLGA-CSMHyK simulation and the POD pipeline, with different selected variables as the uncertain parameters,
method were compared. One case simulation lasted an average of 3 min where the φTran in Eq. (1) is set as 0%. The probability of hydrate for­
using OLGA-CSMHyK. Therefore, simulating five cases 10,000 times in mation changed with time and location, and almost the entire pipe
parallel would last about 100 h in total and require approximately section from the outlet to the 2000 m position was at risk. At 80 min, the
159.37 GB of data storage. However, the calculation for reconstructing probability of hydrate formation in the pipeline section around
the physical field using the POD method lasted about 3.8 h, while only 3000–4000 m was about 0.5 since this represented a low-lying section in

Fig. 10. A comparison between the POD and OLGA simulation results in different conditions.

10
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Fig. 11. The probability of hydrate formation risk along the pipeline after running for 80–160 min with different variables.

the pipeline, promoting hydrate accumulation due to gravity (Fig. 11a). hydrate formation and flow in the pipeline, the probability of hydrate
The probability of hydrate formation risk at the end of the pipeline was formation risk gradually increased, as well as pipeline area where the
about 1.0 since the end of the pipeline entered the hydrate formation probability of hydrate formation was above 0. At 136 min, the proba­
region due to a gradual decrease in the fluid temperature. With the bility of hydrate formation risk in the pipeline segment after 3000 m was

11
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

1.0 since the fluid temperature in the pipeline constantly decreased, risk exhibited almost no fluctuation. Contrarily, the probability of hy­
allowing more sections to enter the hydrate formation region. Figs. 11b drate formation risk shown in Fig. 11c was less than 1.0 at about 3500 m
and d show a similar trend regarding the change in hydrate formation during 136–180 min, which was consistent with the results shown in
risk probability. Fig. 11e shows that the probability of hydrate formation Fig. 9c. Q generally had the most significant influence on the risk of

Fig. 12. The probability of low risk of hydrate blockage along the pipeline after running for 80–160 min with different variables.

12
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

hydrate formation, which was reflected in the larger fluctuation of the set as 10%. At 80 min, the probability of low risk of hydrate blockage
probability curve, while the probability of hydrate formation risk was along the pipeline was at 0, indicating that the hydrate volume fraction
lower than that of other factors. The influence of Dp was minimal, as was below 10% (Fig. 12a) even though the hydrate formation risk
indicated by the relatively stable probability curve. probability exceeded 0 (Fig. 11a). The low-risk areas of hydrate
Fig. 12 shows the probability of low risk of hydrate blockage along blockage were mainly concentrated in the low-lying pipeline section at
the pipeline exposed to different variables, where the φTran in Eq. (1) was about 3500 m. At 84 min, the probability of low risk of hydrate blockage

Fig. 13. The probability of high risk of hydrate blockage along the pipeline after running for 80–160 min with different variables.

13
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

in the low-lying section was approximately 0.4, and 1.0 at around 88 these levels can be predicted using the proposed quantitative
min, indicating that the hydrate volume fraction of the low-lying sec­ assessment method.
tions reached, or even exceeded, 10%. Although the end of the pipeline (2) When calculating the hydrate blockage probability in a reduced
remained perpetually in the hydrate formation region (Fig. 11a), the production scenario, one of five key parameters is selected to
probability of low risk is basically 0 since the hydrate accumulated in the serve as the uncertain parameter, while the others are considered
low-lying sections due to gravity. A similar trend was evident for the deterministic, including T, P, Q, K, and Dp. The results indicate
probability of low risk of hydrate blockage, as shown in Fig. 12b, d, and that Q has the most significant impact on the hydrate blockage
e. Similar to the results shown in Fig. 11, the uncertainty of Q had the risk.
most significant impact on the low risk of hydrate blockage, while the (3) The results indicate that the end of the pipeline perpetually enters
probability was lower than that of other factors (Fig. 12c). the hydrate formation region first since it presents the lowest
Fig. 13 shows the probability of high risk of hydrate blockage, where temperature with a higher hydrate formation risk. The low-lying
the φTran in Eq. (1) was set as 30%. The probability variation trend under sections present the highest blockage risk due to more significant
each variable was generally consistent. Except for the low-lying sections, hydrate particle accumulation.
the other pipeline sections were not at high risk of hydrate blockage.
Compared with the low-risk probability calculation results in Fig. 12, the Therefore, this quantitative analysis of hydrate blockage risk can
high-risk range of hydrates was narrower, mainly concentrated in low- guide the implementation of the necessary management measures by
lying spots, with a probability above 0.8, presenting the highest pipe­ injecting hydrate inhibitors to avoid hydrate blockage. Furthermore, it
line blockage risk. The high-risk area gradually narrowed, slowly can provide a valuable reference for designing cost-effective hydrate risk
decreasing the probability over time since the hydrate formation and management guidelines. Future research should consider multivariate
flow in the pipeline represented a dynamic process, and the hydrates uncertainties to establish a multi-factor limit state equation of hydrate
were transported downstream with the flow direction. The probability blockage that can efficiently combine the reliability and hydrate prob­
of high risk of hydrate blockage shown in Fig. 13c was lower than other lem in flow assurance.
variables, which was similar to the hydrate formation risk and low-risk
of hydrate blockage characteristics mentioned above. Credit author statement
In summary, the hydrate risk probability at different pipeline posi­
tions varied and changed with time. The low-lying sections and outlets Xu Duan: Conceptualization, Methodology, Software, Writing –
of the pipeline preferentially entered the hydrate formation area, original draft. Bohui Shi: Writing – review & editing, Supervision,
resulting in a higher probability of hydrate formation. The low-lying Funding acquisition. Chaoyu Ruan: Methodology, Visualization. Jun­
pipeline sections displayed both the low-risk and high-risk with high peng Jia: Data curation, Formal analysis. Shangfei Song: Validation.
probability, necessitating careful attention when implementing hydrate Qingyun Liao: Data curation. Yuchuan Chen: Validation. Sihang Chen:
control measures. Compared with other parameters, the change in Q Data curation. Junao Wang: Software. Haiyuan Yao: Software. Jing
directly affected the transient process, although had a lower risk, dis­ Gong: Supervision, Resources, Funding acquisition.
played the most significant impact on hydrate blockage risk. Dp
exhibited a minimal impact, while the risk probability curve remained
relatively stable without much fluctuation. Declaration of competing interest
Calculating the risk probability at any time and any geographical
location can provide a reference for hydrate risk management, such as The authors declare that they have no known competing financial
determining the injection time and injection rate of hydrate inhibitors. interests or personal relationships that could have appeared to influence
When hydrate formation in pipelines is unacceptable, more hydrate the work reported in this paper.
inhibitors should be injected before reducing production to decrease the
probability of hydrate formation. When the pipeline is allowed to
operate at a low risk of blockage, control measures should focus on Acknowledgments
injecting inhibitors into the low-lying sections to maintain the proba­
bility of low risk of hydrate blockage within an acceptable range. When This work was supported by the National Natural Science Foundation
of China [grand number 51874323, U20B6005, 52104069]; the Beijing
the pipeline is allowed to operate at a high risk of blockage, the injection
of fewer inhibitors renders the high risk of hydrate blockage probability Municipal Natural Science Foundation [grand number 3192027]; the
National Key Research and Development Plan [grand number
acceptable. Therefore, this method for quantitatively assessing hydrate
blockage risk based on reliability is used to determine the injection time 2016YFC0303704]; the Science Foundation of China University of Pe­
troleum, Beijing [grand numbers 2462020YXZZ045,
and rate of inhibitors, rendering hydrate risk control more time-efficient
and economical. 2462020XKBH012]; and the 111 Project [grand number B18054], all of
which are gratefully acknowledged.
4. Conclusion
References
This paper proposes a novel method for quantitatively assessing the
Aman, Z.M., et al., 2016. Hydrate formation and deposition in a gas-dominant flowloop:
hydrate risk in multiphase transportation pipelines based on reliability
initial studies of the effect of velocity and subcooling. J. Nat. Gas Sci. Eng. 35,
theory in the form of probability. Probabilistic calculations are per­ 1490–1498. https://doi.org/10.1016/j.jngse.2016.05.015.
formed by establishing the limit state equation involved with hydrate And, H.M.P., Lee, M.W., 1998. An efficient method of solving the Navier–Stokes
equations for flow control. Int. J. Numer. Methods Eng. 41 (6), 1133–1151.
volume fractions and combining with parameter uncertainty via LHS
Berkooz, G., Holmes, P., Lumley, J.L., 1993. The proper orthogonal decomposition in the
simulation while adopting efficient POD algorithms, which are verified analysis of turbulent flows. Annu. Rev. Fluid Mech. 25 (1), 539–575.
using a case study. The main conclusions are as follows: Boxall, J., Davies, S., Koh, C.A., Sloan, E.D., 2008. Predicting when and where hydrate
plugs form in oil-dominated flowlines. In: Offshore Technology Conference. https://
doi.org/10.4043/19514-MS.
(1) Three levels of hydrate blockage risk are defined based on three Camargo, R., Palermo, T., 2002. Rheological properties of hydrate suspensions in an
different critical hydrate volume fractions for the limit state asphaltenic crude oil. In: 4th International Conference on Gas Hydrates.
equation, namely hydrate formation risk, low risk of hydrate Chaudhari, P., Zerpa, L.E., Sum, A.K., 2018. A correlation to quantify hydrate plugging
risk in oil and gas production pipelines based on hydrate transportability
blockage, and high risk of hydrate blockage. The probability of parameters. J. Nat. Gas Sci. Eng. 58, 152–161. https://doi.org/10.1016/j.
jngse.2018.08.008.

14
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Chen, J., et al., 2015. Insights into the formation mechanism of hydrate plugging in Mckay, M.D., Beckman, R.J., Conover, W.J., 1979. A comparison of three methods for
pipelines. Chem. Eng. Sci. 122, 284–290. https://doi.org/10.1016/j. selecting values of input variables in the analysis of output from a computer code.
ces.2014.09.039. Technometrics 21 (2), 239–245.
Chen, Y., et al., 2018. Progress of influence mechanism of kinetic hydrate inhibitors. Melchers, R.E., Jeffrey, R.J., 2008. Probabilistic models for steel corrosion loss and
Chem. Ind. Eng. Prog. 37 (5), 1726–1743. pitting of marine infrastructure. Reliab. Eng. Syst. Saf. 93 (3), 423–432. https://doi.
Chen, Y., et al., 2020. Kinetic and rheological investigation of cyclopentane hydrate org/10.1016/j.ress.2006.12.006.
formation in waxy water-in-oil emulsions. Fuel 287. https://doi.org/10.1016/j. Minasny, B., McBratney, A.B., 2006. A conditioned Latin hypercube method for sampling
fuel.2020.119568, 119568-119568. in the presence of ancillary information. Comput. Geosci. 32 (9), 1378–1388.
Chen, G.J., Sun, C.Y., Ma, M.L, 2008. Natural Gas Hydrate Science and Technology. https://doi.org/10.1016/j.cageo.2005.12.009.
Chemical Industry Press. Mochizuki, T., Mori, Y.H., 2006. Clathrate-hydrate film growth along water/hydrate-
Cizmas, P.G., Palacios, A., O’Brien, T., Syamlal, M., 2003. Proper-orthogonal former phase boundaries—numerical heat-transfer study. J. Cryst. Growth 290 (2),
decomposition of spatio-temporal patterns in fluidized beds. Chem. Eng. Sci. 58 (19), 642–652. https://doi.org/10.1016/j.jcrysgro.2006.01.036.
4417–4427. https://doi.org/10.1016/S0009-2509(03)00323-3. Mori, Y.H., 2001. Estimating the thickness of hydrate films from their lateral growth
Davies, S.R., et al., 2009. Predicting hydrate-plug formation in a subsea tieback. SPE rates: application of a simplified heat transfer model. J. Cryst. Growth 223 (1),
Prod. Oper. 24 (4), 573–578. https://doi.org/10.2118/115763-PA. 206–212. https://doi.org/10.1016/S0022-0248(01)00614-5.
Dejean, J.P., Averbuch, D., Gainville, M., Doux, F., 2005. Integrating flow assurance into Mu, L., et al., 2014. Experimental and modeling investigation of kinetics of methane gas
risk management of deep offshore field developments. In: Offshore Technology hydrate formation in water-in-oil emulsion. Fluid Phase Equil. 362, 28–34. https://
Conference. Offshore Technology Conference, Houston, Texas, p. 15. https://doi. doi.org/10.1016/j.fluid.2013.08.028.
org/10.4043/17237-MS. Peng, B.Z., et al., 2007. Hydrate film growth on the surface of a gas bubble suspended in
Dejean, J.P., et al., 2007. Famus I: risk-based design of offshore oil and gas production water. J. Phys. Chem. B 111 (43), 12485–12493. https://doi.org/10.1021/
system - management of uncertainties by integrating flow assurance and reliability jp074606m.
aspects into a stochastic Petri nets model. In: Proceedings of the European Safety and Perrin, A., Musa, O.M., Steed, J.W., 2013. The chemistry of low dosage clathrate hydrate
Reliability Conference 2007, ESREL 2007 - Risk, Reliability and Societal Safety, 2 3, inhibitors. Chem. Soc. Rev. 42 (5), 1996–2015. https://doi.org/10.1039/
pp. 2443–2450. c2cs35340g.
Dellecase, E., et al., 2008. Hydrate Plugging or Slurry Flow: Effect of Key Variables. Qin, H., Srivastava, V., Wang, H., Zerpa, L.E., Koh, C.A., 2019. Machine learning models
Di Lorenzo, M., et al., 2014a. Underinhibited hydrate formation and transport to predict gas hydrate plugging risks using flowloop and field data. In: Offshore
investigated using a single-pass gas-dominant flow loop. Energy Fuel. 28 (11), Technology Conference. Offshore Technology Conference, Houston, Texas, p. 12.
7274–7284. https://doi.org/10.1021/ef501609m. https://doi.org/10.4043/29411-MS.
Di Lorenzo, M., et al., 2014b. hydrate formation in gas-dominant systems using a single- Rathinam, M., Petzold, L.R., 2003. A new look at proper orthogonal decomposition.
pass flowloop. Energy Fuel. 28 (5), 3043–3052. https://doi.org/10.1021/ef500361r. SIAM J. Numer. Anal. 41 (5), 1893–1925. https://doi.org/10.1137/
Freer, E.M., Sami Selim, M., Dendy Sloan, E., 2001. Methane hydrate film growth s0036142901389049.
kinetics. Fluid Phase Equil. 185 (1), 65–75. https://doi.org/10.1016/S0378-3812 Sa, J.-H., et al., 2019. Investigating the effectiveness of anti-agglomerants in gas hydrates
(01)00457-5. and ice formation. Fuel 255. https://doi.org/10.1016/j.fuel.2019.115841.
Frostman, L.M., Thieu, V., Crosby, D.L., Downs, H.H., 2003. Low-dosage hydrate Shi, B., Liu, Y., Lv, X.F., Gong, J, 2018. New simulator for gas–hydrate slurry stratified
inhibitors (LDHIs): reducing costs in existing systems and designing for the future. flow based on the hydrate kinetic growth model. J. Energy Resour. Technol. https://
In: International Symposium on Oilfield Chemistry. doi.org/10.1115/1.4040932.
Hammerschmidt, E., 1934. formation of gas hydrates in natural gas transmission lines. Shi, G., Song, S., Shi, B., Gong, J., Chen, D., 2021. A new transient model for hydrate
Ind. Eng. Chem. 26 https://doi.org/10.1021/ie50296a010. slurry flow in oil-dominated flowlines. J. Petrol. Sci. Eng. 196 https://doi.org/
Hasan, S., Khan, F., Kenny, S., 2012. Probability assessment of burst limit state due to 10.1016/j.petrol.2020.108003.
internal corrosion. Int. J. Pres. Ves. Pip. 89, 48–58. https://doi.org/10.1016/j. Shindo, Y., Lund, P.C., Fujioka, Y., Komiyama, H., 1993. Kinetics and mechanism of the
ijpvp.2011.09.005. formation of CO2 hydrate. Int. J. Chem. Kinet. 25 (9), 777–782. https://doi.org/
Holmes, P., 1996. Turbulence, Coherent Structures, Dynamical Systems and Symmetry. 10.1002/kin.550250908.
Cambridge University Press. Singh, A., Suri, A., 2020. A review on gas hydrates and kinetic hydrate inhibitors based
Huo, Z., et al., 2001. Hydrate plug prevention by anti-agglomeration. Chem. Eng. Sci. 56 on acrylamides. J. Nat. Gas Sci. Eng. 83 https://doi.org/10.1016/j.
(17), 4979–4991. https://doi.org/10.1016/s0009-2509(01)00188-9. jngse.2020.103539.
Iman, R.L., Conover, W.J., 1980. Small sample sensitivity analysis techniques for Sirovich, L., 1987. Turbulence and the dynamics of coherent structures. I - coherent
computer models.with an application to risk assessment. Commun. Statist. Theory structures. Q. Appl. Math. 45 (3), 561–590.
Methods 9 (17), 1749–1842. Skovborg, P., Rasmussen, P., 1994. A mass transport limited model for the growth of
Joshi, S.V., et al., 2013. Experimental flowloop investigations of gas hydrate formation in methane and ethane gas hydrates. Chem. Eng. Sci. 49 (8), 1131–1143. https://doi.
high water cut systems. Chem. Eng. Sci. 97, 198–209. https://doi.org/10.1016/j. org/10.1016/0009-2509(94)85085-2.
ces.2013.04.019. Sloan, D., Creek, J., Sum, A.K., 2011. Chapter two - where and how are hydrate plugs
Ke, W., Svartaas, T.M., Chen, D., 2019. A review of gas hydrate nucleation theories and formed? In: Sloan, D., et al. (Eds.), Natural Gas Hydrates in Flow Assurance. Gulf
growth models. J. Nat. Gas Sci. Eng. 61, 169–196. https://doi.org/10.1016/j. Professional Publishing, Boston, pp. 13–36. https://doi.org/10.1016/B978-1-85617-
jngse.2018.10.021. 945-4.00002-9.
Kelland, M.A., 2006. History of the development of low dosage hydrate inhibitors. Sloan, E.D., 2005. A changing hydrate paradigm—from apprehension to avoidance to
Energy Fuel. 20 (3), 825–847. https://doi.org/10.1021/ef050427x. risk management. Fluid Phase Equil. 228–229, 67–74. https://doi.org/10.1016/j.
Kelland, M.A., 2018. A review of kinetic hydrate inhibitors from an environmental fluid.2004.08.009.
perspective. Energy Fuel. 32 (12), 12001–12012. https://doi.org/10.1021/acs. Sloan, E.D., Koh, C.A., 2007. Clathrate Hydrates of Natural Gases, Third Edition. CRC
energyfuels.8b03363. Press, New York.
Kim, J., Noh, Y., Ryu, J., Seo, Y., Chang, D., 2016. Determination of hydrate inhibitor Sohn, Y.h., et al., 2015. Hydrate plug formation risk with varying watercut and inhibitor
injection rate based on the life-cycle cost of the injection facility and mitigating concentrations. Chem. Eng. Sci. 126, 711–718. https://doi.org/10.1016/j.
measures. J. Nat. Gas Sci. Eng. 34, 552–562. https://doi.org/10.1016/j. ces.2015.01.016.
jngse.2016.07.030. Song, G., et al., 2017. Investigation of hydrate plugging in natural gas+diesel oil+water
Lekvam, K., Ruoff, P., 1993. A reaction kinetic mechanism for methane hydrate systems using a high-pressure flow loop. Chem. Eng. Sci. 158, 480–489. https://doi.
formation in liquid water. J. Am. Chem. Soc. 115 (19), 8565–8569. https://doi.org/ org/10.1016/j.ces.2016.10.045.
10.1021/ja00072a007. Srivastava, V., 2018. Quantitative Risk Modeling of Gas Hydrate Bedding Using
Li, S.-X., et al., 2009. A method of probabilistic analysis for steel pipeline with correlated Mechanistic, Statistical, and Artificial Neural Network Frameworks. Colorado School
corrosion defects. Corrosion Sci. 51 (12), 3050–3056. https://doi.org/10.1016/j. of Mines, Golden, Colorado.
corsci.2009.08.033. Turner, D., Boxall, J., Yang, S., Kleehamer, D., Talley, L., 2005. Development of a hydrate
Li, W., et al., 2013. A study of hydrate plug formation in a subsea natural gas pipeline kinetic model and its incorporation into the OLGA2000 transient multiphase flow
using a novel high-pressure flow loop. Petrol. Sci. 10 (1), 97–105. https://doi.org/ simulator. In: Fifth International Conference on Gas Hydrates.
10.1007/s12182-013-0255-8. Turner, D.J., Miller, K.T., Dendy Sloan, E., 2009. Methane hydrate formation and an
Li, Z., et al., 2020. Assembly pin factor parameterization method based on the proper inward growing shell model in water-in-oil dispersions. Chem. Eng. Sci. 64 (18),
orthogonal decomposition. Ann. Nucl. Energy 139. https://doi.org/10.1016/j. 3996–4004. https://doi.org/10.1016/j.ces.2009.05.051.
anucene.2019.107262. Vysniauskas, A., Bishnoi, P.R., 1983. A kinetic study of methane hydrate formation.
Liu, C., Zhang, C., Zhou, C., Mi, Y., Wang, Z., 2020. Effects of the solidification of Chem. Eng. Sci. 38 (7), 1061–1072. https://doi.org/10.1016/0009-2509(83)80027-
capillary bridges on the interaction forces between hydrate particles. Energy Fuel. 34 X.
(4), 4525–4533. https://doi.org/10.1021/acs.energyfuels.0c00463. Vysniauskas, A., Bishnoi, P.R., 1985. Kinetics of ethane hydrate formation. Chem. Eng.
Liu, Y., et al., 2018. Investigation of hydrate agglomeration and plugging mechanism in Sci. 40 (2), 299–303. https://doi.org/10.1016/0009-2509(85)80070-1.
low-wax-content water-in-oil emulsion systems. Energy Fuel. 32 (9), 8986–9000. Wang, Y., Fan, S., Lang, X., 2019a. Reviews of gas hydrate inhibitors in gas-dominant
https://doi.org/10.1021/acs.energyfuels.8b01323. pipelines and application of kinetic hydrate inhibitors in China. Chin. J. Chem. Eng.
Lou, X., et al., 2012. Synthesis of effective kinetic inhibitors for natural gas hydrates. 27 (9), 2118–2132. https://doi.org/10.1016/j.cjche.2019.02.023.
Energy Fuel. 26 (2), 1037–1043. https://doi.org/10.1021/ef201463v. Wang, Y., Koh, C.A., White, J., Patel, Z., Zerpa, L.E., 2019b. Hydrate formation
Lumley, J.L., 1981. Coherent structures in turbulence. In: Meyer, R.E. (Ed.), Transition management simulations with anti-agglomerants and thermodynamic inhibitors in a
and Turbulence. Academic Press, pp. 215–242. https://doi.org/10.1016/B978-0-12- subsea tieback. Fuel 252, 458–468. https://doi.org/10.1016/j.fuel.2019.04.146.
493240-1.50017-X.

15
X. Duan et al. Journal of Natural Gas Science and Engineering 98 (2022) 104345

Wang, Z., et al., 2018a. Flow assurance during deepwater gas well testing: hydrate Yin, Z., Khurana, M., Tan, H.K., Linga, P., 2018. A review of gas hydrate growth kinetic
blockage prediction and prevention. J. Petrol. Sci. Eng. 163, 211–216. https://doi. models. Chem. Eng. J. 342, 9–29. https://doi.org/10.1016/j.cej.2018.01.120.
org/10.1016/j.petrol.2017.12.093. Zerpa, L.E., et al., 2013. Multiphase flow modeling of gas hydrates with a simple
Wang, Z., et al., 2018b. Quantitatively assessing hydrate-blockage development during hydrodynamic slug flow model. Chem. Eng. Sci. 99, 298–304. https://doi.org/
deepwater-gas-well testing. SPE J. 23 (4), 1166–1183. https://doi.org/10.2118/ 10.1016/j.ces.2013.06.016.
181402-pa. Zerpa, L.E., Sloan, E.D., Sum, A.K., Koh, C.A., 2012. Overview of CSMHyK: a transient
Webber, P.A., Nagappayya, S., 2015. Anti-agglomerants for Controlling Gas Hydrates. hydrate formation model. J. Petrol. Sci. Eng. 98–99, 122–129. https://doi.org/
US. 10.1016/j.petrol.2012.08.017.
Xu, C., et al., 2005. Latin hypercube sampling and geostatistical modeling of spatial Zhang, J., et al., 2020a. Real-time estimation and management of hydrate plugging risk
uncertainty in a spatially explicit forest landscape model simulation. Ecol. Model. during deepwater gas well testing. SPE J. 25 (6), 3250–3264. https://doi.org/
185 (2–4), 255–269. https://doi.org/10.1016/j.ecolmodel.2004.12.009. 10.2118/197151-PA.
Yang, D., Le, L.A., Martinez, R.J., Currier, R.P., Spencer, D.F., 2011. Kinetics of CO2 Zhang, J., Wang, Z., Duan, W., Fu, W., Tong, S., 2020b. Real-time estimation and
hydrate formation in a continuous flow reactor. Chem. Eng. J. 172 (1), 144–157. management of hydrate plugging risk during deepwater gas well testing. SPE
https://doi.org/10.1016/j.cej.2011.05.082. Journal, Preprint(Preprint) 1–15. https://doi.org/10.2118/197151-PA.

16

You might also like