Journal of Energy Storage 76 (2024) 109718

Contents lists available at ScienceDirect

Journal of Energy Storage

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

Research papers

Optimal scheduling of hydrogen blended integrated electricity–gas system

considering gas linepack via a sequential second-order cone programming
methodology
Chenghu Wu a , Weiwei Li a , Tong Qian a , Xuehua Xie a , Jian Wang a , Wenhu Tang a ,∗,
Xianfu Gong b
a
School of Electric Power Engineering, South China University of Technology, Guangzhou 510640, China
b
Grid Planning and Research Center, Guangdong Power Grid Corporation, Guangzhou 510080, China

ARTICLE INFO ABSTRACT

Keywords: In the context of increasing global environmental pollution and constant increase of carbon emission, hydrogen
Integrated electricity–gas system production from surplus renewable energy and hydrogen transportation using existing natural gas pipelines
Sequential solution method are effective means to mitigate renewable energy fluctuation, build a decarbonized gas network, and achieve
Blending hydrogen into natural gas pipelines
the goal of ‘‘carbon peak and carbon neutral’’ in China. Existing gas network modeling with hydrogen blended
Gas linepack
is divided into steady-state and dynamic. The steady-state modeling oversimplifies the physical process, while
the dynamic modeling is computationally intensive. Thus, a quasi-steady-state modeling method of a hydrogen
blended integrated electricity–gas system (HBIEGS) considering gas linepack is developed in this paper. The
gas linepack refers to the gas stored in natural gas pipelines due to the compressibility of the gas. As a form
of gas energy storage, linepack can enhance system flexibility by coping with wind power uncertainty. In
the face of nonlinear gas flow equations in HBIEGS, conventional relaxation or approximate methods cannot
provide a sufficiently feasible optimal solution. Therefore, a sequential second-order cone programming (S-
SOCP) method is proposed to solve the developed model. It can effectively strike a balance between solution
speed and the quality of the solution obtained. Finally, the modified IEEE-24-node power system and 12-node
natural gas system are used to validate the effectiveness of the developed model and the proposed method. The
optimization results show that the proposed method improves the computational efficiency by 91% compared
with a general nonlinear solver.

1. Introduction density, cleanliness, efficiency and flexibility [7]. However, the further
application of hydrogen energy has been hindered by the high cost as-
The growth in the use of fossil energy and the increase in carbon sociated with the construction and maintenance of dedicated hydrogen
emissions have prompted governments to take measures to cope with pipelines, which are the traditional large-scale hydrogen transportation
global environmental climate change [1]. In recent years, China has methods [8]. An effective approach to overcome this dilemma is to
continuously increased the installed capacity of renewable energy, transport hydrogen-enriched compressed natural gas (HCNG) which is
which plays a significant role in its proposed ambitious goals of ‘‘carbon composed of a certain proportion of hydrogen blended into the existing
peak and carbon neutrality’’. Green hydrogen injection by making full natural gas network. Recently, many hydrogen blending demonstration
use of its flexibility is beneficial to energy conservation and emission projects have been carried out around the world. The UK initiated its
reduction, which is in line with the international policies Sustainable first hydrogen blending demonstration project, Hydeploy, in 2019. In
Development Goals (SDG) [2,3] and Paris Agreement [4]. With the the first phase of the project, a blend of up to 20% H2 by volume was
increasing penetration rate of renewable energy, the mature electrolysis used to support the natural gas demand of 100 households and 30 uni-
technology is utilized to convert surplus electricity from renewable versities [9]. In Chaoyang, China, a hydrogen blending demonstration
energy into green hydrogen [5,6]. It is a practical pathway to achieve project was conducted in the same year and has been running steadily
energy conservation, emission reduction and address the fluctuations in
for one year with 10% H2 by volume [10].
renewable energy. Hydrogen energy has the advantages of high energy

∗ Corresponding author.
E-mail address: [email protected] (W. Tang).

https://doi.org/10.1016/j.est.2023.109718
Received 14 September 2023; Received in revised form 8 November 2023; Accepted 12 November 2023
Available online 23 November 2023
2352-152X/© 2023 Elsevier Ltd. All rights reserved.
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Recently, there is an increasing amount of research exploring the are convexified into SOC constraints, transforming the original problem
impact of green hydrogen injection on gas interchangeability. In [11, into a more easily solvable second-order cone programming (SOCP)
12], the short-term reliability and resilience of gas interchangeability problem. However, the optimal solution obtained is less feasible for the
is studied. Security indices about ensuring gas interchangeability in original problem because the relaxation technique enlarges the feasi-
integrated electricity and gas systems (IEGS) considering hydrogen ble domain of the original problem. Therefore, a widespread concern
injection are introduced to restrain the gas composition in [13]. These is how to tighten the feasible domain of the relaxation so that the
literature demonstrate the feasibility of hydrogen blended into IEGS. optimal solution obtained approximates that of the original problem
Existing studies on optimal scheduling models for HBIEGS are mainly sufficiently.
classified into steady-state modeling and dynamic modeling. In [14], The sequential solution technique is a practical approach to tighten
the steady-state optimal operation of HBIEGS was modeled with green the relaxation constraints. The feasible region of the relaxation problem
hydrogen injection and gas composition tracking capabilities. The ap- is iteratively tightened through a sequential solving strategy, ultimately
plication of the model can satisfy hydrogen content limit and pipeline approximating the feasible region of the original problem. In [26],
network operation constraints. An optimal scheduling approach for a a sequential linear programming method is proposed, which involves
hydrogen-penetrated integrated energy system (HPIES) was proposed linearizing the original problem and performing sequential iterations.
in [15], whose gas network was modeled using steady-state mod- However, it lacks robustness. A sequential cone programming (SCP)
eling. However, different subsystems have their own time scales. A technique to deal with the optimal power problem in power systems
power system can reach a steady state in a few seconds, while the was proposed in [27], which exhibits an excellent performance. How-
hydraulic processes in a natural gas pipeline network take several ever, it is applied to power systems only. A sequential SOCP method
minutes. Therefore, these simplified steady-state models will lead to was proposed in [13]. It effectively tightens the SOCP constraints and
a significant deviation of the operational results from actual situation. improves the feasibility of the solution. However, it only deals with
Some research focuses on the dynamic modeling of natural gas systems. steady-state models of IEGS.
For example, discrete partial differential equations were employed to In summary, steady-state or dynamic modeling has been developed
model the dynamic pipeline network of natural gas in [16]. However, for HBIEGS. Based on Table 1, there is a lack of research considering
the dynamic model results in significant computational complexity, gas linepack modeling in HBIEGS, which aims to strike a balance
making it impractical for complex systems. between oversimplified physical processes in steady-state models and
Fluid dynamics in natural gas pipeline networks rely on linepack, the computational burden of dynamic models. Besides, existing solution
which refers to the amount of natural gas that can be ‘‘stored’’ within methods for optimization problems in HBIEGS either suffer from insuf-
the gas pipelines [17]. Due to natural gas pipelines’ significant energy ficient feasibility of the obtained solutions or high computational cost.
storage potential, power-to-hydrogen (P2H) technology can convert There is a requirement for faster and more efficient solution methods
surplus renewable energy into hydrogen for storage within natural gas tailored for solving the optimization of quasi-steady-state HBIEGS. In
pipelines. It enhances the energy system’s flexibility and allows for order to fill this gap, a new modeling and solution method is proposed
better integration of volatile renewable energy sources. The modeling in this research, whose contributions are summarized as follows.
approach that takes into account linepack is known as quasi-steady-
• Green hydrogen is blended into the natural gas pipelines whose
state modeling. There is some research on quasi-steady-state modeling
impact on wind power consumption and system operation costs
of natural gas systems or IEGS to investigate the impact of gas linepack
are investigated.
on the integration of renewable energy and system flexibility. The
• A quasi-steady-state model for hydrogen blended integrated
IEGS coordinated optimal scheduling model considering gas linepack
electricity–gas system (HBIEGS) is proposed, where gas linepack
was developed in [18] which has revealed the contribution of gas
are considered. The bidirectional gas flow and impact of the gas
linepack to accommodate unsteady renewable energy. The day-ahead
linepack are considered in this model, which enables a more
optimal scheduling model for integrated energy system (IES) consider-
accurate characterization of hydrogen injection into natural gas
ing linepack was constructed in [19] whose results validate the role of
pipelines.
linepack in improving system flexibility.
• In order to cope with the complexity of the proposed model,
Although the quasi-steady-state model has made considerable sim-
a tailored sequential second-order cone programming (S-SOCP)
plifications compared to the dynamic model, the presence of nonlinear
method is introduced. Compared with previous studies, (1) the
gas flow equations still poses a challenge in solving the optimiza-
proposed method reformulates the original, sophisticated non-
tion problem. Several methods are currently available to solve this
linear problem into a more tractable second-order form. Thus,
problem, including intelligent algorithms, piecewise linear approxima-
the computation efficiency can be improved significantly. (2) a
tion method, relaxation methods, interior point method, and Newton’s
vary penalty factor is used in the S-SOCP method to balance the
method. The interior point and Newton’s methods may depend on the
computation speed and feasibility of the constraints.
initial point. Besides, considering the unknown gas flow direction in
natural gas systems poses a mixed-integer problem, which is challeng- The rest of this paper is organized as follows. The optimal schedul-
ing to handle with the interior point method. Therefore, the assumption ing model of HBIEGS is developed in Section 2. In Section 3, the
about fixed gas flow direction has to be made. However, it is unsuitable nonlinear constraints of the HBIEGS model are reformulated into SOCP
for all IEGS systems, especially for gas transmission networks [20]. The constraints, and the S-SOCP method is proposed to solve the refor-
segmented linear approximation method is standard for dealing with mulated model. In Section 4, case studies are carried out to show
nonlinear functions. An incremental segmented linearization method the effectiveness of the developed model and the proposed solution
is developed in [21] by reformulating nonlinear gas flow equations. method. Finally, in Section 5, conclusions are drawn.
However, a large number of linear segments had to be introduced in
order to achieve a high-precision approximation. It would introduce 2. Quasi-steady-state modeling of hydrogen blended integrated
many binary variables to the problem, thus reducing the computational electricity-gas system
efficiency. Classical or improved intelligent algorithms, such as particle
swarm algorithms, have been applied in [22,23] to solve the nonlin- 2.1. System structure
earity of the gas flow equation. However, the intelligent algorithms are
time-consuming and easy to fall into local optima. Besides, the second- The system architecture of HBIEGS is depicted in Fig. 1. The elec-
order cone (SOC) relaxation technique is widely developed to cope with trolysis device with hydrogen storage module (HSM) and gas turbines
nonlinear gas flow equations. In [24,25], quadratic gas flow equations are the electrical system’s coupling components, enabling bidirectional

2
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Table 1
Comparison between the developed model and solving method with existing literature.
Reference Modeling Flow direction Solving method
[13] Steady state bi-direction Sequential SOCP
[15] Steady state bi-direction Improved piecewise McCormick relaxation
[16] Dynamic bi-direction IPOPT solver
[18] Quasi-steady-state bi-direction SOCP relaxation
[19,21] Steady state bi-direction Piecewise linear approximation
[22,23] Steady state bi-direction Intelligent algorithm
[24,25] Steady state fixed direction SOCP relaxation
[26] Steady state bi-direction Sequential linear programming
[27] Steady state bi-direction Sequential cone programming
This paper Quasi-steady-state bi-direction Sequential SOCP

Fig. 1. Architecture of the studied HBIEGS.

energy conversion between the two systems. Hydrogen generated from • The gases considered are chemically inert, indicating that there
surplus wind power is injected into the natural gas pipelines. It is are no chemical reactions occurring between them.
mixed with natural gas produced from gas wells and transported as
HCNG through natural gas pipelines. The HCNG mixture can be used According to [28], the natural gas follow-up hydrogen mixing tech-
to supply energy to both gas loads and gas turbines. Gas turbines, coal- nology is adopted in some demonstration projects in practice. Consid-
fired generators, and wind farms provide energy to power systems. ering system safety, a fixed hydrogen blending ratio is adopted in this
Therefore, wind power is not only directly involved in supplying energy research.
to the power system but also indirectly contributes to the power system
by generating hydrogen and then providing energy to gas turbines. 2.2.1. Power gird constraints
DC power flow is adopted in this paper. The Eqs. (1a) is the
2.2. Hydrogen blended integrated electricity-gas system modeling node power balance constraint. The DC flow constraint is denoted
by Eq. (1b). The Eq. (1c) is the phase constraint of the reference
HBIEGS mainly consists of a power grid, natural gas system, and node. The Eq. (1d)–(1e) denote output limit constraint of the coal-
coupling components. The coupling components include electrolysis fired generators and gas turbines, respectively. The Eq. (1f) denote the
devices with HSM and gas turbines. The model assumptions are as constraint that the wind power output does not exceed the maximum
follows: available power generation capacity.
∑ ∑ ∑ ∑ ∑
• Considering that the dynamic process in the power system is rapid c
𝑃𝑘,𝑡 + u
𝑃𝑟,𝑡 + w
𝑃𝑠,𝑡 + 𝑃𝑗𝑖,𝑡 − 𝑃𝑖𝑗,𝑡
compared to the natural gas system, the power system modeling 𝑘∈ 𝑟∈ 𝑠∈ 𝑖𝑗∈𝛼𝑖+ 𝑗𝑖∈𝛼𝑖−
𝑖 𝑖 𝑖
is steady state. ∑
e P2H
= 𝐷𝑖,𝑡 + 𝑃𝑝,𝑡 , ∀𝑖 ∈ , ∀𝑡 ∈  (1a)
• Hydrogen is injected into natural gas pipelines with a fixed hy-
𝑝∈
𝑖
drogen blending ratio.
• Hydrogen is pressurized to natural gas pressure before being
injected into natural gas pipelines. 𝑃𝑖𝑗,𝑡 = (𝜃𝑖,𝑡 − 𝜃𝑗,𝑡 )∕𝑋𝑖𝑗 , ∀𝑖𝑗 ∈ e , ∀𝑡 ∈  (1b)

3
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

According to [29], the Weymouth equation can be used to describe

𝜃𝑅𝑒𝑓 ,𝑡 = 0, ∀𝑡 ∈  (1c) the relationship between the mass flow rate of mixed gas in a pipeline
and the gas pressure at both ends of the pipeline, which is constrained
min
𝑃c,𝑘 c
≤ 𝑃𝑘,𝑡 max
≤ 𝑃c,𝑘 , ∀𝑘 ∈ , ∀𝑡 ∈  (1d) by (4).

𝑝𝑟2𝑚,𝑡 − 𝑝𝑟2𝑛,𝑡 = 𝐾𝑚𝑛 𝑄𝑚𝑛,𝑡 ||𝑄𝑚𝑛,𝑡 || , ∀𝑚𝑛 ∈ g ⧵ c , ∀𝑡 ∈  (4a)

min u max
𝑃u,𝑟 ≤ 𝑃𝑟,𝑡 ≤ 𝑃u,𝑟 , ∀𝑟 ∈ , ∀𝑡 ∈  (1e)
𝜆𝑚𝑛 𝐿𝑚𝑛
𝐾𝑚𝑛 = 𝑉 (𝛾) (4b)
w 𝑎𝑣𝑎 𝐷𝑚𝑛 𝐴2𝑚𝑛
0 ≤ 𝑃𝑠,𝑡 ≤ 𝑃w,𝑠,𝑡 , ∀𝑠 ∈ , ∀𝑡 ∈  (1f)
where 𝜆𝑚𝑛 , 𝐿𝑚𝑛 , 𝐷𝑚𝑛 and 𝐴𝑚𝑛 denote the friction coefficient, length,
where 𝑃𝑘,𝑡c denotes the active power output of the coal-fired generator
diameter, and cross-sectional area of pipeline 𝑚𝑛, respectively. 𝑝𝑟𝑚,𝑡
𝑘 at time 𝑡. 𝑃𝑟,𝑡u and 𝑃 w denote the active power output of the gas
𝑠,𝑡 and 𝑝𝑟𝑛,𝑡 denote the gas pressure at the gas node 𝑚 and gas node 𝑛 at
turbine 𝑟 and the wind farm 𝑠 at time 𝑡, respectively. 𝑃𝑗𝑖,𝑡 and 𝑃𝑖𝑗,𝑡 time 𝑡, respectively. 𝑄𝑚𝑛,𝑡 denote the average mass flow rate of mixed
denote the active power transmitted on a transmission line with node gas flowing through pipeline 𝑚𝑛 at time 𝑡. g denotes the set of gas
𝑖 as the ending and starting point at time 𝑡, respectively. 𝐷𝑖,𝑡 e denotes
network pipelines. c denotes the set of compressor pipelines. 𝑉 (𝛾) is
P2H denotes the active power
the electric load in the node 𝑖 at time 𝑡. 𝑃𝑝,𝑡 constrained by (5).
consumed by the P2H unit 𝑘 at time 𝑡.  𝑖 denotes the set of coal-fired
generators connected to node 𝑖.  𝑉 (𝛾) ≜ 𝛾 ⋅ 𝑎2ℎ + (1 − 𝛾) ⋅ 𝑎2𝑛𝑔 (5)
𝑖 denotes the set of gas-fired turbines
connected to node 𝑖. 𝑖 denotes the set of wind farms connected to where 𝑎ℎ and 𝑎𝑛𝑔 represent the velocity of sound for hydrogen and
node 𝑖.  𝑖 denotes the set of P2H units connected to node 𝑖.  denotes natural gas, respectively, in ideal conditions. They are 1092 m/s and
the set of electric nodes.  denotes the set of 24 h in this paper. 𝛼𝑖+ and 372 m/s, respectively.
𝛼𝑖− denote the sets of transmission lines with node 𝑖 as the ending and Considering the flow direction of the gas mixture, a binary variable
starting point, respectively. In Eq. (1b), 𝜃𝑖,𝑡 and 𝜃𝑗,𝑡 denote the voltage indicating the direction of gas flow is introduced, where 𝑦𝑚𝑛,𝑡 = 1
phase of node 𝑖 and node 𝑗, respectively. 𝑋𝑖𝑗 denotes the reactance of indicates that the gas flows from gas node 𝑚 to 𝑛, and 𝑦𝑚𝑛,𝑡 = 0 indicates
transmission line 𝑖𝑗. e denotes the set of transmission lines. In Eq. (1c), otherwise. 𝑄𝑚𝑛,𝑡 can be expressed as the difference of two non-negative
𝜃𝑅𝑒𝑓 ,𝑡 denotes the voltage phase of the reference node. In Eq. (1d), variables which is constrained by (6a)–(6e).
min and 𝑃 max denotes the lower and upper limit of active output of
𝑃c,𝑘 c,𝑘
the coal-fired generator 𝑘, respectively.  denotes the set of coal-fired 𝑄𝑚𝑛,𝑡 = 𝑄+ −
𝑚𝑛,𝑡 − 𝑄𝑚𝑛,𝑡 , ∀𝑚𝑛 ∈ g , ∀𝑡 ∈  (6a)
generators. In Eq. (1e), 𝑃u,𝑟min and 𝑃 max denote the lower and upper
u,𝑟
limit of active output of the gas turbine 𝑟, respectively.  denotes the 0 ≤ 𝑄+
𝑚𝑛,𝑡 ≤ 𝑀𝑦𝑚𝑛,𝑡 , ∀𝑚𝑛 ∈ g , ∀𝑡 ∈  (6b)
𝑎𝑣𝑎 denotes the maximum available
set of gas turbines. In Eq. (1f), 𝑃w,𝑠,𝑡
generation output of the wind farm 𝑠 at time 𝑡.  denotes the set of 0 ≤ 𝑄−
𝑚𝑛,𝑡 ≤ 𝑀(1 − 𝑦𝑚𝑛,𝑡 ), ∀𝑚𝑛 ∈ g , ∀𝑡 ∈  (6c)
wind farms.
( )
𝑄+
𝑚𝑛,𝑡 = 𝑄𝑖𝑛 𝑜𝑢𝑡
𝑚𝑛,𝑡 + 𝑄𝑚𝑛,𝑡 ∕2, ∀𝑚𝑛 ∈ g , ∀𝑡 ∈  (6d)
2.2.2. Natural gas system constraints
( )
The mass conservation equation for a homogeneous mixture of
𝑄− 𝑖𝑛 𝑜𝑢𝑡
𝑚𝑛,𝑡 = 𝑄𝑛𝑚,𝑡 + 𝑄𝑛𝑚,𝑡 ∕2, ∀𝑚𝑛 ∈ g , ∀𝑡 ∈  (6e)
natural gas and hydrogen at a node can be constrained by (2).
∑ g ∑ pipe ∑ ∑ where 𝑀 is a sufficiently large number, it is taken as 1 × 104 in this
𝑄𝑓 ,𝑡 + 𝑄𝑙,𝑡 + 𝑄𝑜𝑢𝑡
𝑛𝑚,𝑡 − 𝑄𝑖𝑛
+
𝑚𝑛,𝑡
− paper. 𝑄+ −
𝑚𝑛,𝑡 and 𝑄𝑚𝑛,𝑡 denote the forward average flow and reverse
𝑓 ∈
𝑚 𝑙∈
𝑚 𝑛𝑚∈𝛽𝑚 𝑚𝑛∈𝛽𝑚
∑ average flow of pipe 𝑚𝑛, respectively. 𝑄𝑖𝑛 𝑜𝑢𝑡
𝑚𝑛,𝑡 and 𝑄𝑚𝑛,𝑡 denote the
g
= 𝐷𝑚,𝑡 + 𝑄u𝑟,𝑡 , ∀𝑚 ∈ , ∀𝑡 ∈  (2) forward inlet flow and forward outlet flow of pipe 𝑚𝑛, respectively.
𝑟∈ 𝑄𝑖𝑛 𝑜𝑢𝑡
𝑚 𝑛𝑚,𝑡 and 𝑄𝑛𝑚,𝑡 denote the reverse inlet flow and reverse outlet flow of
g pipe 𝑚𝑛, respectively.
where 𝑄𝑓 ,𝑡 denotes the output mass flow rate of the gas source 𝑓 at
pipe Due to gas compressibility, the flow rate at the beginning and end of
time 𝑡. 𝑄𝑙,𝑡 denotes the mass flow rate of hydrogen blended into the a gas pipeline is different. The difference between the flow rate at the
pipeline from the hydrogen injection point 𝑙 at time 𝑡. 𝑄𝑜𝑢𝑡 𝑛𝑚,𝑡 denotes beginning and end of a gas pipeline is stored in the pipeline as linepack
the mass flow rate of the mixed gas at the outlet of the pipeline nm storage, which denotes the amount of gas stored in the pipeline. The
with the node 𝑚 as the ending point at time 𝑡. 𝑄𝑖𝑛 𝑚𝑛,𝑡 denotes the mass formula for calculating linepack storage of mixed gas and the mass
flow rate of the mixed gas at the inlet of the pipeline 𝑚𝑛 with the node balance equation can be found in Eqs. (7a)–(7c), and its derivation can
g
𝑚 as the starting point at time 𝑡. 𝐷𝑚,𝑡 denotes the gas demand at the be found in Appendix.
u
gas node 𝑚 at time 𝑡. 𝑄𝑟,𝑡 denotes the gas consumption of the coal-fired
𝐴 𝑝𝑟𝑚,𝑡 + 𝑝𝑟𝑛,𝑡
unit ℎ connected to the gas node 𝑚 at time 𝑡.  denotes the set of gas ℎ𝑚𝑛,𝑡 = 𝑚𝑛 ⋅ , ∀𝑚𝑛 ∈ g ⧵ c , ∀𝑡 ∈  (7a)
𝑉 (𝛾) 2
nodes. 𝑚 denotes the set of gas sources connected to the gas node 𝑚
while  denotes the set of gas sources.  𝑚 denotes the set of hydrogen ℎ𝑚𝑛,𝑡 −ℎ𝑚𝑛,𝑡−1 = 𝑄𝑖𝑛 𝑜𝑢𝑡 𝑖𝑛 𝑜𝑢𝑡
𝑚𝑛,𝑡 −𝑄𝑚𝑛,𝑡 +𝑄𝑛𝑚,𝑡 −𝑄𝑛𝑚,𝑡 , ∀𝑚𝑛 ∈ g ⧵c , ∀𝑡 ∈  ⧵{𝑡 = 1}
injection nodes connected to the gas node 𝑚 while  denotes the set of
hydrogen injection nodes.  𝑚 denotes the set of gas turbines connected (7b)
to the gas node 𝑚. 𝛽𝑚+ and 𝛽𝑚− denote the set of gas network pipelines
with the gas node 𝑚 as the ending and starting point, respectively. Since ℎ𝑚𝑛,𝑡 − 𝐻𝑚𝑛,0 = 𝑄𝑖𝑛 𝑜𝑢𝑡 𝑖𝑛 𝑜𝑢𝑡
𝑚𝑛,𝑡 − 𝑄𝑚𝑛,𝑡 + 𝑄𝑛𝑚,𝑡 − 𝑄𝑛𝑚,𝑡 , ∀𝑚𝑛 ∈ g ⧵ c , 𝑡 = 1 (7c)
the calorific value of HCNG is different from that of natural gas, the gas
demand flow rate is modified by Eq. (3). where ℎ𝑚𝑛,𝑡 denotes the mass of linepack storage of the pipeline 𝑚𝑛 at
time 𝑡. 𝐻𝑚𝑛,0 denotes the mass of initial linepack storage of the pipeline
g NG
𝐷𝑚,𝑡 (𝛾𝑅H2 + (1 − 𝛾)𝑅NG ) = 𝐷𝑚,𝑡 𝑅NG (3) 𝑚𝑛.
In order to prevent the linepack storage from being depleted, the
where 𝑅H2 and 𝑅NG denote the calorific value of the hydrogen and
safety limit satisfied by the linepack storage of the gas mixture is
natural gas, respectively. They are 141.8 MJ/kg and 44.2 MJ/kg,
NG denotes the expressed in Eq. (8).
respectively. 𝛾 denotes the concentration of hydrogen. 𝐷𝑚,𝑡
| |
gas load flow rate when the working medium is natural gas. ℎ𝑚𝑛,𝑡 ≥ 𝐻𝑚𝑛,0 , ∀𝑚𝑛 ∈ g ⧵ 𝐿c , 𝑡 = | | (8)
| |

4
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

The compressor’s operation in a gas network is required to meet

max
compression ratio limits. The gas flow direction in compressors is 0 ≤ 𝑄𝑖𝑛
𝑝,𝑡 ≤ 𝑏𝑝,𝑡 ⋅ 𝑄𝑝,𝑖𝑛 , ∀𝑝 ∈ , ∀𝑡 ∈  (14c)
consistent with that of increasing gas pressure. Assuming that there is
( )
no gas loss in the compressor and that the concentration of hydrogen in 0 ≤ 𝑄𝑜𝑢𝑡 max
𝑝,𝑡 ≤ 1 − 𝑏𝑝,𝑡 ⋅ 𝑄𝑝,𝑜𝑢𝑡 , ∀𝑝 ∈ , ∀𝑡 ∈  (14d)
the pipeline is equal everywhere due to the short compressor branch.
The constraints are shown in (9a)–(9c).
𝑉𝑝,𝑡 = 𝑉𝑝,𝑡−1 + 𝑄𝑖𝑛 𝑜𝑢𝑡
𝑝,𝑡 − 𝑄𝑝,𝑡 , ∀𝑝 ∈ , ∀𝑡 ∈  (14e)
max
𝑝𝑟𝑚,𝑡 ≤ 𝑝𝑟𝑛,𝑡 ≤ 𝛤𝑚𝑛 ⋅ 𝑝𝑟𝑛,𝑡 , ∀𝑚𝑛 ∈ c (9a)
𝑉𝑝,| | = 𝑉𝑝,0 (14f)
𝑦𝑚𝑛,𝑡 = 1, ∀𝑚𝑛 ∈ c , ∀𝑡 ∈  (9b)
Hydrogen produced by P2H units is supplied to natural gas pipelines
in priority, which are shown in (14a)–(14d). The rest of the hydrogen is
𝑄𝑖𝑛
𝑚𝑛,𝑡 = 𝑄𝑜𝑢𝑡
𝑛𝑚,𝑡 , ∀𝑚𝑛 ∈ c , ∀𝑡 ∈  (9c)
stored in HSMs. If the wind power does not produce enough hydrogen
where max
denotes the maximum compression ratio of the compressor
𝛤𝑚𝑛 to inject into natural gas pipelines, the shortfall is supplied by HSMs.
branch 𝑚𝑛. Eq. (14e) is the mass conservation constraint of HSMs. At the end of a
The gas network node pressure should meet the corresponding scheduling cycle, the hydrogen in the HSM should return to its initial
upper and lower limits constraints shown in (10). state to ensure scheduling for the next cycle, shown in (14f). 𝑄𝑝,𝑡 is
the portion of the hydrogen produced in the P2H unit 𝑝 that is injected
𝑝𝑟min max
𝑚 ≤ 𝑝𝑟𝑚,𝑡 ≤ 𝑝𝑟𝑚 , ∀𝑚 ∈ , ∀𝑡 ∈  (10)
into the HSM. 𝑄𝑖𝑛𝑝,𝑡 is the mass flow rate injected into the HSM from the
where 𝑝𝑟min
and 𝑝𝑟max
denote the minimum and maximum gas pressure hydrogen produced by the P2H unit 𝑝. 𝑄𝑜𝑢𝑡 𝑝,𝑡 is the mass flow rate of the
𝑚 𝑚
allowed at the gas network node 𝑚, respectively. HSM 𝑝 injected into the natural gas pipeline. 𝑉𝑝,𝑡 denotes the capacity
The flow in a pipeline is required to meet a specific safety limit, of the HSM 𝑝 at time 𝑡.
shown in (11).
2.2.4. Objective function
𝑄min max
𝑚𝑛 ≤ 𝑄𝑚𝑛,𝑡 ≤ 𝑄𝑚𝑛 , ∀𝑚𝑛 ∈ g , ∀𝑡 ∈  (11)
The objective function (15) is to minimize the total system cost,
where 𝑄min max
𝑚𝑛 and 𝑄𝑚𝑛 denote the minimum and maximum gas flow rate which includes the power purchase cost of coal-fired generators, the gas
allowed to flow through the pipe 𝑚𝑛, respectively. purchase cost of natural gas sources, and the cost of wind abandonment
At the hydrogen injection point in the gas network, the injection penalty.
of natural gas and hydrogen is subject to the constraint of hydrogen ∑∑ ∑∑ g g ∑∑
injection concentration, shown in (12). min 𝑓 = 𝐶𝑘c 𝑃𝑘,𝑡
c
+ 𝐶𝑟 𝑄𝑟,𝑡 + 𝐶𝑠w (𝑃w,𝑠,𝑡
𝑎𝑣𝑎 w
− 𝑃𝑠,𝑡 ) (15)
𝑡∈ 𝑘∈ 𝑡∈ 𝑟∈ 𝑡∈ 𝑠∈
g pipe
𝑄𝑚,𝑡 ⋅ 𝛾 = 𝑄𝑚,𝑡 ⋅ (1 − 𝛾) , ∀𝑚 ∈ , ∀𝑡 ∈  (12) where 𝐶𝑘c ,
g
and
𝐶𝑟 𝐶𝑠w
denote the unit cost of purchased power for
Combining Eq. (12), Eq. (2) can be derived into two equations: one the coal-fired generator 𝑘, the unit cost of purchased gas for the gas
for the mass flow conservation equations of natural gas and another for source 𝑟 and unit abandonment penalty cost of wind farm 𝑠. Moreover,
the mass flow conservation equations of hydrogen. 𝑓 denotes the total cost.
For the sake of convenience, the original model is referred to as
2.2.3. Coupled equipment constraints Model 1.
The coupled equipment consists of gas turbines and P2H devices.
𝐌𝐨𝐝𝐞𝐥 𝟏 ∶ min 𝑓 𝑠.𝑡. (1a) − (14f) (16)
The fuel for the gas turbine is hydrogen blended natural gas rather than
natural gas. Therefore, the constraint on the relationship between gas
3. Model reformulation and solution
consumption and active power output of the gas turbine is modified
to (13a). The constraint on the relationship between the active power
consumption of the P2H device and hydrogen production power is 3.1. MISOCP Reformulation of the mixed gas flow equation
shown in (13b). The constraint on hydrogen production power of the
P2H device is shown in (13c). Model 1 is nonlinear and non-convex. The nonlinearity of the model
( ) mainly comes from the mixed gas flow equation denoted by Eq. (4a).
𝑄u𝑟,𝑡 𝛾𝑅H2 + (1 − 𝛾) 𝑅NG = 𝜙u𝑟 𝑃𝑟,𝑡
u
𝑅NG , ∀𝑟 ∈ , ∀𝑡 ∈  (13a) Therefore, it is necessary to reformulate Eq. (4a). Auxiliary variables
𝛷𝑚𝑛 and 𝑝𝑟𝑠𝑚 are introduced. Then, Eq. (4b) is reformulated as
𝑄P2H P2H P2H
𝑝,𝑡 = 𝜙𝑝 𝑃𝑝,𝑡 , ∀𝑝 ∈ , ∀𝑡 ∈  (13b)
𝛷𝑚𝑛 = 𝐾𝑚𝑛 𝑄2𝑚𝑛 (17a)
min P2H max
𝑃P2H,𝑝 ≤ 𝑃𝑝,𝑡 ≤ 𝑃P2H,𝑝 , ∀𝑝 ∈ , ∀𝑡 ∈  (13c)
𝑝𝑟𝑠𝑚 = 𝑝𝑟2𝑚 (17b)
where 𝜙u𝑟 and 𝜙P2H
𝑝 denote the conversion efficiency coefficients of the
min
gas turbine 𝑟 and the P2H device 𝑝, respectively. 𝑃P2H,𝑝 max
and 𝑃P2H,𝑝
𝛷𝑚𝑛 = |𝑝𝑟𝑠𝑚 − 𝑝𝑟𝑠𝑛 | (17c)
denote the lower and upper limit of the hydrogen production power
of the P2H device 𝑝, respectively. Eqs. (17a)–(17b) are equivalent to convex constraints
Hydrogen generated by wind power can be equipped with a certain
capacity of HSM. Since the HSM can only be charged or discharged 𝛷𝑚𝑛 ≥ 𝐾𝑚𝑛 𝑄2𝑚𝑛 (18a)
at a specific moment, it is necessary to introduce a binary variable
that characterizes the state of charging and discharging, where 𝑏𝑝,𝑡 = 1 𝑝𝑟𝑠𝑚 ≥ 𝑝𝑟2𝑚 (18b)
indicates that the HSM is charged and 𝑏𝑝,𝑡 = 1 indicates that the HSM
is discharged. The constraints of HSM are shown in (14a)–(14f). and concave constraints

𝑄P2H 𝑖𝑛
𝑝,𝑡 = 𝑄𝑝,𝑡 + 𝑄𝑝,𝑡 , ∀𝑝 ∈ , ∀𝑡 ∈  (14a) 𝛷𝑚𝑛 ≤ 𝐾𝑚𝑛 𝑄2𝑚𝑛 (19a)

pipe 
𝑄𝑙,𝑡 = 𝑄𝑝,𝑡 + 𝑄𝑜𝑢𝑡
𝑝,𝑡 , ∀𝑝 ∈ , ∀𝑡 ∈  , 𝑙 = 𝑙 (14b) 𝑝𝑟𝑠𝑚 ≤ 𝑝𝑟2𝑚 (19b)

5
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Fig. 2. Flowchart of the proposed S-SOCP algorithm.

The convex constraints are left only to convexify the nonlinear 3.2. Solution method
Eq. (17a) and (17b). Eq. (17c) is transformed into equivalent forms
(20a)–(20d). The nonlinearity of the model arises mainly from the Weymouth
equation for the gas mixture expressed in Eqs. (4a). The S-SOCP tech-
𝛷𝑚𝑛 ≥ 𝑝𝑟𝑠𝑛 − 𝑝𝑟𝑠𝑚 + (2𝑦𝑚𝑛 )(𝑝𝑟𝑠,min − 𝑝𝑟𝑠,max ) (20a)
𝑚 𝑛 nique is adopted to tighten the constraints (18) in Model 2. Constraints
(19a)–(19b) are linearized based on Taylor expansion which makes
𝛷𝑚𝑛 ≥ 𝑝𝑟𝑠𝑚 − 𝑝𝑟𝑠𝑛 + (2𝑦𝑚𝑛 − 2)(𝑝𝑟𝑠,max
𝑚 − 𝑝𝑟𝑠,min
𝑛 ) (20b) constraints (18) tighter.
(𝜈−1)
𝛷𝑚𝑛 ≤ 𝐾𝑚𝑛 (2𝑄𝑚𝑛 𝑄𝑚𝑛 − (𝑄(𝜈−1) 2
𝑚𝑛 ) ) + 𝛿𝑚𝑛 (24a)
𝛷𝑚𝑛 ≤ 𝑝𝑟𝑠𝑛 − 𝑝𝑟𝑠𝑚 + (2𝑦𝑚𝑛 )(𝑝𝑟𝑠,max
𝑚 − 𝑝𝑟𝑠,min
𝑛 ) (20c)
𝑝𝑟𝑠𝑚 ≤ 2𝑝𝑟𝑠,(𝜈−1)
𝑚 𝑝𝑟𝑠𝑚 − (𝑝𝑟𝑠,(𝜈−1)
𝑚 )2 + 𝜎𝑚 (24b)
𝛷𝑚𝑛 ≤ 𝑝𝑟𝑠𝑚 − 𝑝𝑟𝑠𝑛 + (2𝑦𝑚𝑛 − 2)(𝑝𝑟𝑠,min
𝑚 − 𝑝𝑟𝑠,max
𝑛 ) (20d)
where 𝛿𝑚𝑛 and 𝜎𝑚 denote the slack variables introduced which are non-
(𝜈−1)
Eq. (21) is the range of auxiliary variables. negative strictly. 𝑄𝑚𝑛 and 𝑝𝑟𝑠,(𝜈−1)
𝑚 denote the optimization value of
𝑄𝑚𝑛 and 𝑝𝑟𝑠𝑚 in the last iteration, respectively. The convergence proof
𝑝𝑟𝑠,min
𝑚 ≤ 𝑝𝑟𝑠𝑚 ≤ 𝑝𝑟𝑠,max
𝑚 (21) of the S-SOCP algorithm can refer to [13]. The main procedures of the
S-SOCP algorithm are shown in Algorithm 1 and its flowchart is shown
Eqs. (22a)–(22b) donate the constraint of the difference of pressure.
in Fig. 2.
(1 − 𝑦𝑚𝑛 )(𝑝𝑟min max max min
𝑚 − 𝑝𝑟𝑛 ) ≤ 𝑝𝑟𝑚 − 𝑝𝑟𝑛 ≤ 𝑦𝑚𝑛 (𝑝𝑟𝑚 − 𝑝𝑟𝑛 ) (22a)
4. Case study
(1 − 𝑦𝑚𝑛 )(𝑝𝑟𝑠,min
𝑚 − 𝑝𝑟𝑠,max
𝑛 ) ≤ 𝑝𝑟𝑠𝑚 − 𝑝𝑟𝑠𝑛 ≤ 𝑦𝑚𝑛 (𝑝𝑟𝑠,max
𝑚 − 𝑝𝑟𝑠,min
𝑛 ) (22b)
The modified IEEE-24 node power system and 12-node natural gas
where 𝑝𝑟𝑠,min
𝑚 and 𝑝𝑟𝑠,max
denote the lower bound and upper bound
𝑚 system are used to validate the effectiveness of the developed model
of the auxiliary variable 𝑝𝑟𝑠𝑚 , respectively. Then, Model 1 can be and solution method. The detailed topology is shown in Fig. 3. As
reformulated as Model 2. depicted in Fig. 3, the 24-node power system consists of 34 transmission
lines and five coal-fired generators with a total installed capacity of
𝐌𝐨𝐝𝐞𝐥 𝟐 ∶ min 𝑓 𝑠.𝑡. (1a) − (3), (4b) − (14f), (18a) − (18b), (20a) − (22b) 2.05 GW. Besides, there are seven gas turbines with a total installed ca-
(23) pacity of 1.03 GW and three wind farms with a total installed capacity
of 1.5 GW. The 12-node natural gas system comprises three gas sources,
Due to the relaxation of constraints in Eq. (18) not always being 12 pipelines, and a loop network. Three P2H units equipped with HSM
exact, a sequential solution procedure is introduced to tighten the are installed, and each is connected to a wind farm and a gas source
constraints. It allows for violations of the constraints of the original directly to avoid gas transmission blockage. The specific information of
problem in the early iterations of the algorithm. As the iterations coal-fired generators, gas turbines, gas sources and P2H are depicted
continue, the relaxed feasible region gradually approaches that of the in Tables 2–5. The system’s 24 h power demand, gas demand, and
original problem by applying penalties to constraints violated. maximum available wind power are depicted in Fig. 4. Besides, the

6
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Fig. 3. Structure of HBIEGS with P2H units.

Algorithm 1: The S-SOCP algorithm for optimization of the fraction of 20%. According to Ref. [30], 25 vol% of hydrogen could be
HBIEGS added to the natural gas pipeline network. While in [31], the highest
Input: input the growth factors of penalty 𝜅𝛼 and 𝜅𝛽 , the upper hydrogen blending ratio in urban natural gas systems could reach up
bound of penalty factors 𝛼 max and 𝛽 max , convergence to 40 vol%. Research on the upper limit of the hydrogen blending ratio
condition 𝜀𝑓 . is ongoing, and a consensus has yet to be reached. Considering existing
Output: optimal value of continuous variables 𝒙∗ and discrete literature, a 30 vol% blending ratio (namely 5% by mass) is adopted in
variables 𝑰 ∗ . this paper.
1 Set iteration index 𝜈 ← 0; Three scenarios are set up for comparative analysis to validate the
2 Solve the Model 2 to obtain an initial solution; developed model’s accuracy and effectiveness. The three scenarios are
| (𝜈) | described as follows.
3 while |𝑓 − 𝑓 (𝜈−1) | > 𝜀𝑓 do
| | S1: steady-state model without considering hydrogen injection.
4 // solve the following model
S2: steady-state model considering 5% mass fraction of hydrogen
∑ ∑
Model 3: min 𝑓 + 𝛼 (𝑣) 𝛿 + 𝛽 (𝑣) 𝜎 injection.
S3: quasi-steady-state model considering 5% mass fraction of hydro-
𝑠.𝑡. (1a)–(3), (4b)–(14f), (18a)–(18b), (20a)–(22b), (24a)–(24b)
gen injection.
5 // update penalty factor;
6 4.1. Comparison between steady-state models considering and not consid-
ering hydrogen injection
𝛼 (𝜈+1) ← min(𝜅𝛼 𝛼 (𝜈) , 𝛼 max )
In this subsection, the impact of considering hydrogen injection is
𝛽 (𝜈+1) ← min(𝜅𝛽 𝛽 (𝜈) , 𝛽 max )
studied. Fig. 5 shows the optimization results of total costs, purchased
7 // update iteration index; electricity costs, gas costs, and wind curtailment costs for S1 and S2.
8 𝜈 ← 𝜈 + 1; Based on Fig. 5, it can be concluded that the purchased electricity
9 end cost, purchased gas cost, wind curtailment penalty cost, and total cost
of S2 are all lower than those of S1. This is because P2H units can
convert surplus wind power into hydrogen, which can be injected into
natural gas pipelines, thereby mitigating the penalties imposed for wind
conversion efficiency coefficient of each P2H unit is 10 kg/MWh and curtailment. The injected hydrogen can substitute natural gas as a fuel
the capacity of each HSM is 3 t. source for gas turbines, reducing the costs of purchasing electricity and
The case study is conducted on a desktop computer equipped with natural gas. It leads to more cost-effective operations by enhancing
an AMD Ryzen 5 2600 Six-Core Processor (3.40 GHz) and 32 GB of overall economic efficiency.
memory. MATLAB 2021b is used for programming, and the YALMIP In order to further analyze the impact of hydrogen injection on
package invokes the GUROBI 10.0.1 solver for optimization. wind power consumption, Fig. 6 illustrates a comparison of wind power
According to [15], the relationship between mass fraction and consumption between S1 and S2.
volume fraction is depicted by Eq. (25). According to Fig. 6, in S1, the system experiences a certain degree of
( 𝜌H 𝜌H )−1 curtailed wind power in multiple periods, including 1:00–7:00, 10:00–
𝑐v = 1 − 2 + 2 ⋅ 𝑐𝑚−1 (25) 11:00, 16:00–17:00, and 24:00, totaling 12 time periods. Specifically,
𝜌NG 𝜌NG
the curtailed wind power during 1:00–7:00 is relatively significant. In
where 𝜌H and 𝜌NG are the density of the hydrogen and natural gas S2, the system experiences a certain degree of curtailed wind power in
respectively at standard conditions. They are 0.0899 kg∕m3 and 0.7174 9 time periods, including 1:00–6:00, 10:00–11:00, and 24:00. The wind
kg∕m3 . power curtailment during each period is notably reduced, particularly
Upon conversion, a mass fraction of 5% corresponds to a volume during the 1:00–6:00 time period, compared to S1. Additionally, the
fraction of 30%, while a mass fraction of 3% corresponds to a volume periods with wind power curtailment are reduced by three. It reflects

7
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Fig. 4. Profile of power load, gas load and maximum available wind power.

Fig. 5. Cost of S1 and S2.

Fig. 6. Comparison of wind power consumption between S1 and S2.

8
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Table 2
Parameters of coal-fired generators.
Electric node 𝑃max (MW) 𝑃min (MW) Upward ramp Downward ramp Unit purchasing cost of
rate (MW/h) rate (MW/h) electricity ($/MWh)
7 300 200 60 60 30.82
13 591 100 100 100 30.82
18 400 250 80 80 20.84
21 400 50 80 80 26.90
23 350 50 100 100 32.22

Table 3
Parameters of gas turbines.
Electric node Gas node 𝑃max (MW) 𝑃min (MW) Upward ramp Downward ramp Conversion efficiency
rate (MW/h) rate (MW/h) coefficient (kg/MWh)
1 12 152 20 50 50 15
2 12 152 30 50 50 15
15 10 60 10 20 20 15
15 10 155 30 40 40 15
16 7 155 30 40 40 15
22 6 300 50 150 150 15
23 6 60 10 20 20 15

Table 4 Table 6
Parameters of gas sources. Gas purchase cost and wind curtailment cost between S2 and S3.
Gas node 𝑄max (kg) 𝑄min (kg) Unit purchasing cost Scenario Gas purchase Wind curtailment
of gas ($/kg) cost (×106 $) cost (×106 $)
1 8000 0 2.0 S2 0.7331 0.1433
3 8000 0 2.0 S3 0.6982 0.1417
11 8000 0 2.0

Table 5 output of natural gas from the gas source. Besides, using surplus wind
Parameters of P2H units.
P2H (MW) P2H
power to produce hydrogen can increase wind power consumption.
Electric node Gas node 𝑃max 𝑃min (MW)
The hydrogen-blended natural gas produced can be used to fuel gas
5 3 150 0 turbines. As its unit energy cost is lower than that of natural gas
7 1 150 0
and thermal power units, the output of gas turbines will increase and
18 11 150 0
replace coal-fired generators as a source of energy supply to the power
system. It leads to a reduction in the output of coal-fired generators
indirectly, making the operation more economically efficient.
that injecting hydrogen into natural gas pipelines can effectively reduce
wind power curtailment and decrease the periods with wind power 4.2. Comparison between quasi-steady-state and steady-state modeling of
curtailment in the system. hydrogen blended integrated electricity-gas system
To further analyze how hydrogen injection affects system operation
and thus wind power consumption, Fig. 7 compares the electricity and Hydrogen produced during surplus wind power periods can be
gas dispatch results between S1 and S2. stored by linepack storage. When the system load is high, linepack
According to the electricity and gas dispatch results in S2 shown in storage can release the stored energy to replace expensive power and
Fig. 7, during the 1:00–8:00 time period, the P2H converts a portion of gas sources to supply energy to the system. The linepack storage and
wind power that is not fully absorbed by the power grid into hydrogen, its variation in S3 is shown in Fig. 8.
with some of it injected into the natural gas pipelines and the rest Fig. 8 shows that during the period 1:00–7:00, the system linepack
stored in the HSM. As a result, the amount of hydrogen stored in storage is increased to absorb more hydrogen to reduce the wind power
the HSM increases during this period. During the 8:00–12:00 time curtailment penalty since the system faces significant wind power
period, the amount of hydrogen produced is reduced compared to the curtailment during this period. During the 7:00–12:00 time period, due
previous period, but it is still sufficient to maintain injection into the to the decrease in the maximum available wind power compared to
natural gas pipelines. The hydrogen capacity in the HSM exhibits an the previous period and a slight increase in the total electrical load,
overall fluctuating trend during this period. During the 12:00–24:00 the linepack storage remains stable with some fluctuations. During
time period, as most of the wind power is absorbed by the system, the 12:00–24:00 time period, as the overall maximum available wind
the hydrogen in the HSM is injected into the natural gas pipelines power is lower than that in the 7:00–12:00 time period, the amount
to maintain the fixed hydrogen concentration. As a result, the overall of hydrogen produced is reduced. The total electric and gas load also
amount of hydrogen in the HSM decreases during this period. reaches peak values during this period. As a result, the linepack storage
Based on the dispatch results of gas turbines and coal-fired gener- exhibits an overall decreasing trend, and stored energy is released to
ators in S1 and S2 shown in Fig. 7(a), it can be observed that during supplement the insufficient energy supply from power and gas sources
most of the dispatch period, the output of gas turbines in S2 is higher to meet the electric demand.
than that in S1. In comparison, the output of coal-fired generators in Table 6 compares gas purchasing cost and wind curtailment penalty
S2 is lower than in S1. According to the gas source dispatch results cost between S2 and S3 to analyze further the impact of linepack on the
in S1 and S2 shown in Fig. 7(b), it can be observed that during most system operating cost.
of the dispatch period, the output of the gas source in S2 is lower According to Table 6, after considering linepack storage charac-
than that in S1. It reflects that injecting hydrogen produced by P2H teristics in the hydrogen blended quasi-steady-state model, both the
technology into the natural gas grid will replace a portion of natural system’s gas purchasing cost and wind curtailment penalty cost have
gas as a source of energy for the natural gas system, reducing the decreased.

9
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Fig. 7. Optimal dispatch result of electricity–gas system between S1 and S2.

Fig. 8. Linepack storage and its variation of S3.

The comparison of hydrogen injected into the pipeline within 24 h into the natural gas pipelines during high wind curtailment periods.
between S2 and S3 is shown in Fig. 9. During high gas demand periods, linepack storage replaces a portion
According to Fig. 9, during the 1:00–6:00 time period, the amount of of gas source output to supply energy to the load, thereby reducing gas
hydrogen injected into the pipeline in S3 is higher than that in S2. This purchasing cost and wind curtailment penalty cost, which validates the
is because the linepack storage can store more hydrogen during high effectiveness of the hydrogen blended quasi-steady-state model.
wind curtailment periods, thus reducing expensive wind curtailment
penalties. However, during the 12:00–23:00 time period, the amount of 4.3. Impact of hydrogen blending concentration on wind power consumption
hydrogen injected into the pipeline in S3 is lower than that in S2. The
fixed blending ratio injection strategy adopted in S3 results in a lower To further analyze the impact of hydrogen blending concentration
gas source output in S3 compared to S2. During high wind curtailment on wind power consumption, two additional scenarios, S4 and S5, are
periods, the system stores a certain amount of gas as linepack storage, introduced, whose specific descriptions are as follows:
which can be released during high gas demand periods to reduce S4: quasi-steady-state model considering 3% mass fraction of hydro-
expensive gas purchasing costs. Therefore, overall, linepack storage gen injection.
enables the system to inject more surplus hydrogen from wind power S5: quasi-steady-state model without hydrogen injection.

10
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Fig. 9. Comparison of hydrogen into the pipeline between S2 and S3.

Fig. 10. Comparison of wind power output in different scenarios.

Table 7 Table 8
Wind curtailment rates among dif- Parameters of the S-SOCP method.
ferent scenarios. 𝜅𝛼 𝜅𝛽 𝛼 max 𝛽 max 𝜀𝑓
Scenario 𝐷𝑐𝑢𝑟𝑡 4 4
2 2 1 × 10 1 × 10 1 × 10−6
S3 13.44%
S4 16.74%
S5 24.53%
4.4. Effectiveness of the proposed solution method

To validate the effectiveness of the proposed S-SOCP method, S3

The wind power output curves of S3, S4, and S5 and the maximum
is solved using this method as an example. Parameters of the S-SOCP
available wind power curve are shown in Fig. 10.
method are depicted in Table 8.
According to Fig. 10, during high wind curtailment periods such as
The convergence curve of operation cost is shown in Fig. 11. The
1:00–7:00, 10:00–11:00, 15:00–17:00, and 23:00–24:00, injecting the
S-SOCP method is an iterative approach that relaxes and expands the
hydrogen produced by P2H into natural gas pipelines can significantly
reduce wind curtailment. Additionally, as hydrogen blending concen- feasible region of the original problem at the beginning of the iteration
tration increases, more wind power can be integrated through P2H, to find a feasible solution close to the optimal solution quickly. As the
reducing wind curtailment and improving wind power consumption. iteration progresses, the feasible region of the problem is gradually
During the 15:00–17:00 period, both S4 and S5 experiences some reduced by penalizing the penalty term, ultimately approaching the
wind curtailment, while in S3, increasing the hydrogen blending con- feasible region of the original problem. Therefore, as shown in Fig. 11,
centration can achieve almost complete wind power consumption. A it can be observed that as the iterations progress, the value of the
comparison of wind curtailment rate between S3, S4, and S5 is shown operation cost gradually increases and eventually converges. In the
in Table 7. Wind curtailment rate 𝐷𝑐𝑢𝑟𝑡 is defined as the ratio of the fourth iteration, the relative error of the operation cost has already
total curtailed wind power to the available wind power during a day. converged to below 1 × 10−4 . The relative error of auxiliary variables
∑ ∑ 𝑎𝑣𝑎 w 𝛷 and 𝑝𝑟𝑠 also rapidly decrease to below 1% as well. It indicates
𝑡∈ 𝑠∈ (𝑃w,𝑠,𝑡 − 𝑃𝑠,𝑡 )
𝐷𝑐𝑢𝑟𝑡 = ∑ ∑ (26) that the S-SOCP method converges rapidly, and the tightened feasible
𝑎𝑣𝑎
𝑡∈ 𝑠∈ 𝑃w,𝑠,𝑡 region quickly approaches the feasible region of the original problem.
According to Table 7, increasing the hydrogen blending concentra- Additionally, it is worth noting that starting from the 5th iteration,
tion by 5% reduces the wind curtailment rate by 10.09% compared the convergence trend line is already very close to the reference line.
to the scenario without hydrogen blending. As the hydrogen blending However, as the iterations continue, the convergence speed towards
concentration increases, the wind curtailment rate further decreases. the reference value slows down. Since each iteration involves solving a

11
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

Fig. 11. Convergence curve of the proposed S-SOCP method.

Table 9 Table 10
Comparison of computation time and optimality of different methods. Comparison of feasibility of different methods.
Method Computation time (s) Relative error of operation Method A Method B Method C
cost to Method D
𝐸𝑚𝑟 0.02% 11.96% 16.85%
Method A 48.19 5.13 × 10−5 𝐸𝑎𝑟 0.0002% 1.53% 1.70%
Method B 196.77 2.59 × 10−4
Method C 223.90 1.87 × 10−4
Method D 535.42 − Table 11
Solution results with different penalty factors.
Penalty factor Solution time (s) Iteration steps
1.5 86.25 8
mixed-integer problem, the computation is relatively time-consuming. 2.0 48.19 4
Therefore, it is necessary to strike a balance between solution time and 2.5 50.32 4
optimality. 5.0 200.56 3
In this section, the proposed method is compared with three meth-
ods in order to verify the effectiveness of the proposed solution method.
Method A is the approach proposed in this paper. Method B makes of constraint violation. According to Ref. [32], a constraint violation
a segmented linear approximation to the nonlinear equation of the index 𝐸𝑅𝑚𝑛,𝑡 is defined for this purpose.
original problem, where ten segments are used, transforming the non-
| 2 |
linear problem into a mixed-integer linear programming problem that |𝑝𝑟𝑚,𝑡 − 𝑝𝑟2𝑛,𝑡 − 𝐾𝑚𝑛 𝑄𝑚𝑛,𝑡 |𝑄𝑚𝑛,𝑡 ||
is easier to solve. In Method C, the nonlinear gas flow equality is 𝐸𝑅𝑚𝑛,𝑡 = | | , ∀𝑚𝑛 ∈  ∖ , ∀𝑡 ∈ 
g c (27)
𝐾𝑚𝑛 𝑄𝑚𝑛,𝑡 |𝑄𝑚𝑛,𝑡 |
relaxed into piecewise McCormick constraints, which are tightened by
adjusting the number of segments. Twenty segments are used in Method where 𝐸𝑅𝑚𝑛,𝑡 reflects the degree of violation of the nonlinear constraint
C. The Gurobi solver is used in Method D to obtain a globally optimal (5). The closer it is to zero, the smaller the violation of the nonlinear
constraint, which indicates better solution feasibility. The maximum
solution to the nonlinear problem based on the spatial branch-and-
constraint violation 𝐸𝑚𝑟 and the average constraint violation 𝐸𝑎𝑟 are
bound method. Table 9 shows the solution time for four methods and
the relative error of the optimal operation cost compared to Method D defined as below.
for the first three methods. 𝐸𝑚𝑟 = max 𝐸𝑅𝑚𝑛,𝑡 (28a)
𝑚𝑛,𝑡
According to Table 9, the relative error of the obtained optimal
∑ [ ]
objective function value for the proposed method is around 5 × 10−5 𝐸𝑎𝑟 = 𝐸𝑅𝑚𝑛,𝑡 ∕ (|g | − |c |) ∗ | | (28b)
compared to Method D, while the relative error for Method B and 𝑚𝑛,𝑡
Method C is around 5 × 10−4 . It indicates that the proposed method where 𝐸𝑚𝑟 reflects the extreme cases of constraint violation, while 𝐸𝑎𝑟
achieves better optimality than the other two. The solution quality denotes the overall constraint violation. The closer 𝐸𝑚𝑟 and 𝐸𝑎𝑟 are
for Method B and Method C is achieved at the expense of increased to zero, the smaller the constraint violation in the obtained solution,
solution time. In other words, the number of segments needs to be indicating better feasibility.
increased to obtain higher-quality solutions. However, as the number of The feasibility comparison among different methods is presented in
segments increases, the corresponding increase in the number of binary Table 10.
variables unavoidably prolongs the solution time. From Table 9, it can According to Table 10, Method B and Method C yield the maximum
be observed that the solution time for Method A is around 50s, while constraint violation above 10% and the average constraint violation
Method B and Method C both take around 200s. The proposed method above 1%. In contrast, the proposed method achieves the maximum
has a solution time that is one-fourth of the other two methods. Due to constraint violation below 0.1% and the average constraint violation
the global solver’s use of spatial branch and bound method, obtaining below 0.001%. It shows that the proposed method significantly reduces
high-quality global optimal solutions requires a significant number the degree of constraint violation compared to the other two methods,
of branching steps, resulting in longer solution time. The proposed leading to improved solution feasibility.
method, on the other hand, achieves a 91% improvement in solution To investigate the impact of penalty factors on the performance of
speed compared to the global solver. Therefore, the proposed method the solution performance, multiple sets of penalty coefficients are set,
exhibits a shorter solution time and lower relative error in optimality. and the solution results are depicted in Table 11.
To validate the feasibility of the proposed method, a comparison is From Table 11, it can be concluded that when the penalty coefficient
made between the proposed method, Method B, and Method C in terms is small, an initial solution can be obtained quickly. However, it can

12
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

result in an increasing number of iterations significantly, leading to an Appendix. Derivation of the linepack storage calculation formula-
increase in solution time. When the penalty factor increases, although tion
the total number of iterations decreases, the difficulty of obtaining
feasible solutions at each step increases, leading to a significant increase The calculation formulation for the linepack storage of pipeline 𝑚𝑛
in solution time. Therefore, multiple attempts are needed to find a at time 𝑡 is reflected by Eq. (A.1).
moderate penalty coefficient that minimizes the solution time.
𝐿𝑚𝑛
In summary, the proposed method outperforms Method B and ℎ𝑚𝑛,𝑡 = 𝐴𝑚𝑛 𝜌(𝑥)𝑑𝑥 (A.1)
Method C regarding optimality and feasibility. It has a solution time of ∫0
one-fourth of Method B and Method C, and it is 91% faster than Method According to Ref. [29], the relationship between density of mixed
D. Besides, the obtained optimal solution by Method A has achieved a hydrogen–natural gas and the pressure at both ends is donated by
minor constraint violation of the nonlinear equation (4) than Method B Eq. (A.2).
and Method C. These results validate the effectiveness of the proposed
method. 𝑝 = 𝜌 ⋅ 𝑉 (𝛾) (A.2)

Based on Eqs. (A.1)–(A.2), the calculated formula of gas linepack

5. Conclusion can be obtained shown in Eq. (A.3) via a numerical integration using
the trapezoidal rule.
This paper proposes a quasi-steady-state optimization scheduling
𝐴𝑚𝑛 𝐿𝑚𝑛 𝐴 𝑝𝑟𝑚,𝑡 + 𝑝𝑟𝑛,𝑡
model for HBIEGS that considers gas linepack, with a bi-directional ℎ𝑚𝑛,𝑡 = 𝑝𝑟(𝑥)𝑑𝑥 = 𝑚𝑛 ⋅ (A.3)
flow of gas modeled within the model. The proposed model is refor- 𝑉 (𝛾) ∫0 𝑉 (𝛾) 2
mulated into a MISOCP model by convex relaxation. It is further solved
using a sequential solution method, which exhibits an excellent solution References
performance. The main conclusions of this paper are as follows.
[1] R. Warren, O. Andrews, S. Brown, F.J. Colón-González, N. Forstenhäusler, D.E.
• By converting surplus renewable energy into hydrogen through Gernaat, P. Goodwin, I. Harris, Y. He, C. Hope, et al., Quantifying risks avoided
water electrolysis and injecting it into natural gas pipelines, the by limiting global warming to 1.5 or 2 ◦ C above pre-industrial levels, Clim.
Change 172 (3–4) (2022) 39, http://dx.doi.org/10.1007/s10584-021-03277-9.
amount and duration of wind curtailment are reduced, and the
[2] C. Gebara, A. Laurent, National SDG-7 performance assessment to support
operating costs of the system are lowered, thereby making the achieving sustainable energy for all within planetary limits, Renew. Sustain.
system more economically efficient. Energy Rev. 173 (2023) 112934, http://dx.doi.org/10.1016/j.rser.2022.112934.
• The developed HBIEGS quasi-steady state model can accurately [3] B.M. Campbell, J. Hansen, J. Rioux, C.M. Stirling, S. Twomlow, E. (Lini)
reflect the impact of hydrogen blended into gas network pipelines Wollenberg, Urgent action to combat climate change and its impacts (SDG 13):
on wind power consumption. Specifically, the results show that transforming agriculture and food systems, Curr. Opin. Environ. Sustain. 34
blending hydrogen with a mass fraction of 5% can improve (2018) 13–20, http://dx.doi.org/10.1016/j.cosust.2018.06.005.
[4] P.R. Liu, A.E. Raftery, Country-based rate of emissions reductions should increase
wind power consumption by 10.09%. Compared to steady-state
by 80% beyond nationally determined contributions to meet the 2 C target,
models, the results of the proposed model show that the wind
Commun. Earth Environ. 2 (1) (2021) 29, http://dx.doi.org/10.1038/s43247-
consumption rate has been increased. 021-00097-8.
• The proposed S-SOCP method reformulates the original nonlinear [5] W. Huang, B. Zhang, L. Ge, J. He, W. Liao, P. Ma, Day-ahead optimal scheduling
problem into a MISOCP form. The computation results demon- strategy for electrolytic water to hydrogen production in zero-carbon parks type
strate that the feasibility of the solution obtained is further im- microgrid for optimal utilization of electrolyzer, J. Energy Storage 68 (2023)
proved. Moreover, the proposed method is 91% more computa- 107653, http://dx.doi.org/10.1016/j.est.2023.107653.
[6] R. Wang, R. Zhang, Techno-economic analysis and optimization of hybrid energy
tionally faster than a general-purpose nonlinear solver.
systems based on hydrogen storage for sustainable energy utilization by a
biological-inspired optimization algorithm, J. Energy Storage 66 (2023) 107469,
This paper mainly focuses on a quasi-steady state modeling method
http://dx.doi.org/10.1016/j.est.2023.107469.
with a fixed hydrogen blending ratio and simplified hydrogen usage.
[7] C. Tarhan, M.A. Çil, A study on hydrogen, the clean energy of the future:
In the future, the quasi-steady state modeling methods considering Hydrogen storage methods, J. Energy Storage 40 (2021) 102676, http://dx.doi.
hydrogen load and a variable hydrogen blending ratio will be further org/10.1016/j.est.2021.102676.
investigated. [8] B. Lebrouhi, J. Djoupo, B. Lamrani, K. Benabdelaziz, T. Kousksou, Global hydro-
gen development - A technological and geopolitical overview, Int. J. Hydrogen
Declaration of competing interest Energy 47 (11) (2022) 7016–7048, http://dx.doi.org/10.1016/j.ijhydene.2021.
12.076.
[9] T. Isaac, Hydeploy: The UK’s first hydrogen blending deployment project, Clean
The authors declare that they have no known competing finan- Energy 3 (2) (2019) 114–125, http://dx.doi.org/10.1093/ce/zkz006.
cial interests or personal relationships that could have appeared to [10] G. Jia, M. Lei, M. Li, W. Xu, R. Li, Y. Lu, M. Cai, Hydrogen embrittlement in
influence the work reported in this paper. hydrogen-blended natural gas transportation systems: A review, Int. J. Hydrogen
Energy (2023) http://dx.doi.org/10.1016/j.ijhydene.2023.04.266.
Data availability [11] S. Wang, H. Hui, P. Siano, Resilience of gas interchangeability in hydrogen-
blended integrated electricity and gas systems: A transient approach with
dynamic gas composition tracking, iEnergy 2 (2) (2023) 143–154, http://dx.
Data will be made available on request. doi.org/10.23919/IEN.2023.0016.
[12] S. Wang, H. Hui, J. Zhai, Short-term reliability assessment of integrated power-
Acknowledgments gas systems with hydrogen injections using universal generating function, IEEE
Trans. Ind. Appl. 59 (5) (2023) 5760–5773, http://dx.doi.org/10.1109/TIA.2023.
3286821.
This work was supported in part by the National Natural Sci-
[13] S. Wang, J. Zhai, H. Hui, Optimal energy flow in integrated electricity and gas
ence Foundation of China (51977082), Guangdong Basic and Ap-
systems with injection of alternative gas, IEEE Trans. Sustain. Energy 14 (3)
plied Basic Research Foundation, China (2021A1515110675), Science (2023) 1540–1557, http://dx.doi.org/10.1109/TSTE.2023.3237229.
and Technology Project of China Southern Power Grid Co., Ltd [14] I. Saedi, S. Mhanna, P. Mancarella, Integrated electricity and gas system
(037700KK52220040 (GDKJXM20220893)), Guangzhou Science and modelling with hydrogen injections and gas composition tracking, Appl. Energy
Technology Plan Project, China (202201010577). 303 (2021) 117598, http://dx.doi.org/10.1016/j.apenergy.2021.117598.

13
C. Wu et al. Journal of Energy Storage 76 (2024) 109718

[15] W. Zheng, J. Li, K. Lei, Z. Shao, J. Li, Z. Xu, Optimal dispatch of HCNG [24] Y. He, M. Shahidehpour, Z. Li, C. Guo, B. Zhu, Robust constrained operation
penetrated integrated energy system based on modelling of HCNG process, Int. of integrated electricity-natural gas system considering distributed natural gas
J. Hydrogen Energy 48 (51) (2023) 19437–19449, http://dx.doi.org/10.1016/j. storage, IEEE Trans. Sustain. Energy 9 (3) (2018) 1061–1071, http://dx.doi.org/
ijhydene.2023.02.056. 10.1109/TSTE.2017.2764004.
[16] K. Liu, L.T. Biegler, B. Zhang, Q. Chen, Dynamic optimization of natural gas [25] C. Wang, W. Wei, J. Wang, L. Bai, Y. Liang, T. Bi, Convex optimization based
pipeline networks with demand and composition uncertainty, Chem. Eng. Sci. distributed optimal gas-power flow calculation, IEEE Trans. Sustain. Energy 9
215 (2020) 115449, http://dx.doi.org/10.1016/j.ces.2019.115449. (3) (2018) 1145–1156, http://dx.doi.org/10.1109/TSTE.2017.2771954.
[17] L. Chi, H. Su, E. Zio, M. Qadrdan, J. Zhou, L. Zhang, L. Fan, Z. Yang, F. Xie, L. [26] S. Mhanna, I. Saedi, P. Mancarella, Z. Zhang, Coordinated operation of electricity
Zuo, J. Zhang, A systematic framework for the assessment of the reliability of and gas-hydrogen systems with transient gas flow and hydrogen concentration
energy supply in integrated energy systems based on a quasi-steady-state model, tracking, Electr. Power Syst. Res. 211 (2022) 108499, http://dx.doi.org/10.1016/
Energy 263 (2023) 125740, http://dx.doi.org/10.1016/j.energy.2022.125740. j.epsr.2022.108499.
[27] W. Wei, J. Wang, N. Li, S. Mei, Optimal power flow of radial networks and its
[18] A. Schwele, C. Ordoudis, J. Kazempour, P. Pinson, Coordination of power and
variations: A sequential convex optimization approach, IEEE Trans. Smart Grid
natural gas systems: Convexification approaches for linepack modeling, in: 2019
8 (6) (2017) 2974–2987, http://dx.doi.org/10.1109/TSG.2017.2684183.
IEEE Milan PowerTech, 2019, pp. 1–6, http://dx.doi.org/10.1109/PTC.2019.
[28] Z. Liye, D. Haitao, S. Guijun, N. Chen, S. Gang, L. Wei, S. Chen, L. Xueying, L.
8810632.
Yanghui, J. Guanwei, et al., Research progress of natural gas follow-up hydrogen
[19] Y. Chen, J. Zhu, H. Ji, X. Zhou, Day-ahead optimal scheduling for integrated en-
mixing technology, Mech. Eng. 44 (4) (2022) 755–766, http://dx.doi.org/10.
ergy system considering the line pack of gas network, in: 2021 IEEE Sustainable
6052/1000-0879-22-056.
Power and Energy Conference, ISPEC, 2021, pp. 965–969, http://dx.doi.org/10.
[29] S.R. Kazi, K. Sundar, S. Srinivasan, A. Zlotnik, Modeling and optimization of
1109/iSPEC53008.2021.9735863.
steady flow of natural gas and hydrogen mixtures in pipeline networks, 2022,
[20] F. Liu, Z. Bie, X. Wang, Day-ahead dispatch of integrated electricity and natural http://dx.doi.org/10.48550/arXiv.2212.00961, arXiv preprint arXiv:2212.00961.
gas system considering reserve scheduling and renewable uncertainties, IEEE [30] Y. Zhao, V. McDonell, S. Samuelsen, Experimental assessment of the combustion
Trans. Sustain. Energy 10 (2) (2019) 646–658, http://dx.doi.org/10.1109/TSTE. performance of an oven burner operated on pipeline natural gas mixed with
2018.2843121. hydrogen, Int. J. Hydrogen Energy 44 (47) (2019) 26049–26062, http://dx.doi.
[21] J. Zhou, S. Li, X. Zhou, C. Li, Z. Xiong, Y. Zhao, Operation optimization for gas- org/10.1016/j.ijhydene.2019.08.011.
electric integrated energy system with hydrogen storage module, Int. J. Hydrogen [31] C. Zhou, Z. Yang, G. Chen, Q. Zhang, Y. Yang, Study on leakage and explosion
Energy 47 (86) (2022) 36622–36639, http://dx.doi.org/10.1016/j.ijhydene.2022. consequence for hydrogen blended natural gas in urban distribution networks,
08.224. Int. J. Hydrogen Energy 47 (63) (2022) 27096–27115, http://dx.doi.org/10.
[22] C.A. Saldarriaga, R.A. Hincapié, H. Salazar, A holistic approach for planning 1016/j.ijhydene.2022.06.064.
natural gas and electricity distribution networks, IEEE Trans. Power Syst. 28 (4) [32] Y. He, M. Yan, M. Shahidehpour, Z. Li, C. Guo, L. Wu, Y. Ding, Decentralized
(2013) 4052–4063, http://dx.doi.org/10.1109/TPWRS.2013.2268859. optimization of multi-area electricity-natural gas flows based on cone reformu-
[23] Y. Hu, Z. Bie, T. Ding, Y. Lin, An NSGA-II based multi-objective optimization lation, IEEE Trans. Power Syst. 33 (4) (2017) 4531–4542, http://dx.doi.org/10.
for combined gas and electricity network expansion planning, Appl. Energy 167 1109/TPWRS.2017.2788052.
(2016) 280–293, http://dx.doi.org/10.1016/j.apenergy.2015.10.148.

14