Risk-averse distributed optimization for integrated electricity-gas systems considering

ed
uncertainties of Wind-PV and Power-to-Gas.

v iew
re
er
pe
ot
tn
rin
ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

ed
iew
rev
er
pe
ot
tn
rin
ep
Pr

This Project Supported by the National Natural Science Foundation of China

(51877174).
Fan Liu, 15229338767@163.com; Jiandong Duan, duanjd@xaut.edu.cn;
Chen Wu, wuchen606@126.com; Qinxing Tian, 25234249437@qq.com.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

1 Risk-averse distributed optimization for integrated

electricity-gas systems considering uncertainties of Wind-PV

ed
2

3 and Power-to-Gas
4
5

iew
6 Fan Liua, Jiandong Duana, Chen Wua, and Qinxing Tiana

7 a School of Electrical Engineering, Xi’an University of Technology, Xi’an 710048, Shaanxi Province, China
8

9 ABSTRACT

ev
10 The uncertainties of wind power and photovoltaic (Wind-PV) challenges to the stable operation of integrated electricity-gas
11 systems, and limits its efficient consumption. Under this circumstance, an optimal operation model considering the cost risk of
12 Wind-PV uncertainties is proposed. Firstly, the cost risk of uncertainties in typical scenarios is measured by conditional value at
13 risk(CVaR). Secondly, considering that the power system and the natural gas system as different stakeholders have barriers in
14 information exchange, a distributed optimization operation model is established, and the uncertainty of the conversion efficiency

r
15 of the coupling unit from Power-to-Gas (P2G) is considered. Then, the integrated electricity-gas systems are decoupled through
16 consistency constraints, and the model is solved by an algorithm framework based on adaptive penalty ADMM. Finally, a case
17 study shows that: 1) the proposed model can measure the system operation risk caused by uncertainty factors and provide guidance
18
19
20
21
3) the convergence of the proposed method is excellent.

Keywords
er
for decision makers; 2) in order to reduce the operation cost of P2G, the uncertainty of conversion efficiency should be controlled;
pe
22 Wind power and photovoltaic uncertainty;Risk-averse;Conditional value at risk(CVaR);Adaptive penalty ADMM;Conversion
23 efficiency of Power-to-Gas(P2G)
24

25 1 Introduction

26 The integrated electricity-gas systems contains energy with multiple operating characteristics, which is an effective way to
ot

27 achieve multi energy coordination and complementarity, promote energy efficiency and reduce environmental pollution[1]. At the
28 same time, more and more new energy power generation which promotes the diversity of energy and alleviates the energy and
29 environmental problems caused by traditional generation[2]. However, the uncertainty of Wind-PV will bring challenges to their
30 efficient consumption, affect the supply and demand balance of the real-time operation of the system, and further affect the system's
tn

31 operation and the formulation of scheduling plans, and increase the cost of system rescheduling [2],[4]. Therefore, it is necessary
32 to assess the risks brought by Wind-PV uncertainties to the integrated electricity and natural gas system, and comprehensively
33 balance the operating costs and risks of the system.
34 At present, there are many models for optimal operation considering the uncertainty of Wind-PV, which can be divided into
35 expected value model, chance constraint model and relevant chance constraint model from the perspective of modeling mechanism
rin

36 [5]. The expected value model is the main research method for stochastic optimization problems. It is based on the expected
37 constraint conditions to transform the stochastic programming problem into a deterministic programming problem in order to
38 achieve the optimal expected value of the objective function[6]. The chance constraint model is the optimization result obtained
39 when the output meets the constraint conditions based on a certain probability description uncertainty under a certain confidence
40 level[7]. The relevant chance constraint model is to obtain the optimal decision result by taking advantage of the chance of
ep

41 maximizing the establishment of random events under the uncertainty environment[8]. The above model can well describe and
42 deal with uncertainty from the perspective of cost expectation and probability constraint, but it is difficult to accurately reflect the
43 risk attitude of dispatchers when making decisions, and ignores the description of risk losses beyond the confidence level, thus
44 affecting the accuracy of optimal decisions[9].
45 In addition, the integrated electricity and natural gas system has attracted more and more attention, and a large number of
Pr

46 research results have emerged. Ref. [10] proposed a distributed robust optimization model for multiple uncertainties in power and
47 natural gas systems. Ref. [11] used P2G technology to optimize the use of renewable energy at the distribution network level,
48 including natural gas and electricity system. Ref. [12] established a two-stage optimization model based on risk for wind power
49 uncertainty in integrated electricity-gas systems, which can effectively reduce operational risks. To address carbon emissions and
50 renewable energy consumption issues, the Ref. [13] carried out optimal dispatching considering uncertainty for compressed natural

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

51 gas and power system. Ref. [14] proposed a peak load regulation model to smooth the net load curve by coordinating the operation
52 of gas turbine generator units and generators. In [15], a hybrid model based on distributed robust chance constraint and interval
53 optimization is proposed to consider the impact of wind power uncertainty. As the main coupling unit of the integrated electricity

ed
54 and natural gas system, the uncertainty of conversion efficiency exists in the process of power to gas conversion [16], but there is
55 little research on it at present.
Nomenclature , Maximum intake rate and minimum intake
rate of air reservoir k
Abbreviation , Maximum exhaust rate and minimum
exhaust rate of air reservoir k

iew
CVaR Conditional value at risk Cost of purchasing power and natural gas
,
from superior systems
P2G Power to gas Gas turbine purchase cost coefficient
ADMM Alternating direction multiplier method Revenue coefficient of P2G sold to natural
gas network
GS Gas storage , Upper and lower limits of node voltage

ev
WT Wind turbine 、 Upper and lower limits of absorbed power
of the main network
PT Photovoltaic unit 、 Upper and lower limits of reactive power
absorbed by the main network
Sets and parameters Daily operating cost of P2G

r
Operating cost coefficient of gas storage
tank
t Index of the period The upper limit of gas well output
p
g
Index of P2G equipment
Index of gas turbine
er ,
,
Lagrange primary of P2G and gas storage
Secondary penalty coefficients of P2G and
gas storage
pe
Set of all branches with node j as the head , Equivalent resistance and equivalent
node reactance on branch ij
Set of all branches with node j as the end Current flowing on branch ij
node
P2G device set at node j
Gas turbine set at node j Decision variables
ot

Transmission parameters of pipeline mn Generation capacity of gas turbine g

Collection of wind turbines on node j New energy power generation consumed
by P2G
Collection of photovoltaic units on node j Volume of gas purchased by the natural gas
tn

system from the supplier

, Node pressure at node m and node n Volume of natural gas sold by P2G to the
gas network
, Upper and lower output limits of gas turbine Volume of gas sold from natural gas system
g to electric power system
rin

, Upper and lower ramping limits of gas Purchase power from the main network
turbine g
, Upper and lower output limits of P2G , Wind-PV output
, Upper and lower ramping limits of P2G
Conversion efficiency of gas turbines
ep

Conversion efficiency of P2G

High calorific value of natural gas
, Maximum and minimum gas storage
capacity of gas storage tank k
Pr

56 Conditional value at risk (CVaR) is a financial risk assessment method. At present, there are many researches in power system
57 using it to measure the impact of uncertainty on optimal dispatching. Ref. [17] proposes to effectively measure the operational risk
58 caused by the uncertain changes of renewable energy generation and load in the integrated energy system based on CVaR, and to
59 balance the risk and cost. Ref. [18] established a risk assessment model for park level integrated energy system considering the
60 uncertainty of energy price, and used CVaR to quantify the impact of extreme uncertainty of energy price on the planning of
2

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

61 integrated energy system. Ref. [19] quantifies the uncertainty risk of new energy generation through CVaR to support the economic
62 operation of integrated energy systems. Ref. [20] introduced CVaR as an appropriate risk aversion index in the multi microgrid
63 system to avoid the low profit risk caused by the uncertainties. In [21], CVaR is used to measure the risk of abandoned wind power,

ed
64 and a day ahead scheduling model of power system is proposed. In [22], distributed robust CVaR constraints are proposed in the
65 optimal power flow problem to deal with the risks brought by the uncertainty of wind power generation. Ref. [23] used CVR to
66 quantitatively evaluate uncertainty risks and achieved risk avoidance decisions for renewable energy investment. In [24], an
67 adaptive risk averse stochastic programming method was proposed to manage the comprehensive energy demand on the ship, so
68 as to minimize the navigation cost and conditional risk value. From the perspective of the impact caused by uncertainties such as
69 renewable energy output, energy price and load demand, the above research uses CVaR to study the impact of different

iew
70 uncertainties on optimal operation. Therefore, CVaR can comprehensively weigh the economy and risk of the operation of the
71 integrated energy system, and assess the risk loss of abandoned Wind-PV and load shedding caused by the uncertainty. However,
72 these studies tend to analyze the integrated energy system as a unified subject and carry out centralized optimization, without
73 considering that each subsystem belongs to different subjects and there are barriers to information interaction between them [25].
74 In addition, the amount of information involved in centralized optimization is too complex, which will increase the computing
75 burden of the dispatching center.
76 Therefore, more and more research tends to distributed optimization to achieve the unity of autonomy of different systems.

ev
77 Distributed optimization does not require any central unit. All subsystems only need to interact with boundary information, which
78 not only ensures data privacy, but also reduces the cost of information interaction. At the same time, it has scalability, and can
79 better adapt to the optimization problem of multiple systems [26],[27]. The game based method describes the original problem in
80 the form of game, and realizes the overall optimization by ensuring the optimal interests of each subject[28]. The advantage of this
81 method is that it is completely distributed in structure and can be extended to more complex occasions, but it has disadvantages in

r
82 solving complexity[29]. The decoupling method is mainly based on the Lagrange multiplier method, including the augmented
83 Lagrange multiplier method, the principle method of auxiliary problems, and the alternating direction multiplier method
84 (ADMM)[30]. Lagrange method and auxiliary problem principle method can relax constraints, but the convergence is slow
85
86
87
88
er
[31],[32]. ADMM is widely used because of its fast convergence when it alternately optimizes the variables of subproblems[33].
However, due to the fixed penalty factor of ADMM, the convergence is still not ideal[31],[34].
To sum up, in this paper, the optimal operation of the integrated electricity and natural gas system considering the uncertainty
of the Wind-PV is conducted, and CVaR is used to evaluate the cost risk. In addition, an affine policy is proposed to model the
pe
89 uncertainty of P2G conversion efficiency. In the optimization method, the adaptive penalty ADMM is utilized to solve the
90 optimization model, and the excellent effect is achieved in convergence. The differences between our work and the aforementioned
91 researchs are shown in Table 1, and the main contributions can be summarized as follows.
92 (1) A quantitative method for cost risk caused by uncertainties of Wind-PV based on CVaR theory has been presented, and
93 impact of the uncertainty of Wind-PV on the optimal operation is investigated. The proposed method can balance the risk and
94 economy of system, providing risk-averse scheduling schemes for decision-makers to avoid the uncertainty risk.
95 (2) The uncertainty of P2G conversion efficiency is modeled by affine policy, and its effect is verified through simulation, which
ot

96 provides a reference for the application of P2G.

97 (3) Risk-averse distributed optimal operation of the integrated electricity and natural gas system is proposed, which can achieve
98 decomposition of the optimal operation and benefit the privacy and confidentiality protection of different stakeholders. The
99 proposed method is better than the standard ADMM for solving distributed optimization model in terms of convergence.
tn

100 The rest of this paper is organized as follows: Section 2 introduces the risk principle based on CVaR for Wind-PV uncertainties.
101 Section 3 puts forward the risk-averse optimizations for integrated electricity-gas systems. Section 4 introduces the adaptive
102 penalty ADMM-based optimal framework for integrated electricity-gas system. In Section 5, numerical examples verify the
103 effectiveness of the model. Finally, Section 6 draws the conclusions.
rin

104 2 CVaR based Risk principle for Wind-PV uncertainties

105 Firstly, the BP neural network model is used to practice the historical data, and the prediction error and initial prediction value
106 of the Wind-PV are obtained. Then Latin hypercube sampling is used to obtain numerous groups of prediction errors. Considering
107 the correlation of historical data prediction error, the auto regression moving average model of prediction error is established, and
ep

108 numerous groups of prediction error scenarios are obtained. Finally, a large number of scenes are reduced to obtain typical scenes.
109 After setting up the typical scenario set of the Wind-PV, CVaR is used to evaluate the cost risk.
110 Suppose is the feasible set of objective combination, is the dimensional optimization decision variable, is the
111 dimensional random vector that determines the system risk loss, is the probability density function of , and is the
112 loss function. Assuming that is the boundary value of , its distribution function is shown as
Pr

113 Table 1
114 Differences between this work and the aforementioned research.
Ref. Renewable energy Uncertainty Integrated electricity Risk Theory Distributed algorithm
and natural gas system
[10] √ √ √ Benders

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

[11] √ √
[12] √ √ √ Scenario-based Benders
[13] √ √ √
[14] √ √ √ √

ed
[15] √ √ √
[16] √ √ √
[17] √ √ √ CVaR Benders
[18] √ √ √ CVaR
[19] √ √ CVaR

iew
[20] √ √ CVaR
[21] √ √ CVaR
[22] √ √ CVaR
[23] √ √ CVaR
[24] √ CVaR
[25] √ √ √ ATC
[27] √ √ √ ATC
[28] √ √ Benders

ev
[29] √ √
[30] √ ADMM
[31] ADMM
[32] ADMM
[33] √ √ CVaR ADMM
[34] √ √ √ ADMM

r
This work √ √ √ CVaR ADMM

115 (1)

116
117
118
Given a certain confidence level

The expression for CVaR is shown as

, when
er, the set minimum value of is VaR, that is shown as
(2)
pe
119 (3)

120 where, means .

121 For discrete uncertain scene sets, the estimated value of equation (3) is formulated as
ot

122 (4)

123 where is the probability of scenario , and represents the risk loss on scenario .
124 When the abandoned Wind-PV and electrical load cut occur, the penalty cost is equal to the product of the abandoned
tn

125 Wind-PV and penalty coefficient . The electrical and gas load loss borne by the system is equal to the product of
126 the penalty coefficient and the total amount of electrical and gas load cut . As shown in the following formula.
127 (5)
rin

128 (6)
129 Assuming that the total scene set of the Wind-PV is , where is the total number of scenes,
130 the CVaR value of abandoning the Wind-PV and shedding off the electrical and gas load is shown as

131 (7)
ep

132 (8)

133 where and are the critical values of risk loss for the abandoned Wind-PV and electrical load loss respectively.
Pr

134 Therefore, the risk cost of considering the uncertainties of the Wind-PV is expressed as

135 (9)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

136 3 Risk-averse optimizations for integrated electricity-gas systems

137 A schematic diagram of a typical integrated electricity-gas system with new energy through coupling gas turbines and P2G is

ed
138 shown as Fig.1. Under the framework of the diagram, the integrated energy system of electricity and natural gas system is divided
139 into three functional modules: energy supply module, energy conversion and distribution module, and energy demand module. The
140 entire integrated electricity-gas system can ensure flexible operation in the conversion and distribution of electricity and natural
141 gas to meet the various energy needs of users.
Energy supply Energy conversion and distribution Energy demand

iew
Distribution network Power load

Main grid
GT P2G

ev
New energy GS

Gas well Gas distribution network Gas load

r
142 Power flow Natural gas flow
143 Fig.1 Schematic diagram of an integrated electricity-gas system
144

145
146
1) Optimal operation model of new energy power system: er
3.1 Optimization modelling considering the risk cost of Wind-PV uncertainties

CVaR is used to measure the risk cost of the Wind-PV uncertainty, which is added to the objective function to minimize the
pe
147 operation cost and reduce the risk cost as much as possible.
148 The optimization objective is to minimize the total cost of power system operation. The first part is the operation cost, mainly
149 including the power purchase cost from the main power grid, the gas consumption cost of gas turbine and the net cost of P2G operation.
150 The second part is the risk cost reflected by the risk aversion coefficient.

151 (10)
ot

152 where is the power purchase cost coefficient; is the gas purchase cost coefficient of gas turbine; T is the general dispatching
153 period; refers to the power purchased from the main network; is the total number of gas turbines; is the total number of
154 P2G; is the return coefficient of P2G selling gas to the gas network; is the volume of natural gas sold to the gas network;
tn

155 is the daily operating cost of P2G; is the risk preference coefficient; is the risk cost of the power system.
156 The optimized constraints mainly include distflow power flow constraints (11) - (13), node voltage constraints (14), power absorbed
157 from the main network constraints (15) - (16), P2G and gas turbine operation constraints (17) - (22).
rin

158 (11)

159 (12)
ep

160 (13)

161 (14)
Pr

162 (15)
163 (16)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

164 (17)

ed
165 (18)
166 (19)

167 (20)

iew
168 (21)
169 (22)
170 2) Optimal operation model of natural gas system:
171 The optimization objective is to minimize the total cost of natural gas system risk operation. The first part is the operating cost,
172 mainly including the cost of purchasing gas from natural gas suppliers and P2G, the operating cost of gas storage tanks, and the

ev
173 revenue of selling natural gas to the power system. The second part is the risk cost reflected by the risk aversion coefficient.

174

175 (23)

r
176 where and are the cost coefficient and volume of gas purchased from the supplier.
177 The optimized constraints mainly include node flow balance constraint (24), Weymouth equation constraint of pipeline (25), gas
178 storage tank operation constraint (26) - (28), node pressure and pipeline flow constraint (29), and gas purchase constraint (30).
179
er (24)
pe
180

181 (25)
ot

182 (26)
tn

183 (27)

184 (28)
rin

185 (29)
ep

186
(30)

187 3.2 The uncertainty of P2G conversion efficiency

Pr

188 As the main coupling unit of the integrated electricity and natural gas system, P2G strengthens the coupling between the power
189 system and the natural gas system and its conversion efficiency will also affect the operation of the integrated energy system. At
190 present, Power-to-Gas conversion is mainly divided into two processes: electric-hydrogen conversion and hydrogen-to-methane
191 conversion. There are uncertainties in the process, such as electrolysis efficiency, hydrogen carbon ratio, and reaction temperature in
6

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

192 hydrogen-to-methane conversion. However, existing research considers its conversion efficiency as a fixed value. In equation (17),
193 is considered a fixed value in conventional situations, and this paper proposes an uncertainty model for P2G conversion efficiency
194 based on affine policy. The P2G conversion efficiency can be obtained as

ed
195 (31)
196 Where, and represent the affine center values of the conversion efficiency and conversion efficiency of P2G respectively.
197 represents the affine noise element of P2G efficiency with a value of . is the affine noise element coefficient of P2G

iew
198 efficiency.
199 By substituting (31) into (17), one can derive
200 (32)

201 4 Adaptive penalty ADMM-based optimal framework for integrated electricity-gas system

ev
202 4.1 Decomposition mechanism

203 The optimization model established in Section 3 includes the electricity and gas subsystems, and the information interaction
204 between them is privacy. Therefore, ADMM is used to solve the optimization problem of electricity gas integrated energy system.
205 The decomposition mechanism of the integrated electricity and natural gas system is shown in Fig. 2.

r
The
The
natural
power
system
er gas
system
pe
The
The
natural
power
gas
system
system

206
ot

207 Fig.2 Decomposition mechanism of distributed optimization

208 As shown, the integrated energy system is divided into the power system and the natural gas system representing the interests of
209 different subjects. The two subsystems realize the overall operation of the system through the interactive coupling of variable
210 information. and represent the transmission variable and feedback variable of the power system respectively. The
tn

211 transmission variable is the electricity to gas output of the power system to the natural gas system, and the feedback variable is the
212 electricity to gas output required by the natural gas system. and represent the transmission variable and feedback
213 variable of the natural gas system, in which the transmission variable is the amount of natural gas supplied by the natural gas
214 system to the power system, and the feedback variable is the amount of natural gas required by the power system. The two
215 subsystems are decomposed according to the information interaction of the two sets of coupling variables, and the coupling
rin

216 consistency constraint of equation (33) is satisfied.

217 (33)
ep

218 4.2 Distributed solution framework based on adaptive penalty ADMM

219 In order to ensure that the decomposed power system and natural gas system can meet the consistency constraint equation (33) as
220 much as possible, the core idea of ADMM is to add the relaxed coupling consistency constraint as a penalty term to the objective
221 function of each subsystem[35].
Pr

222 1) Distributed optimization model of the power system: The distributed optimization model of the power system is shown in
223 equations (34) and (35), where equation (34) is the optimization objective function after adding the consistency constraint penalty
224 term.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

ed
225 (34)

iew
226 s.t. formula: (11) - (22), (32) (35)
227 where, and are Lagrange multipliers, and and are penalty factors.

ev
228 2) Distributed optimization model of the natural gas system: The distributed optimization model of natural gas system is shown
229 in equations (36) and (37), where equation (36) is the optimization objective function after adding consistency constraint penalty.

r
230
er (36)
pe

231 s.t. formula: (24) - (30) (37)

ot

232 To sum up, a distributed optimization model based on ADMM decomposition is obtained, and the risk cost of system satisfies
233 equation (38).
234 (38)
tn

235 3) Solution framework: In the above-mentioned distributed optimization model of the integrated electricity and natural gas system,
236 the nonconvex nonlinearity of distflow power flow constraint and pipeline Weymouth equation constraint increases the difficulty of
237 solving the model. Second order cone relaxation is applied to equations (13) [36]and incremental linearization is applied to equation
238 (25), than the mixed integer nonconvex nonlinear programming problem is transformed into a mixed integer linear programming
239 problem[37]. The penalty factor of standard ADMM is a fixed value, and its value will have a great impact on the iterative process.
240 The original residual of the standard ADMM is related to the solvability of the original problem, and the dual residual is related to the
rin

241 convergence of the original problem. As the number of iterations increases, the original and dual residuals will also change, and the
242 fixed penalty factor may cause the solution to be slow or not convergent. The adaptive penalty ADMM updates the penalty factor
243 according to the size relationship between the original residual and the dual residual: when the original residual is large, the penalty
244 factor can be increased to increase the weight of the penalty term and promote the consistency constraint to be satisfied; When the
245 dual residual is large, the penalty factor can be reduced to reduce the oscillation of the original objective function and promote the
ep

246 convergence of the solution.

247 Therefore, a distributed solution framework based on adaptive penalty ADMM is proposed. See algorithm 1.
248
249
250
Pr

251
Algorithm 1：Distributed solution framework based on adaptive penalty ADMM

1.Initialize settings. Preset the subsystem equipment and topology parameters. Set the convergence

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

margins and of the original residual and dual residual. Set the initial values of the coupling
variables, Lagrange multipliers and penalty factors, and set the maximum number of iterations.
2.Solve the optimization model of power system (34) - (35) and the optimization model of natural gas

ed
system optimization model (36) - (37) respectively to obtain the values of coupling variables ,
, , , , , , .
3.Calculate the original residual and dual residual according to equations (39) - (40), and update the
Lagrangian multipliers , and penalty factors , aaa according to equations (42) - (43).

iew
4.Check whether the residual meets the convergence according to equation (41), and update the number
of iterations. If the convergence condition is met or the maximum number of iterations is reached,
the next step is executed; otherwise, step 2 is returned.
5.Output the final results.
252 The original residual and dual residual are calculated as follows.
(39)

ev
253

254 (40)

r
255 The convergence condition is calculated as follows.

er
pe
256 (41)

257 where, is defined as the residual flag variable whose original residual and dual residual exceed the convergence margin, and its
ot

258 value represents the number of residual excesses. When the residual out of limit flag is less than 1, it means that all residuals meet
259 the convergence margin.
260 The Lagrange multiplier is updated according to equation (42), and the penalty factor of ADMM is adaptively updated
261 according to equation (43).
tn

262 (42)
rin
ep

263
(43)
Pr

264 Fig. 3 is a flow chart for the distributed optimization solution of an electric-gas coupled gamete network considering the
265 uncertainty of the demand side.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

Start

ed
Initialize system parameters, preset coupling variables and initial
multiplier values, and set the maximum iteration number K

k=0

iew
Optimal operation of Optimal operation
power system model of natural gas
(34)~(35) system(36)~(37)

Exchange coupling variable information and

Update multipliers
calculate the original and dual residuals accordinto

ev
according to equations
equations (39)~(40)
(42)~(43)

No k=k+1

r
Does the result satisfy the convergence condition (41) or
reach the maximum iteration number K?

er Yes

Output final results

pe
End
266
267 Fig. 3 Flow chart of the optimization algorithm

268 5 Numerical results

269 5.1 Testing System

ot

270 An integrated electricity and natural gas system as shown in Fig. 4 includes an IEEE-33 bus distribution system and a 24-node
271 natural gas systems[38]. The system consists of two gas turbines GT1 and GT2, one wind turbine WT and one photovoltaic unit PT,
272 two P2G facilities P2G1 and P2G2, and two gas storage tanks GS1 and GS2. The electricity load and natural gas loads are shown in
273 Fig. 5. The electricity price and natural gas price are from [25]. The parameters of gas turbine and gas storage tank is assumed as the
tn

274 same[27], and the capacities of Wind-PV units are 1.6MW respectively. See Fig. 6 for typical scenarios of obtaining 10 groups of
275 Wind-PV output.

19 20 21 22
WT
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
rin

Main grid PT
23 24 25 27 28 29 30 31 32 33
26 Wind power
GS2

GT2 P2G2 P2G1 Photovoltaic

GT1

17 18 19 22 23 24 GS1 Gas turbine

16 20 21
ep

Power to gas
2 3 4 5 6 7 8 9 10 11 Gas storage
tank
Natural gas well 12 13 14 15
276
277 Fig. 4 Test system topology
Pr

10

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

Active power demand

Natural gas demand
7

ed
6

Power demand/MW
5

iew
3

2
0 4 8 12 16 20 24

278 Time/h
279 Fig. 5 The power demand and gas demand

ev
S(1) S(6)
1.0
S(2) S(7)
S(3) S(8)
S(4) S(9)
0.8 S(5) S(10)
Wind power output/p.u.

r
0.6

0.4

0.2

0.0
er
pe
0 4 8 12 16 20 24

280 Time/h
281 (a) Typical scenarios of wind power output
S(1)
1.0
S(2)
S(3)
S(4)
0.8
Photovoltaic output/p.u.

S(5)
S(6)
ot

S(7)
0.6
S(8)
S(9)
S(10)
0.4
tn

0.2

0.0
0 4 8 12 16 20 24

282 Time/h
283 (b) Typical PV output scenarios
rin

284 Fig. 6 Wind-PV output curves

285 5.2 Impact analysis of Wind-PV uncertainty
286 In order to study the impact of Wind-PV uncertainty on the operation of integrated energy system using CVaR, the following two
287 cases are set for comparative analysis:
ep

288 Case 1) Traditional optimal operation under the determination of Wind-PV.

289 Case 2) the confidence level is 0.9, the risk preference coefficient is 0.5, and the CVaR is used to consider the optimal operation of
290 the uncertainty of the Wind-PV.
291 Fig. 7 and Fig. 8 show the power purchase and gas turbine output curves in two cases.
292 It can be seen from the figure that at the time of high wind power output before 8:00, the power system demand is preferentially
293 supplied by wind power. In addition, the electricity price in this period is lower, so the remaining demand is borne by the power
Pr

294 purchased. The output of the gas turbine is 0, and the dispatching results of the two cases are the same. At 10:00-13:00, the wind
295 power output decreases sharply and the photovoltaic output increases sharply, so the uncertainty of Wind-PV output increases. In case
296 1, the wind power output gives priority to the supply of power demand. At this time, the electricity price is in the flat section and the
297 gas price is in the peak section. The residual power demand is preferentially borne by the purchased power. At 13:00-18:00, the

11

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

298 electricity price is high and the gas price is low. The gas turbine output and the main grid share the load demand other than the
299 Wind-PV output. After 20:00, Wind-PV output supply part of the electrical load.
4.0 180
Power price($/MW.h)

ed
3.6
Case1 165
3.2 Case2
150

Power price/($/MW.h)
2.8
135
2.4

Power/MW
120

iew
2.0
105
1.6

1.2 90

0.8 75

0.4
60
0.0
0 4 8 12 16 20 24

ev
300 Time/h
301 Fig. 7 Power purchase of case 1 and case 2
65
2.4 Case1
2.2 Case2
60
2.0 Gas price
1.8

r
55

Gas price/($/MW.h)
1.6
1.4
Power/MW

50
1.2
1.0
0.8
0.6
0.4
0.2
er 45

40
pe
35
0.0
-0.2 30
0 4 8 12 16 20 24
302 Time/h
303 Fig. 8 Gas turbine output of case 1 and case 2
304 Table 2
305 Operating cost results of case 1 and case 2
ot

Case1 Case2
Power purchase cost/$ 4245.54 4322.15
Operating cost of gas turbine / $ 2221.85 2353.23
Abandoned Wind-PV cost / $ 90.62 276.62
Power load shedding cost/$ 242.31 0
tn

Income of P2G/$ 24.15

Cost of power system/$ 7449.85 7523.08
Cost of natural gas system/$ 3848.77 3718.62
The total cost/$ 11435.54 11241.69
VaR/$
CVaR/$ - 278.31
rin

306 Table 2 shows the cost comparison of the two cases. It shows that when uncertainty is not taken into account, the load will be
307 supplied by the output of the Wind-PV. Therefore, the cost of abandoned Wind-PV in case 1 is slightly 186 $ less than that in case 2,
308 but the adjustable margin of case 1 is small, and there will be a penalty of 242.3 $ for power load shedding, which will affect the user
309 experience and power quality. In addition, the total cost of case 1 is 56.9 $ higher than that of case 2, and in the actual operation,
310 because of the actual uncertainty of the Wind-PV, the operation cost is further increased, resulting in economic and security losses. In
ep

311 contrast, CVaR is used in case 2 to evaluate the cost risk caused by the uncertainty, which provides a reference for the scheduling
312 decision. It has an appropriate amount of adjustable margin while ensuring to reduce the risk cost, and has a strong ability to deal with
313 the uncertainty.
314 Fig. 9 shows the power load shedding of each node in Case 1. The load shedding time in the figure is mainly concentrated at 8:00-
315 12:00. This period is the time when Wind-PV are highly volatile. If the uncertainty is not considered, the load shedding penalty cost
Pr

316 of 242.31 $ will be generated. On the contrary, if the uncertainty is considered and CVaR is used to measure uncertainty risk, load
317 shedding events will be avoided in decision-making.

12

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

ed
iew
ev
318
319 Fig. 9 Power load shedding at each node on case 1

320 5.3 Analysis of the influence of the P2G conversion uncertainty

r
321 To analyze the influence of the uncertainty in the process of electricity to gas conversion on the operation economy, the affine
322 policy is introduced to establish the uncertainty of P2G efficiency. On the basis of case 2, the affine center value of electricity to gas
323
324
325
326
327
er
conversion efficiency was set to 0.65, and the step length of the uncertainty coefficient was set to 0.03. Five groups of simulation
experiments were carried out to study the impact of the change of P2G efficiency within [0.5,0.8] on the optimal operation.
In Table 3, with the gradual increase of uncertainty, the income from P2G is lower. The stability of P2G equipment's conversion
efficiency has a great impact on its economy. This has an impact on the gas purchase cost of the natural gas system, mainly because
with the increase of uncertainty, the natural gas production of P2G becomes more uncertain and inefficient, and the natural gas system
pe
328 will instead purchase natural gas from natural gas suppliers at a higher price than P2G. At the same time, because the remaining Wind-
329 PV output were consumed by P2G, so that Wind-PV output that should be discarded can be used, no matter how uncertain the P2G
330 conversion efficiency is, it will not affect the abandoned Wind-PV. However, the efficiency of P2G should not be too random. This
331 is because the cost of P2G is still high and the economy is not obvious. The steady increase of gas production of P2G to obtain
332 economic benefits is conducive to the promotion and use of P2G.
333 Table 3
334
ot

The costs under different uncertainty coefficients

Abandoned Income Gas
Uncertainty The total
Wind-PV of purchase
coefficient cost/$
cost / $ P2G/$ cost /$
0 10826.46 276.62 20.03 5918.46
tn

0.03 11244.77 276.62 19.77 5918.46

0.06 11246 276.62 19.49 5920.92
0.09 11247.23 276.62 19.17 5922.15
0.12 11249.69 276.62 18.8 5923.23
335 Fig. 10 shows the natural gas production of P2G under different uncertainty coefficients. It can be seen that with the gradual increase
rin

336 of uncertainty, the natural gas production has a great probability of decreasing, which is not conducive to reducing the operating cost
337 of P2G. Therefore, in order to improve the economic benefits of P2G, conversion efficiency should be improved and its uncertainty
338 should be reduced.
7 0
0.03
6 0.06
ep

0.09
Gas production/m3

5 0.12

2
Pr

0
0 4 8 12 16 20 24
339 Tim/h
340 Fig. 10 The gas production of P2G under different uncertainty coefficients
13

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

341 5.4 Analysis of influencing factors of optimized operation

342 1) Influence of risk coefficient on optimal operation:
343 In Table 4, with the gradual increase of risk preference coefficient, the cost of power system is increasing, while the cost of natural

ed
344 gas system decreases first and then increases, and the cost of power system is more sensitive to the change of risk preference
345 coefficient. At the same time, it can be seen that the cost of abandoned Wind-PV is decreasing, because the increase of risk coefficient
346 makes the system consume renewable energy as much as possible. Decision makers tend to be conservative and allocate more power
347 reserves, thus increasing the cost of the power system. In addition, the cost of abandoned wind and CVaR are decreasing, which
348 indicates that the risk cost of the system is decreasing. In addition, the cost of the natural gas system changes with the risk coefficient,

iew
349 which indicates that the uncertainty will affect the natural gas network side, and the cost of the natural gas system will increase if the
350 risk coefficient is too large or too small. Meanwhile, with the gradual increase of the risk preference coefficient, the total cost is also
351 increasing, while the cost of abandoned Wind-PV is decreasing. This is because the increase of the risk coefficient makes the
352 system consume renewable energy as much as possible. The decision-makers tend to be conservative and allocate more power
353 reserves, thus increasing the total cost. Moreover, the cost of abandoned Wind-PV and CVaR are also decreasing, which also
354 shows that the risk cost of the system is decreasing. In addition, there is a certain difference between VaR and CVaR, which means
355 the tail risk loss under a certain confidence level. The CVaR is closer to the actual risk cost, providing a more reasonable risk

ev
356 assessment basis for the day ahead scheduling decision of the integrated energy system.
357 Table 4
358 The costs under different risk preferences
Risk Abandoned
The total
preference Wind-PV VaR/$ CVaR/$
cost/$

r
coefficient cost / $
0.1 11116.15 310.15 310 313.54
0.2 11150.46 301.54 301.08 306.62
0.3 11184.62 293.23 292.77 296.46
0.4
0.5
0.6
0.7
11217.54
11241.69
11266.62
11287.69
284.92
276.62
268.46
260.15
er 284.77
276.46
267.54
260.92
287.07
278.31
277.85
271.23
pe
0.8 11303.08 251.85 251.54 255.85
359 The value of the risk preference coefficient is determined by the system decision-maker's preference attitude towards risk cost.
360 From Fig. 11, it can be seen that the ratio of risk cost to total cost is approximately [2.3%, 2.8%]. As the risk coefficient increases, the
361 risk cost gradually decreases and the total system cost gradually increases. This indicates that when decision-makers prefer the lowest
362 total cost, they will face a gradually increasing risk cost, and when they prefer the lowest risk cost, they will face a significant loss in
363 the total system cost. When the risk coefficient is set to 0.8, it indicates that the decision-maker bears the least risk and the system
364 avoids risk more effectively. When the risk coefficient is set to 0.1, although the total cost is the smallest, there is a lack of consideration
ot

365 for risk factors at this time. The most abandoned Wind-PV faces the greatest risk, and decision-makers pursue a reduction in the total
366 cost at greater operational risk. In addition, due to considering uncertainty and avoiding load shedding events, the risk mainly manifests
367 in the cost of abandoning wind and solar energy, and the cost of abandoned Wind-PV and risk change in the same direction as the risk
368 coefficient. In real-time scheduling, decision-makers need to weigh risks and costs to determine the final operational plan.
tn

The total cost 320

Abandoned wind power / photovoltaic cost/$

11320 Abandoned wind power / photovoltaic cost

0.8
0.1
0.7
0.2
11280 0.6 300
0.3
The total cost/$

0.5
11240
rin

0.4
0.5 0.4
280
11200 0.3
0.6
11160 0.7 0.2
260
0.8 0.1
11120
ep

250 260 270 280 290 300 310 320

369 CVaR/$
370 Fig. 11 Relationship between total cost and risk cost curve
371 In Fig. 12, with the gradual increase of the risk preference coefficient, the power purchase and gas purchase of the system show the
372 same change trend, which are increased first and then decreased, and there is the highest value point of power or gas purchase. This
Pr

373 is different from the positive change law of each subsystem with the risk preference coefficient. Therefore, the decision-maker can
374 also determine the specific operation scheme by weighing the risk cost, power or gas purchase cost, operation cost of each subsystem
375 and the integrated system.

14

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

4324
Power purchase cost
Gas purchase cost 5920

4322

ed
Power purchase cost/$

Gas purchase cost/$

5915
4320

4318

iew
5910

4316

5905
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Risk preference coefficient
376
377 Fig. 12 Relation curve between energy purchase cost and risk coefficient

ev
378 2) Influence of confidence level on optimal operation:
379 In order to study the influence of confidence level on simulation results, four groups of experiments with different confidence
380 levels were set up.
381 Table 5
382 The costs under different confidence levels

r
Abandoned
Confidence The total
Wind-PV VaR/$ CVaR/$
level cost/$
cost / $
0.9 11241.85 276.62 276.46 278.31
0.85
0.8
0.75
0.7
11272.31
11302.15
11333.38
11359.08
er
297.38
318.77
341.54
368.31
296.62
319.38
340.46
367.23
301.85
321.69
345.38
371.08
pe
0.65 11409.38 400 399.69 400.62
0.6 11453.85 436.15 435.85 437.08
383 It can be seen from Table 5 that with the increase of confidence level, the cost of each system and CVaR are increasing, which is
384 different from the impact of the increase of risk preference coefficient in Table 4. It shows that the greater the confidence level is, the
385 greater the risk the system faces, and the total cost of the system is also increasing. This is because the potential risk borne by the
386 system gradually moves to the tail of the normal distribution, resulting in the corresponding cost expectation of this part gradually
387 increasing.
ot

388 It can be seen from Table 5 that with the increase of confidence level, the total cost and CVaR are decreasing, which is different
389 from the impact of the increase of risk preference coefficient in Table 4. It shows that the greater the confidence level, the smaller the
390 risk faced by the system, and the lower the total cost of the system. This is because the potential risk borne by the system decreases
391 with the increase of the confidence level, that is, the cost of abandoned Wind-PV decreases, which increases the consumption rate of
tn

392 Wind-PV and reduces the cost of energy purchase, so the total cost also decreases.

393 5.5 Performance analysis of distributed algorithm

394 To analyze the effectiveness and advantages of the proposed methods, this section sets the following two cases for comparative
395 analysis:
rin

396 Case 3) the confidence level is 0.9, the risk preference coefficient is 0.5, and the distributed optimization operation based on the
397 standard ADMM;
398 Case 4) the confidence level is 0.9, the risk preference coefficient is 0.5, and the distributed optimization operation based on adaptive
399 penalty ADMM. Set the initial values of the Lagrange multiplier and penalty factor as 1.
×102
14
sign of out-of-limit residual error
ep

Case 3 20 Case 3
Case 4 Case4
12
15
10
Time/h

10
8
5
Pr

4 0

1 2 1 2
Iterations Iterations
400
401 Fig. 13 Calculation time and sign of out-of-limit residual error when the convergence margin is 0.1

15

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

×102

Sign of out-of-limit residual error

18 Case 3 Case 3
24
Case 4 Case 4
15

ed
18
12

Time/h
12
9
6
6

iew
3 0
1 2 1 2
Iterations Iterations
402
403 Fig. 14 Calculation time and sign of out-of-limit residual error when the convergence margin is 0.01
×102
240 36

Sign of out-of-limit residual error

Case 3 Case 3
Case 4 Case 4
200 30

160 24

ev
Time/h

120 18

80 12

40 6

0 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14

r
Iterations Iterations
404
405 Fig. 15 Calculation time and sign of out-of-limit residual error when the convergence margin is 0.001
406 Fig. 13, Fig. 14, and Fig. 15 show the calculation time and the sign of out-of-limit residual error of the two cases when the
407
408
409
410
er
convergence margin is 0.1, 0.01 and 0.001 respectively. It can be seen from Fig. 13 and Fig. 14 that when the accuracy is low, the two
algorithms have no difference in the number of iterations, but the calculation time will be different. The calculation time of adaptive
penalty ADMM is shorter and the solution speed is faster. When the convergence margin is 0.001, the calculation accuracy is further
improved. As shown in Fig. 15, the adaptive penalty ADMM can still be solved in a short number of iterations and calculation time,
pe
411 but the standard ADMM cannot meet this requirement. When the two algorithms are solved for 13 times in the same iteration, the
412 adaptive penalty ADMM can meet the convergence requirements and the calculation speed is faster in the second and 13th iteration,
413 while the standard ADMM cannot meet the convergence requirements, which is more obvious when the accuracy is further improved,
414 and even will not converge. Therefore, it can be concluded that the method proposed in this paper has better convergence and
415 advantages.

416 6 Conclusions
ot

417 In this paper, a risk-averse optimal operation model and method considering uncertainty are investigated for the integrated
418 electricity and natural gas system with Wind-PV.
419 (1) An optimized operation model of integrated electricity and natural gas system considering the uncertainty of Wind-PV based
tn

420 on CVaR is proposed, which can effectively measure the loss of abandoned Wind-PV, avoid the occurrence of load shedding
421 events, and balance the operation cost and operation risk of the integrated energy system;
422 (2) Based on CVaR theory, the system operation risk caused by uncertain factors can be measured, and the decision-maker can
423 determine the optimization operation plan according to the risk preference and risk tolerance;
424 (3) A large number of simulation tests and analysis indicate that the proposed method is faster in speed and better in convergence,
rin

425 in comparison with the classical ADMM algorithm.

426 Note that the adaptive penalty ADMM distributed optimization framework established in this paper protects the data information
427 of each subject, which is more suitable for the actual scheduling of integrated electricity and natural gas system, and also provides
428 a new reference for privacy protection in optimization operation.
ep

429 REFERENCES
430 [1] Xiang Y, Guo YT, Wu G, et al. Low-carbon economic planning of integrated electricity-gas energy systems. Energy,2022;249.
431 [2] Zhang Q, Wang ZS, Shi YP, et al. The Optimal Dispatch with Combination of Wind Power and Photovoltaic Power Systems. Energy
432 Procedia, 2016; 103:94-99.
433 [3] Tan W, Tan QB, Zhang Amin, et al. A Bi-level optimization model of integrated energy system considering wind power uncertainty.
Pr

434 Renewable Energy, 2023;202:973-991.

435 [4] Bo YL, Xia YH, Wei W, et al. Hyperfine optimal dispatch for integrated energy microgrid considering uncertainty. Applied Energy,2023;334.
436 [5] Liu Baoding. Uncertain programming with applications. Tsinghua University Press, 2003; 277-278.
437 [6] Zhang JT, Cheng CT, Shen JJ, et al. Short-term joint optimal operation method for high proportion renewable energy grid considering wind-
438 solar uncertainty. Proceedings of the CSEE,2020;5921-5932.

16

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653
 > Renewable Energy<

439 [7] Yang L, Xu Y, Gu W, et al. Distributionally Robust Chance-constrained Optimal Power-Gas Flow under Bidirectional Interactions
440 Considering Uncertain Wind Power. IEEE Transactions on Smart Grid, 2021;12(2):1722-1735.
441 [8] Li ZW, Zhao SQ, Liu JS. Optimal Scheduling of Power System Based on Dependent-chance Goal Programming. Proceedings of the CSEE
442 ,2019, 39(10):2803-2816.

ed
443 [9] Yang JC, Zhai XH, Tan ZF, et al. Day-ahead Bidding Optimization for High-uncertainty Units Based on Relatively Robust Conditional Value
444 at Risk. Power System Technology,2021 ,45(11):4366-4376.
445 [10] Zhang YC, Liu W, Huang ZH, et al. Distributionally robust coordination optimization scheduling for electricity-gas-transportation coupled
446 system considering multiple uncertainties. Renewable Energy, 2021;163:2037-2052.
447 [11] Fambri G, Diaz-Londono C, Mazza A, et al. Techno-economic analysis of Power-to-Gas plants in a gas and electricity distribution network
448

iew
system with high renewable energy penetration. Applied Energy, 2022; 312:118743.
449 [12] Bao ML, Hui HY, Ding Y, et al. An efficient framework for exploiting operational flexibility of load energy hubs in risk management of
450 integrated electricity-gas systems. Applied Energy,2023; 338: 120765.
451 [13] Ye JN, Xie M, Zhang SP, et al. Stochastic optimal scheduling of electricity–hydrogen enriched compressed natural gas urban integrated
452 energy system. Renewable Energy, 2023;211:1024-1044.
453 [14] Zhang B, Hu W, Li J, et al. Dynamic energy conversion and management strategy for an integrated electricity and natural gas system with
454 renewable energy: Deep reinforcement learning approach. Energy Conversion and Management, 2020, 220(6321):113063.
455 [15] Fang X, Cui HT, Yuan HY, et al. Distributionally-Robust Chance Constrained and Interval Optimization for Integrated Electricity and Natural
456

ev
Gas Systems Optimal Power Flow with Wind Uncertainties. Applied Energy, 2019;252:113420.
457 [16] Cui Y, Guo FY, Zhong WZ, et al. Interval Multi-objective Optimal Dispatch of Integrated Energy System Under Multiple Uncertainty
458 Environment. Power System Technology, 2022,46(08):2964-2975.
459 [17] Xuan A, Shen XW, Guo QL, et al. A conditional value-at-risk based planning model for integrated energy system with energy storage and
460 renewables. Applied Energy, 2021;294:116971.
461 [18] Mu YF, Wang CS, Cao Y, et al. A CVaR-based risk assessment method for park-level integrated energy system considering the uncertainties

r
462 and correlation of energy prices. Energy, 2022;247:123549.
463 [19] Yao LY, Liu ZY, Chang WG, et al. Multi-level model predictive control based multi-objective optimal energy management of integrated
464 energy systems considering uncertainty. Renewable Energy, 2023;212:523-537.
465
466
467
468
469
er
[20] Navid R, Yasin P, Amirhossein K. Economic-environmental risk-averse optimal heat and power energy management of a grid-connected
multi microgrid system considering demand response and bidding strategy. Energy, 2022;240:122844.
[21] Yang HM, Liang R, Yuan Y, et al. Distributionally robust optimal dispatch in the power system with high penetration of wind power based
on net load fluctuation data. Applied Energy, 2022;313:118813.
[22] Jabr, R. A. Distributionally Robust CVaR Constraints for Power Flow Optimization. IEEE Transactions on Power Systems, 2020;35(5):3764-
pe
470 3773.
471 [23] Renan ADF, Ederson PV, André LK , et al. Stochastic model to aid decision making on investments in renewable energy generation: Portfolio
472 diffusion and investor risk aversion. Renewable Energy, 2020;162:1161-1171.
473 [24] Li Z , Xu Y , Wu L, et al. A Risk-averse Adaptively Stochastic Method for Multi-Energy Ship Operation under Diverse Uncertainties. IEEE
474 Transactions on Power Systems, 2020;33(3):2149-2161.
475 [25] Duan, JD, Liu F,Yang Y. Optimal operation for integrated electricity and natural gas systems considering demand response uncertainties.
476 Applied Energy, 2021;323:119455.
477 [26] Yang T , Yi X , Wu J , et al. A survey of distributed optimization. Annual Reviews in Control, 2019, 47:278-305.
ot

478 [27] Duan, JD, Yang Y, Liu F. Distributed optimization of integrated electricity-natural gas distribution networks considering wind power
479 uncertainties. International Journal of Electrical Power & Energy Systems, 2022;135:107460.
480 [28] Wang YF, Wang XL, Shao CC, et al. Distributed energy trading for an integrated energy system and electric vehicle charging stations: A
481 Nash bargaining game approach. Renewable Energy, 2021;155:513-530.
tn

482 [29] Li Y, Wang CL, Li GQ, et al. Optimal Scheduling of Integrated Demand Response-Enabled Integrated Energy Systems with Uncertain
483 Renewable Generations: A Stackelberg Game Approach. Energy Conversion and Management, 2021;235:113996.
484 [30] Li ZN, Su S, Jin XL, et al. Distributed energy management for active distribution network considering aggregated office buildings. Renewable
485 Energy, 2021;180:1073-1087.
486 [31] Mohamed AM, Hossein C, Emad MA, et al. Stochastic and distributed scheduling of shipboard power systems using MθFOA-ADMM.
487 Energy, 2020;206:118041.
rin

488 [32] Kou Y, Bie ZH, Wang X, et al. ADMM-Based Multiperiod Optimal Energy Flow of A Regional Integrated Multi-Energy Microgrid. Energy
489 Procedia, 2019;159:180-185.
490 [33] Jiang T, Wu CH, Zhang RF, et al. Risk-averse TSO-DSOs coordinated distributed dispatching considering renewable energy and demand
491 response uncertainties. Applied Energy, 2022;327:120024.
492 [34] Chen J, Zhang W, Zhang Y, et al. Day-ahead Scheduling of Distribution Level Integrated Electricity and Natural Gas System Based on Fast-
493 ADMM with Restart Algorithm. IEEE Access, 2018;6:17557-17569.
ep

494 [35]Lan P, Shen XD, Wu G, et al. Distributed Optimal Scheduling for Transmission-Distribution-Natural-Gas System Based on Alternating
495 Direction Method of Multipliers. Automation of Electric Power Systems, 2021,45(23):21-30.
496 [36] Rémy RM, Vannak V. An iterative linear DistFLow for dynamic optimization in distributed generation planning studies. International Journal
497 of Electrical Power & Energy Systems, 2022; 138:107936.
498 [37] Yang L, Zhao X, Xu YL. A Convex Optimization and Iterative Solution Based Method for Optimal Power-Gas Flow Considering Power and
499 Gas Losses. International Journal of Electrical Power & Energy Systems, 2020,121:106023.
Pr

500 [38] Jiang YB, Xu J, Sun YZ, et al. Coordinated operation of gas-electricity integrated distribution system with multi-CCHP and distributed
501 renewable energy sources. Applied energy, 2018; 211:237-248.

17

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4513653