Scheduling of Energy Management Based On Battery Logistics in Pelagic Islanded Microgrid Clusters

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Electrical Power and Energy Systems 127 (2021) 106573

Contents lists available at ScienceDirect

International Journal of Electrical Power and Energy Systems


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

Scheduling of energy management based on battery logistics in pelagic


islanded microgrid clusters
Chuantao Wu, Quan Sui, Xiangning Lin, Zhixun Wang, Zhengtian Li *
State Key Laboratory of Advanced Electromagnetic Engineering and Technology, Huazhong University of Science and Technology, Wuhan, Hubei 430074, China

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

Keywords: According to resource distribution, the pelagic islands can be divided into the load center island and resource-
Energy management system rich islands. Since the islanded microgrids are geographically isolated, an economical and feasible way of en­
Pelagic islanded microgrid clusters ergy flow among islanded microgrids is urgently needed. This paper proposes a novel energy management system
Battery logistics
(EMS) based on battery logistics by electric vessel to optimize the operation of pelagic islanded microgrid clusters
Information gap decision theory
Generalized Benders method
(PIMGC). Firstly, the battery logistics among islanded microgrids and energy flow within islanded microgrids are
modeled. Furthermore, an information gap decision theory-based EMS model is established with the un­
certainties of renewable energy generators. Then, the Generalized Benders method is presented to decompose the
scheduling problem into a master problem of discrete battery logistics optimization and subproblems of power
flow feasibility checking. Finally, case studies with the proposed EMS are conducted with three islands in the
South China Sea. The simulation results show that the proposed EMS is feasible and economical to supply energy
for PIMGCs, providing a reference for the energy supply system of pelagic islands.

predictive control approach for islanded AC microgrid is presented. [16]


The main notations of this paper are listed below; other symbols are proposes a novel decentralized periodic energy trading framework for
defined as required. the pelagic islanded microgrid. The above works focus on the optimal
operation of the single islanded microgrid.
1. Introduction However, due to the small scale of the pelagic islands, it is hard to
meet the land requirements of electricity producers and consumers
The pelagic islands are far from the mainland, and therefore, most simultaneously. According to resource distribution, the pelagic islands
rely on diesel generator (DG) or gas turbine as an independent power could be divided into two types: load center island (LCI) and resource-
supply [1]. Yet, the traditional power supply solely on fossil energy has rich islands (RRIs). Each island has an islanded microgrid and all
various problems, including excessive fuel costs, serious environmental microgrids are combined into the pelagic islanded microgrid cluster
pollution, and poor power supply reliability [2–3]. These restrictions (PIMGC). There are most electricity consumers and DG as backup power
have become stumbling blocks to the economic and sustainable devel­ on the LCI, while the renewable energy generators (REG), e.g. wind
opment of pelagic islands. Therefore, it is crucial to develop the pelagic turbines (WT) and photovoltaics (PV), are equipped on the RRIs. Owing
islanded microgrids with renewable resources [3–7]. to high investment costs and the difficulty of laying, it is not economical
More recently, the energy management system (EMS) in islanded and wise to connect the islanded microgrids by submarine cables.
microgrids has received notable attention. In [8], a scenario-based sto­ Another economic and feasible way of energy flow among microgrids is
chastic multi-energy microgrid investment planning model is presented. proposed in [17], which is to swap battery transported by electric vessel
[9] proposes a 100% renewable energy supply system based on wind (EV). The batteries are fully charged on the RRIs and swapped to LCI
turbines, concentrating solar power plants, and desalination units. In through EV, realizing the energy flows from RRIs to LCI. In other words,
[10], the multi-objective optimization of EMS is presented. The sto­ the energy flow among islanded microgrids is realized through battery
chastic optimization framework for EMS of isolated microgrids is pro­ logistics. Therefore, the EMS for the single islanded microgrid is not
posed in [11–12]. [13] proposes a two-stage robust-based EMS for suitable for the PIMGC.
isolated microgrids. An EMS with an improved genetic algorithm is As a further study, the researches aiming at energy management in
proposed in [14]. In [15], a decentralized EMS based on model

* Corresponding author.
E-mail address: 453874933@qq.com (Z. Li).

https://doi.org/10.1016/j.ijepes.2020.106573
Received 8 May 2020; Received in revised form 16 August 2020; Accepted 6 October 2020
Available online 7 December 2020
0142-0615/© 2020 Elsevier Ltd. All rights reserved.
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

reg
Nomenclature Pn,t Power of REG on the RRI
cg1 /cg2 Fuel consumption coefficients of DG g cg1 = 0.25,cg2 =
Acronyms 0.08
EMS Energy management system dg Price of diesel in the PCIs
PIMGC Pelagic islanded microgrid clusters cb Charging/discharging unit costs of battery
LCI Load center island cev Charging/discharging unit costs of EV
RRI Resource-rich island β Settled cost deviation factor
EV Electric vessel
REG Renewable energy generator Variables
IGDT Information gap decision theory uev
i,t State variable of EV’s location
DG Diesel generator sev
i,t State variable of leaving BMS for EV
WT Wind turbine dev
i,t State variable of reaching BMS for EV
PV Photovoltaics uli,t State variable of batteries’ location
MMGs Multi-microgrids
sli,t State variable of leaving BMS for battery i
BMS Battery management station
dli,t State variable of reaching BMS for battery i
Sets and Indices SOCev State of charge of EV at time t
t
t∈T Time periods
SOClt State of charge of battery l at time t
m∈M BMS on the LCI
Pdev,t Discharging power of EV at time t
n∈N BMS on the RRIs
k∈K All BMS in PCIs,K = N ∪ M Pcev,i,t Charging power of EV at BMS i at time t
l∈L Batteries in PCIs Pcl,i,t /Pdl,i,t Active charging/discharging power of the battery l at BMS i
w∈W Bus nodes of the microgrid on the LCI at time t
g∈G DGs of the microgrid on the LCI Qcl,i,t /Qdl,i,t Reactive charging/discharging power of battery l at BMS i
Parameters at time t
( ) Pg,t /Qg,t Active/reactive power of DG g at time t
ci Location of BMS i,ci = xi , yi
dij Distance between BMS i and j Pcm,t /Pdm,t Active charging/discharging power of BMS m at time t
τij Sailing time between BMS i and j Qcm,t /Qdm,t Reactive charging/discharging power of BMS m at time t
vev Sailing speed of EV Vi,t Voltage amplitude at bus i at time t
tins Installment time of the battery θi,j,t The phase difference between bus i and j at time t
Eev Energy storage capacity of EV ug,t Operating status of DG g at time t
Pcm
ev /P dm
ev Rated charging/discharging (sailing) power of EV f Operating cost of MSEMS
Eb Energy storage capacity of the battery ϕ Decision variable
Pcm dm Uncertain parameter
b /Pb Rated charging/discharging power of the battery ψ
Sb The apparent power of the battery α Robustness coefficient
SOCmin Minimum state of charge λk Weight of parameter k
SOCmax Maximum state of charge Pm,t Active power of BMS m at time t
PDi,t /QDi,t Active/reactive power at bus i at time t Ploss
t Line loss power at time t
Gij /Bij Conductivity/admittance of line ij g,t /εm,t
ε+/− Positive/negative power deviation factors of DG g and
+/−

Vmin /Vmax Minimum/maximum amplitude of bus voltage BMS m at time t


Pmax
g Maximum power of DG g λg,t /λm,t
+/− +/−
Lagrange multipliers of Eq. (40)
Pmin
g The minimum power of DG g,Pmin
g = 0.3Pmax
g
Pmax
ij The maximum transmission power of line ij

multi-microgrids (MMGs) has been the hot spot recently. There are two connected by battery logistics. Because the energy flow through cables is
ways to transmit energy among microgrids in the existing studies: power continuous and real-time, while discrete and delayed in the battery
cable and battery logistics. In terms of EMS in MMGs connected by ca­ logistics.
bles, a bi-level EMS, and a pricing model based on microgrid marginal Some related studies about the EMS in MMGs connected by battery
pricing in MMGs are presented in [18]. A double-layer framework of logistics have also been carried out. In [32], a scenario-based day-ahead
energy transactions based on blockchain in MMGs is proposed in [19]. In EMS based on EV in pelagic islands is proposed. A novel day-ahead EMS
[20], a tri-layer multi-agent system architecture for MMGs is presented. based on EV considering noninteger-hour energy flow is subsequently
The stochastic programming, robust optimization, and distributionally presented in [33]. The EV is integrated with a large-scale battery in the
robust optimization for EMS in MMGs are proposed in [21–24]. An above studies, while the EV in this paper is just used as a vehicle for
intelligent EMS in MMGs is proposed based on deep reinforcement logistics and can transport all batteries in PIMGCs to improve flow ef­
learning in [25]. In [26–27], different real-time market trading mech­ ficiency. In [34], a novel multi-energy management framework for
anisms are proposed. [28–29] design the peer-to-peer trading frame­ PIMGCs based on battery and gas tank swapping is presented. However,
works for MMGs. In [30], a coalitional operation model based on the multi-energy logistics among microgrids is realized once a day, and
cooperative game theory for MMGs is presented. A distributed model the time–space characteristics of logistics are ignored. In practice, the
predictive control strategy based on the dynamic non-cooperative game battery logistics can occur several times in one day to improve the ef­
theory is proposed in [31]. The above works have done comprehensive ficiency of energy flow among microgrids. Therefore, considering these
researches on the EMS in MMGs. However, it is not suitable for MMGs discrepancies, a novel model of battery logistics on an hourly scale is

2
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

needed.
With the aforementioned observations, the detailed comparisons
with the existing literature are shown in Table 1, which shows that no
research has been done to design an energy management framework for
RRI
the PIGMC based on battery logistics on an hourly scale. Therefore, to fill
the research gap in this area, this paper proposed a novel EMS based on
battery logistics for the PIMGC. Especially, the electricity consumers and
RRI
DG are scattered on the LCI, and REGs are equipped on the RRIs. Firstly,
LCI
the battery logistics among microgrids and the energy flow within
microgrids are modeled. Furthermore, the information gap decision
theory (IGDT) [35–37] is applied to cope with the uncertainties of REGs.
Then, the scheduling problem is solved by the Generalized Benders, WT DG Battery Power line
PV
dividing the original problem into the master problem of discrete energy EV BMS Tower Users Shipping route
flow optimization and subproblems of continuous energy flow optimi­
zation. Case studies are presented to verify the feasibility and economy Fig. 1. An energy management framework for the PIMGC.
of the proposed EMS in the PIMGC. The main contributions of this paper
are summarized as follows: microgrids.

1) Considering the time–space transportation of EV, the battery logis­ 2.1. Model of the battery logistics
tics among microgrids based on battery swapping, and the energy
flow within microgrids based on AC power flow is modeled. As shown in Fig. 2, the graph theory is used to describe the topology
2) An IGDT-based EMS for the PIMGC is first proposed to collabora­ of the traffic network of EV. A graph of the channel network is repre­
tively optimize the battery logistics and energy flow with the un­ sented by G (C , D ), where C and D denote the set of BMS’s location
certainties of REGs. and EV route. The corresponding relationship can be formulated by:
3) The Generalized Benders method is applied to solve efficiently,
decomposing the problem into a master problem of discrete energy 𝒞 = {ci , ∀i ∈ K } = {(xi , yi ), (xi , yi ), ⋯, (xK , yK ) } (1)
flow optimization and subproblems of power flow feasibility ⎧ ⎫
checking. { } ⎨ d11 ⋯ d1K ⎬
𝒟 = dij = ⋮ ⋱ ⋮ , ∀i, j ∈ K (2)
4) Using the realistic demand and pelagic island environment, the ⎩ ⎭
dK1 ⋯ dKK
effectiveness and feasibility of the proposed EMS for the PIMGC are
illustrated. ⃦ ⃦
dij = ⃦ci − cj ⃦2 , ∀i, j ∈ K (3)

The rest of this paper is organized as follows: Section 2 describes the


dij /vev + tins
analytic model of the system and the IGDT-based EMS model. Section 3 τij = ⌈ ⌉, ∀i, j ∈ K (4)
Δt
presents the Generalized Benders method for the scheduling problem.
Case studies are conducted in Section 4. Finally, the conclusion is drawn Eqs. (1)-(3) present the distance between each BMS. Eq. (4) repre­
in Section 5. sents the inter-island sailing time, where EV navigation time and battery
installment time are included. And ⌈⋅⌉ is the ceiling function in Eq. (4).
2. System model The swapping and transportation of batteries depend on the EV,
which means that both are deeply coupled in traffic attributes. The
The structure of the energy management framework for the PIMGC is space–time transportation model can be expressed by:
shown in Fig. 1. As has been presented in the above part, we consider a ∑
uev
i,t ⩽1, ∀t ∈ T (5)
PIMGC consisting of one LCI and multiple RRLs. The LCI and RRIs are i∈K
equipped with a setM ≜{1, 2, ⋯, M} and N ≜{1, 2, ⋯, N} battery man­ ∑
agement stations (BMS) to control the charging/discharging of batteries. uli,t ⩽1, ∀l ∈ L , ∀t ∈ T (6)
Besides, a set L ≜{1, 2, ⋯, L} of batteries are distributed at different BMS i∈K

and transported by EV. The EV is powered by its battery and driven by t+τij
∑ ( )
operators to complete battery transportation. uev ev
(7)
j,k ⩽τij 1 − ui,t , ∀i, j ∈ K ⊖, ∀t ∈ T
The system model monitored by EMS consists of two subsystems: (1) k=t+1
a discrete battery logistics model based on batteries and EV, which
represents the dynamic energy flow among microgrids of LCI and RRIs; sev ev ev ev
i,t − di,t = ui,t − ui,t− 1 , ∀i ∈ K , ∀t ∈ T (8)
and (2) a continuous energy flow model which expresses the real-time
balance between energy supply and demand within the islanded sli,t − di,t
l
= uli,t − uli,t− 1 , ∀i ∈ K ⊖, ∀l ∈ L , ∀t ∈ T (9)

Table 1 sev ev
i,t + di,t ⩽1, ∀i ∈ K , ∀t ∈ T (10)
Comparsion with the existing literatures.
Case [8–16] [18–31] [32–33] [34] This
paper BMS 1
Multi-microgrids No Yes Yes Yes Yes
Connection among – 1 2 3 3
microgrids
Time scale of battery – – Hourly Daily Hourly
logistics
Time-space model – – Yes No Yes
BMS 2 BMS 3
Remarks: 1-Cable; 2-EV integrated with battery; 3-Battery logistics transported
by EV. Fig. 2. Inter-island traffic diagram of EV.

3
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

sli,t + di,t
l
⩽1, ∀i ∈ K ⊖, ∀l ∈ L , ∀t ∈ T (11) − Pmax
ij ⩽Vi,t Vj,t Gij cosθi,j,t +
(25)
Vi,t Vj,t Bij sinθi,j,t ⩽Pmax
ij , ∀i, j ∈ W , ∀t ∈ T
sli,t ⩽sev
i,t , ∀i ∈ K , ∀l ∈ L , ∀t ∈ T (12)
Vmin ⩽Vi,t ⩽Vmax , ∀i ∈ W (26)
l
di,t ev
⩽di,t , ∀i ∈ K , ∀l ∈ L , ∀t ∈ T (13)
ev/l ev/l ev/l
ug,t Pmin max
g ⩽Pg,t ⩽ug,t Pg , ∀g ∈ G , ∀t ∈ T (27)
where ui,t , si,t , di,t are state variables of location, action variables
∑ c/d
of leaving/reaching BMS, respectively. When uev/l ev/l ev/l
i,t (si,t , di,t ) is equal to Pc/d
i,t = Pl,i,t , ∀i ∈ K , ∀t ∈ T (28)
1, EV/battery l stays at (leaves/reaches) BMS i at time t, otherwise not. l∈L

Besides, Eqs. (5)-(6) ensure that EV and batteries can only be located at ∑ c/d
one BMS at time t. Eq. (7) is the transportation time constraint of EV. Qc/d
i,t = Ql,i,t , ∀i ∈ K , ∀t ∈ T (29)
Eqs. (8)-(13) depict the coupling relationship between batteries and EV, l∈L

that is, the transportation of batteries must be consistent with EV. The 2 2
battery l is transported from island i, if and only if the EV departs from (Pc/d c/d 2
l,i,t ) + (Ql,i,t ) ⩽Sb , ∀i ∈ K , ∀l ∈ L , ∀t ∈ T (30)
island i, that is, sev
i,t = 1. The battery l reaches island i, when the EV
reaches island i, that is, dev
i,t = 1. 2.3. IGDT-based EMS model
Besides, EV has the same energy properties as batteries, as described
in Eqs. (14)-(21), where Eqs. (14)-(17) and Eqs. (18)-(21) interpret the Without considering the uncertainties of REGs, the objective func­
constraints of capacity and power, respectively. tion consisting of the fuel consumption costs, charging/discharging costs
Pev ∑ Pev
c,i,t η
of batteries and EV, is formally stated as:
(14)
d,t
SOCtev = SOCt−ev 1 − + , ∀t ∈ T [
Eev η i∈K Eev ∑ ∑
min f = dg (cg,1 Pg,t + cg,2 Pmax
g )+
( ) t∈T g∈G
∑ Pdl,i,t Pcl,i,t η (31)
SOCtl = SOCt−l 1 − − , ∀l ∈ L , t ∈ T (15) ∑∑ ( )
(

)]
Eb η Eb
i∈K cb Pdl,i,t + Pcl,i,t + cev Pev
c,i,t + P ev
d,t
i∈K l∈L i∈K
SOCmin ⩽SOCtev ⩽SOCmax , ∀t ∈ T (16)
To sum up, the proposed deterministic scheduling model takes the
SOCmin ⩽SOCtl ⩽SOCmax , ∀l ∈ L ,t ∈ T (17) form as:

( ) min f
ϕ

Pdev,t = 1− uev dm
i,t ⋅Pev , ∀t ∈ T (18) s.t. (1) − (30)
i∈K
{ (32)
ϕ = uev l ev l ev l
i,t , ui,t , si,t , si,t , di,t , di,t , ug,t , Pg,t , Qg,t ,
}
0⩽Pcev,i,t ⩽uev cm
i,t Pev , ∀i ∈ K , l ∈ L , t ∈ T (19)
Pdl,i,t , Pcl,i,t , Qdl,i,t , Qcl,i,t , Pev ev
c,i,t , Pd,t , Vi,t

0⩽Pdl,i,t ⩽uli,t Pdm


b , ∀i ∈ K , l ∈ L , t ∈ T (20) In practice, the output power of REGs has randomness and volatility,
which cannot be ignored. Moreover, the scheduling model of EMS
0⩽Pcl,i,t ⩽uli,t Pcm
b , ∀i ∈ K , l ∈ L , t ∈ T (21) contains nonlinear constraints (23)-(24) and many discrete decision
variables, making it difficult to use the traditional robust method. Sto­
chastic programming is also not suitable due to the huge computing cost.
2.2. Model of the energy flow Therefore, IGDT is applied to address the uncertainties in the EMS. IGDT
is a non-probabilistic and non-fuzzy method for coping with uncertainty
There is a real-time power balance between source, load, and bat­ problems, and it has higher computational efficiency than other sto­
teries within the islanded microgrids. Due to no loads in RRIs, this paper chastic programming methods. The main idea of IGDT is to find a so­
only considers the power balance constraint in the microgrid of RRIs lution that maximizes the range of uncertainties within acceptable
shown in Eq. (22). The AC power flow constraints in the microgrid of LCI system performance [35–37].
{ pv }
are modeled as presented in Eqs. (23)-(30). Especially, Eqs. (23)-(24) The parameter ψ = Pwt t , Pt is uncertain, and their information gap
indicate the power balance conditions. Eq. (25) gives the power flow model can be expressed as follows:
limits on power lines. Eq. (26) presents the amplitude limits of voltage. {
Eq. (27) suggests the power limits of DGs. Eqs. (28)-(30) are the rela­ ψ ∈ Γ(α, ψ̃ )Γ(α, ψ̃ ) = {ψ : |(ψ − ψ̃ )/ψ̃ |⩽α|α⩾0} (33)
tionship between the output of BMS and the power of batteries.

Preg d c ev
(22) where Γ(α, ψ̃ ) indicates that ψ deviates from the predicted value ψ̃ by
n,t + Pn,t − Pn,t − Pc,n,t ⩾0, ∀n ∈ N , ∀t ∈ T
no more than α|ψ̃ |.
∑ ∑( ) ∑ Then we can build a robust equation by:
Pg,t + Pdm,t − Pcm,t − PDi,t −
i∈Ω α̃(ϕ, fp ) = max{α : ∀X ∈ Γ(α, ψ̃ ), f (ψ , ϕ)⩽fp } (34)
(23)
g∈G m∈M

Vi,t Vj,t (Gij cosθi,j,t + Bij sinθi,j,t ) = 0, ∀t ∈ T where α̃ is the robustness coefficient used to characterize the
robustness of the solution. And the larger α̃ is, the more robust solution
i,j∈W

∑ ∑( ) ∑ is.
Qg,t + Qdm,t − Qcm,t − QDi,t − Therefore, the IGDT-based EMS model can be written as:
i∈Ω
(24)
g∈G m∈M
∑ max α
Vi,t Vj,t (Gij sinθi,j,t − Bij cosθi,j,t ) = 0, ∀t ∈ T
(35)
ϕ
i,j∈W s.t. fp = max f (ψ , ϕ) fp ⩽(1 + β)f0 (1) − (30)
ψ ∈Γ(α,ψ̃ )

4
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

Note that the IGDT-based model is a bi-level optimization problem. from subproblems and is set to 0 initially.
The upper level is set to maximize the range of uncertainty within The master problem is a mixed-integer linear problem and can be
acceptable operation cost, while the lower level aims to find an optimal solved efficiently by GUROBI. After optimization of the master problem
solution to deal with any possible variation ψ in the set of Γ(α, ψ̃ ). is completed, the values of P1g,t and P1m,t need to be transferred to
When the actual power of WT and PV on the RRIs are lower than the subproblems.
predicted value, the energy collected by batteries will be reduced. The
resulting power shortage can only be compensated by DGs, which will 3.1.2. Subproblems
increase the operation cost. Therefore, when the power of WT and PV Once the optimal solution is found in the master problem, it’s
are selected according to (36), the operation cost will be the largest. In necessary to check the feasibility of the solution to satisfy the power flow
this situation, max f can be degraded to f in the model. constraints (23)-(26) in subproblems. The form of the feasibility sub­
⎧ problems is given as follows:

⎨ ∑( ) ∑( )
(36)
wt pv pv
Pwt
t = (1 − αw )P̃t Pt = (1 − αp )P̃t minf2,t = ε+g,t + ε−g,t + ε+m,t + ε−m,t

⎩ g∈G m∈M

s.t. P2g,t − ε + 1
g,t ⩽Pg,t : λ+
where αw > 0 and αp > 0. g,t

Besides, the objective of the model is to maximize the value of α, and P2g,t + ε−g,t ⩽P1g,t : λ−g,t
the objective function needs to be processed as follow: (40)
P2m,t − ε+ 1 +
m,t ⩽Pm,t : λm,t
max α = λw αw + λp αp (37) P2m,t + ε−m,t ⩽P1m,t : λ−m,t
To get a more optimal solution, the weighting factor is determined (23) - (26)
by: ∀t ∈ T
α** α*p Although the subproblems have non-convex power flow constraints
λw = w
, λp = (38)
α*w + α*p α*w + α*p (23)-(26), it is a continuous second-order differentiable problem and can
be solved by the interior point method [38].
where α∗w/p is the maximum robustness coefficient under the single
The subproblem is feasible if and only if f2,t is less than θ, where θis
objective. enough smaller value. If f2,t becomes more than θ, that is, the sub­
problem is infeasible, the corresponding infeasibility cut is generated as
3. SOLUTION ALGORITHM Eq. (41) and is added to the master problem with the power loss
determined by Eq. (42).
In this paper, the solution for the IGDT-based EMS model can be ∑( )( )
divided into two stages: f2,t − λ+ −
g,t − λg,t P2g,t − P1g,t −
Stage 1: solving the deterministic scheduling model (32) to obtain g∈G
∑( )( ) (41)
the optimal objective value f0; λ+ −
P2m,t − P1m,t ⩽0
m,t − λm,t
Stage 2: passing f0 to Eq. (35) and using Eq. (37) as the objective m∈M
function to solve the IGDT-based model. ∑
Due to nonlinear constraints (23)-(24), the model is difficult to solve. Ploss
t = Vi,t Vj,t (Gij cosθi,j,t + Bij sinθi,j,t ) (42)
To this end, the Generalized Benders method is applied to decompose i,j∈W

the original problem into a master problem of discrete battery logistics


The master problem and subproblems are solved iteratively until all
optimization and subproblems of AC power flower feasibility checking
the subproblems are feasible. At this time, the optimal solution to the
to reduce the difficulty of solving. The specific decomposition for two
master problem is the optimal solution to the original problem. The
stages is shown as follows.
flowchart of the Generalized Benders method is summarized in Algo­
rithm 1.
3.1. Stage one Algorithm 1. Benders for the deterministic model

3.1.1. Master problem 1: Initialize the parameter θ and set the iteration index k = 0.
The master problem aims to optimize discrete battery logistics 2: Solve the initial master problem (39) by GUROBI, and get the optimal solution ϕ1k .
without considering power flow constraints (23)-(24). The initial master 3: Pass the optimal solution ϕ1k to the subproblems.
problem is formulated as below: 4: Solve the subproblems (41) using the interior point method.
5: Check the feasibility of subproblems:
min f1 = f If ∃t, f2,t > θ, add infeasibility cut and power loss as expressed in (41)-(42) to the master
ϕ problem (39). Then, go to step 2 and make k = k + 1.
∑ ∑ ∑
s.t. Pg,t + Pm,t − PDi,t − Ploss =0 If ∀t, f2,t ⩽θ, then terminate and report the optimal solution ϕ1k and objective value f0.
t
g∈G m∈M i∈Ω

Pm,t = Pdm,t − Pcm,t


(39)
(1) − (22), (27) − (28)
{
ϕ = uev l ev l ev l 3.2. Stage two
i,t , ui,t , si,t , si,t , di,t , di,t , ug,t ,
}
Pg,t , Pdl,i,t , Pcl,i,t , Pev ev
c,i,t , Pd,t
The IGDT-based model is formed according to the optimal objective
value f0 at stage one. And the Generalized Benders method is also
Eq. (39) is the simplification of Eq. (32), where Ploss
t is power loss adopted to solve the problem, which is the same as Algorithm 1 except

5
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

that the objective function of the main problem. The master problem of Considering the installment time of batteries, the sailing time between
the IGDT-based model is cast as: BMS calculated by Eq. (4) is given in Table 2. At the initial time, EV’s

max α = λw αw + λp αp
ϕ

s.t. fp = f (ψ , ϕ) (43)
{ ⃒ } ∑ ∑ ∑
w/p ⃒ w/p w/p
ψ = Pt ⃒Pt = (1 − αw/p )P̃t fp ⩽(1 + β)f0 Pg,t + Pm,t − PDi,t − Ploss
t = 0 Pm,t = Pdm,t − Pcm,t (1) − (22), (27) − (28)
g∈G m∈M i∈Ω

SOC is set to 0.5 and berths on the LCI. Moreover, there are three bat­
teries located at BMS 1 and others at BMS 2. The initial SOC of all bat­
The subproblems are the same as those of the deterministic model. teries is 0.5.

4. Case studies
4.2. Performance analysis of the optimal solution
4.1. Test system
Firstly, we analyze the scheduling plan of EV and batteries as follows:
For numerical simulations, we consider three islands in the South
1) Scheduling Results of EV: EV serves as a vehicle for battery swapping,
China Sea as one LCI and two RRIs, which is shown in Fig. 3(a). The
making multiple voyages between islands during the scheduling
modified IEEE 6-node system [39] illustrated in Fig. 3(b) is adopted on
period. As shown in Fig. 5, the sailing order of EV is “1 → 2 → 1 → 3
the LCI, where both DG capacity is 400 kW. There are 400 kW WT and
→ 1 → 2 → 1” and EV has 12 h of sailing. Besides, EV needs sufficient
600 kW PV on island 2 and island 3, respectively. And Fig. 4 depicts the
energy for battery transportation safely. So, EV would be charging on
output of REG and load [32]. Based on the 7.5 yuan/L average price of
the island, replenishing energy for navigation. And EV prefers to stay
diesel in coastal areas, the price of diesel transported to pelagic islands is
longer time on the RRIs than the LCI, directly replenishing clean
set equal to 9 yuan/L.
energy from WT and PV.
It is assumed that the EV is powered by a lithium-ion battery with a
rated capacity of 200 kW⋅h, rated sailing power of 40 kW, and rated
charging power of 60 kW. And EV’s rated speed is 14 km/h. Meanwhile, Table 2
there are 5 sets of sodium-sulfur batteries with a rated capacity of 2000 Distance and sailing time between BMS.
kW⋅h and rated charging/discharging power of 600 kW. The maximum BMS: i-j 1–2 1–3 2–3
and minimum SOC values of the batteries and EV are 0.9 and 0.1, and Distance dij (km) 22.4 23.8 11.6
charging/discharging efficiencies are set to 90%. The charging/dis­ Sailing time τij (h) 2 2 1
charging degradation costs are 0.8 yuan/kW⋅h and 1.6 yuan/kW⋅h [40],
respectively.
Without loss of generality, each island only has one BMS.
3 0.9
SOC
BMS

BMS EV 2 0.5
LCI
(Isla nd 1) G1
4 1 0.1
5 1
RRI 3 0.9
SOC

(Isla nd 3)
2
Battery 2 0.5
BMS

3
RRI G2 1
(Isla nd 2) 1 0.1
6
(a) The relative position of pelagic islands (b) Microgrid topology diagram 3 0.9
SOC

Battery 2 0.5
BMS

Fig. 3. The relative position of pelagic islands and microgrid topology diagram 2
of LCI. 1
0.1
3 0.9
SOC

Battery
BMS

2 0.5
3
800 1 0.1
Load WT PV 0.9
3
S OC

600
BMS

Battery 2 0.5
Power (kW)

4
1 0.1
400
3 0.9
S OC
BMS

200 Battery 2 0.5


5
1 0.1
0 2 6 10 14 18 22 2 6 10 14 18 22
2 4 6 8 10 12 14 16 18 20 22 24
Time (h) Time (h)
Time (h)
Fig. 5. Time-space distribution and SOC diagram of batteries and EV.
Fig. 4. The output power of REGs and load curve.

6
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

4000 battery during the day.


RRI BMS 2 BMS 3 Moreover, the energy flow of the PIMGC is shown in Fig. 6 and Fig. 7,
3000 and the following characteristics can be found:
Available energy (kW·h)

2000
1000 1) The battery logistics based on EV has obvious time-delayed and
discrete characteristics. As shown in Fig. 6, EV transports batteries
0 with 3187 kW⋅h, 1600 kW⋅h, and 2879 kW⋅h available energy is
4000 transported from the RRIs to LCI by EV in turn, and each process
LCI BMS 1
3000 expends 2 h. Limited by spatial distance and speed, EV navigation
takes a long time, making it show intermittent work in time–space
2000 2h 2h 2h distribution.
1000 2) The energy flow of the PIGMC has the characteristics of real-time
0 power balance within the islanded microgrids and delayed energy
2 6 10 14 18 22 flow among islanded microgrids coexisting. As shown in Fig. 7, the
Time (h) source, storage, and load are located at the microgrid of LCI, and the
Fig. 6. Residual energy curve of BMS. power flow meets the real-time power balance. At 20:00, 2879 kW⋅h
energy in the form of battery on EV is expected to arrive at BMS 1
after one hour, which will participate in the next moment of the
243.10+
microgrid operation of LCI.
-57.20-j36.50 PV
j193.21
170.66+ 54.86+j27.41 Finally, the microgrid operation on the LCI is shown in Fig. 8. The
BMS

BMS
j118.46 G1 batteries at the BMS and DGs are coordinated in the scheduling period to
5 4 jointly meet the load demand of the island. During the period of lower
1
169.33+j115.33

2879 load, i.e. 0:00–8:00 and 22:00–24:00, batteries at BMS 1 supply the load
-0.20-j0.37

independently. However, during other periods, batteries and DG com­


EV
bined energy supply mode is mainly used on the island due to high load
2 power. And two DGs operation is determined by the power flow distri­
6
3 48.97+
G2
bution. It can be seen that energy collected from RRIs can be reasonably
j36.32 83.44+j62.10
distributed with DGs to provide stable and economic operation for the
BMS

24.88+j27.99 microgrid on the LCI.


48.92+j36.77 WT
4.3. Robustness analysis
Fig. 7. System energy flow at 20:00 (kW).
In this case, we analyze the robustness of the IGDT-based EMS model.
2) Scheduling Results of Batteries: The time–space transportation of bat­ Firstly, the uncertainties of wind power and PV are analyzed. The model
teries depends on EV, whose navigation plan is consistent with EV (35) is solved multi-objectively using the Pareto optimality concept with
demonstrated in Fig. 5. Initially, the No. 1 battery is mainly used for the objective of αw and αp, respectively. Suppose that the operator
power supply and EV charging. At 2:00, No. 1 and No. 3 batteries are chooses the given cost target fp = 1.05*f0. The model (35) is solved for
transported to BMS 2 for charging, where the output power of WT is the given value of β and tries to find the maximum value of αw and αp.
large. After replenishment for a short time, EV returns to BMS 1 with This gives a set of solutions as shown in Fig. 9. The solutions conform to
No. 4 and No.5 batteries full of clean energy from RRI. At 8:00, No. 2 Pareto distribution and are relatively uniform. And the maximum values
battery, which is exhausted, would be moved to BMS 3 to collect the of αw and αp can be optimized to be 11.6% and 24.7%, respectively. It is
energy of PV on the RRI. Until No. 2 battery is fully charged at 14:00, worth noting that the latter is more than twice the former. Because there
it immediately returned to the LCI. At 17:00, EV would go to BMS 2 is a difference in the total value of energy collected from island 2 (WT
to replace the empty batteries (No. 4 and No. 5) on the LCI with fully island) and island 3 (PV island), which is 5610 kW⋅h and 1723 kW⋅h in
charged batteries (No. 1 and No. 3) on the RRI and return at 22:00. It the deterministic solution. In practice, the uncertainty of WT should be
can be seen that EV allocates the position of batteries according to roughly the same as that of PV. To make the solution more suitable for
the power generation of RRIs. Due to the output power of WT staying the actual scenario, we set the weighting factors with the maximum
in [100, 400] kW, batteries are always available at BMS 2, while BMS values, i.e. λw = 0.68 and λp = 0.32, respectively.
3 is only issued and equipped with the Then, we explore the impact of the cost deviation factor β on the
solution against the WT and PV uncertainty. The (35) is solved with
different values β. The parameter β is varied from 0 to 20% and the
optimum solutions are found. The robustness coefficient α shows the
800
BMS 1
600 G1
Power/kW

G2
400

200

0
2 6 10 14 18 22
Time (h)

Fig. 8. The power distribution in the microgrid of LCI. Fig. 9. The Pareto optimal front of robustness against WT and PV uncertainties.

7
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

0.6 of the cost deviation factor β is 0.08 in the IGDT optimization. The EV
scheduling plans of the two solutions are different and listed in Table 3.
Furthermore, we use 50 scenarios with large fluctuations for real
0.4
simulation to reflect the robustness of the solution. The operation costs
values

are displayed in Fig. 11. The operation costs of most scenarios with IGDT
0.2 solution are lower than that of stochastic solution. It is demonstrated
that the IGDT solution is more robust than the stochastic solution. Be­
sides, the execution time is 1179 s for IGDT optimization while the
0
0 0.05 0.1 0.15 0.2 stochastic programming runs>10000 s due to a large number of discrete
values decision variables.
In conclusion, the IGDT-based EMS model has obvious advantages in
Fig. 10. Optimum robustness coefficients value versus cost deviation factor. robustness and computational efficiency. And the operator can set a
reasonable cost deviation factor β according to their risk preferences and
get the strategies to their satisfaction.
Table 3
Performance of the EV’s trips under different cases.
Case Trip 1st 2nd 3rd 4th 5th 6th 7th 4.4. Economy analysis
IGDT β= Islands 1 2 1 3 1 2 1
0.08 Stay time 2 1 1 4 1 1 2 In the section, we demonstrate the economic of EMS compared with
(h) the other energy supply system with submarine cable and DGs, respec­
β= Islands 1 2 1 2 1 – – tively. The natural conditions of the pelagic islands are the same as the
0.10 Stay time 2 3 4 5 2 – –
above, and the uncertainties are not considered. The schemes are
(h)
Stochastic Islands 1 2 1 3 1 2 1
described as follows and configuration parameters are shown in Table 4.
Stay time 1 1 1 4 1 1 3
(h) 1) Scheme 1: The EMS is used with the same renewable generators and
batteries as described in Section 4.1.
2) Scheme 2: The submarine cables are used to establish energy flow
channels for islands with the same renewable generators as described
10 4 in Section 4.1. The total length of submarine cables is about 40 km. A
2.06
400 kW/1000kWh battery is also equipped to balance power output
and acts as an emergency power supply.
3) Scheme 3: Only DGs with a total rated capacity of 800 kW are used to
Cost (yuan)

2.04 supply energy on the LCI.

To accurately calculate the costs of each scheme, the investment cost,


operation cost, environmental cost, and maintenance cost are taken into
Stochastic IGDT
2.02 consideration and the uniform annual value method is also applied.
0 10 20 30 40 50 Firstly, the parameters of devices are listed in Table 5 and the discount
Scenario Number rate is set to 5%. The investment costs and maintenance costs can be
calculated according to the data in Table 4 and Table 5, and the results
Fig. 11. The operation costs of 50 scenarios with different solutions. are shown in Table 6. Secondly, we modify the model in this paper for
Scheme 2 and Scheme 3 and optimize the operation costs in Table 6.
uncertainties of WT and PV. When the value α is kept zero, it means that Thirdly, the environmental costs are caused by DG emissions and the
there is no uncertainty. The variation of robustness coefficient α against Table 4
this parameter β begins from 0 to 45%, as shown in Fig. 10. As is ex­ treatment cost of DG is set to 0.225 yuan/kWh [41]. Finally, the
pected, the values α against uncertainties show an increasing behavior annual comprehensive costs are obtained as shown in Table 6.
with the increase of β. It is worth noting that there is a jump at β = 0.10. It can be seen from Table 6 that the annual comprehensive cost of
Because the scheduling plans of EV have undergone major changes. As Scheme 1 is lowest than that of Scheme 2 and Scheme 3, indicating that
present in Table 3, EV no longer travels to island 3 (PV island) to collect the energy supply system proposed in this paper is the most economical.
energy when β = 0.10. The uncertainty of PV can reach the maximum Particularly, although Scheme 2 has lower operation cost due to the
under this circumstance. Therefore, the value of α has made a big
change. Table 5
Finally, scenarios-based stochastic programming is applied to Parameters of devices [32,41,42]
compare with the IGDT model. Assuming that the forecast errors of WT
Distributed Generators
and PV obey normal distribution, 1000 scenarios are generated by
Monte Carlo simulation and then reduced to 20 typical scenarios [41]. Type Captial cost Maintenance cost (yuan/ Economic life
(yuan/kW) kW/yr) (yr)
And the value
WT 6210 200 30
PV 8280 130 30
DG 2070 915 30
Table 4 BMS 750 75 10
Device configuration of each mode. Battery – 241 –
Other Devices
Scheme WT PV DG Battery BMS Submarine EV
Type Captial cost Maintenance cost Economic life
(kW) (kW) (kW) (kWh) (kW) cable (km)
(yr)
1 400 600 800 10,000 1600 0 1 Submarine 5,000,000 (yuan/ 5000 (yuan/km/yr) 30
2 400 600 0 1000 400 40 0 cable km)
3 0 0 800 0 0 0 0 EV 800,000 (yuan) 10,000 (yuan/yr) 10

8
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

Table 6
Annual comprehensive costs.
Scheme Operation cost (yuan) Environmental cost(Yuan) Maintenance cost (yuan) Investment cost (yuan) Comprehensive cost (yuan)

1 7,098,155 328,500 3,000,000 851,498 11,588,153


2 525,600 0 568,000 13,533,902 14,658,502
3 11,963,240 822,710 732,000 107,725 13,625,675

higher energy flow efficiency, Scheme 1 has greater advantages in the [8] Ehsan A, Yang Q. Scenario-based investment planning of isolated multienergy
microgrids considering electricity, heating and cooling demand. Appl Energy 2019;
investment costs and saves approximately 3 million yuan. Besides, the
235:1277–88.
operation cost and environmental cost of Scheme 3 without renewable [9] Wang Zhixun, Lin Xiangning, Tong Ning, Li Zhengtian, Sun Shitong, Liu Chang.
generators are highest, resulting in Scheme 3 spend about 2 million yuan Optimal planning of a 100% renewable energy island supply system based on the
more than Scheme 1. Therefore, in general, Scheme 1 is preferable to integration of a concentrating solar power plant and desalination units. Int J Elec
Power 2020;117:105707.
Scheme 2 and Scheme 3. [10] Yang M, Sun L, Wang J. Multi-objective optimization scheduling considering the
operation performance of islanded microgrid. IEEE Access 2020;8:83405–13.
[11] Ghasemi Ahmad, Enayatzare Mehdi. Optimal energy management of a renewable-
5. Conclusion based isolated microgrid with pumped-storage unit and demand response. Renew
Energy 2018;123:460–74.
In this paper, we have proposed a novel EMS based on battery lo­ [12] Lin S, Wang Y, Liu M, Fan G, Yang Z, Li Q. Stochastic optimal dispatch of PV/wind/
diesel/battery microgrids using state-space approximate dynamic programming.
gistics by EV in the PIMGC. An IGDT-based EMS model is presented to IET Gener Transm Dis 2019;13(15):3409–20.
optimize the day-ahead scheduling of battery logistics among islanded [13] Lara JD, Olivares DE, Ca~nizares CA. Robust energy management of isolated
microgrids and energy flow within islanded microgrids. And the model microgrids. IEEE Syst J 2019;13(1):680-91.
[14] Yeh Wei-Chang, He Min-Fan, Huang Chia-Ling, Tan Shi-Yi, Zhang Xianyong,
could be solved efficiently by the Generalized Benders method. The
Huang Yaohong, et al. New genetic algorithm for economic dispatch of stand-alone
simulation results have shown the effectiveness and economy of the three-modular microgrid in DongAo Island. Appl Energy 2020;263:114508.
proposed EMS in the PIMGC. [15] Jayachandran M, Ravi G. Predictive power management strategy for PV/battery
hybrid unit based islanded AC microgrid. Int J Elec Power 2019;110:487–96.
The proposed energy supply system based on battery logistics can
[16] Hu M, Wang Y, Lin X, Shi Y. A Decentralized Periodic Energy Trading Framework
improve the economy of pelagic islands, saving 3 million yuan and 2 for Pelagic Islanded Microgrids. IEEE Trans Ind Electron 2020;67(9):7595–605.
million yuan than the other energy supply system with submarine cable [17] Wang Zhixun, Lin Xiangning, Liu Chang, Tong Ning, Li Zhengtian, Sui Quan, et al.
and DGs, respectively. And the IGDT-based EMS can collaboratively A hybrid transmission network in pelagic islands with submarine cables and all-
electric vessel based energy transmission routes. Int J Elec Power 2020;120:
optimize the battery logistics and energy flow to ensure the economic 106005.
and reliable operation of PIGMCs. In short, this study provides a new [18] Ahmadi Seyed Ehsan, Rezaei Navid. A new isolated renewable based multi
perspective on the construction of pelagic islands. microgrid optimal energy management system considering uncertainty and
demand response. Int J Elec. Power 2020;118(105760).
Due to a certain margin reserved for sailing time, scheduling of EMS [19] Zhao Z, Guo J, Luo X, Xue J, Lai CS, Xu Z, et al. Energy Transaction for Multi-
could be further refined and the uncertainty of sailing time needs to be Microgrids and Internal Microgrid Based on Blockchain. IEEE Access 2020. https://
considered in future work. doi.org/10.1109/ACCESS.2020.3014520.et.
[20] Chen L, Zhu X, Cai J, Xu X, Liu H. Multi-time scale coordinated optimal dispatch of
microgrid cluster based on MAS. Electr Power Syst Res 2019;177:105976.
[21] Najmeh B, Ahmadreza T, Amjad AM, Josep MGA. hierarchical energy management
Declaration of Competing Interest strategy for interconnected microgrids considering uncertainty. Int J Elec Power
Syst 2019;109:597–608.
[22] Lahanda Purage MIS, Krishnan A, Foo YSE, Gooi HB. Cooperative Bidding-based
The authors declare that they have no known competing financial Robust Optimal Energy Management of Multi-Microgrids. IEEE Trans Ind Inf 2020;
interests or personal relationships that could have appeared to influence 16(9):5757–68.
the work reported in this paper. [23] Qiu Haifeng, You Fengqi. Decentralized-distributed robust electric power
scheduling for multi-microgrid systems. Appl Energy 2020;269:115146.
[24] Huang W, Zheng W, Hill DJ. Distributionally Robust Optimal Power Flow in Multi-
Acknowledgments Microgrids with Decomposition and Guaranteed Convergence. IEEE Trans Smart
Grid 2020. https://doi.org/10.1109/TSG.2020.3012025.
[25] Du Y, Li F. Intelligent Multi-Microgrid Energy Management Based on Deep Neural
This work was supported by the Science and Technology Project of Network and Model-Free Reinforcement Learning. IEEE Trans Smart Grid 2020;11
State Grid Corporation of China (5700-201999495A-0-0-00). (2):1066–76.
[26] Du Y, Li F. A Hierarchical Real-Time Balancing Market Considering Multi-
Microgrids With Distributed Sustainable Resources. IEEE Trans Sustain Energy
References 2020;11(1):72–83.
[27] Liu Z, Gao J, Yu H, Wang X. Operation Mechanism and Strategies for Transactive
Electricity Market With Multi-Microgrid in Grid-Connected Mode. IEEE Access
[1] Yousefzadeh Moslem, Lenzen Manfred, Tyedmers Elijah Keenan, Hassan Ali SM. An
2020;8:79594–603.
integrated combined power and cooling strategy for small islands. J Clean Prod
[28] Xu D, Zhou B, Liu N, Wu Q, Voropai N, Li C, et al. Peer-to-Peer Multi-Energy and
2020;276(122840).
Communication Resource Trading for Interconnected Microgrids. IEEE Trans on
[2] Zhao Bo, Chen Jian, Zhang Leiqi, Zhang Xuesong, Qin Ruwen, Lin Xiangning. Three
Ind Inf 2020. https://doi.org/10.1109/TII.2020.3000906.
representative island microgrids in the East China Sea: Key technologies and
[29] Liu H, Li J, Ge S, He X, Li F, Gu C. Distributed Day-Ahead Peer-to-Peer Trading for
experiences. Renew and Sustain Energy Rev 2018;96:262–74.
Multi-Microgrid Systems in Active Distribution Networks. IEEE Access 2020;8:
[3] Alves M, Segurado R, Costa M. Increasing the penetration of renewable energy
66961–76.
sources in isolated islands through the interconnection of their power systems. The
[30] Du Y, Wang Z, Liu G, Chen X, Yuan H, Wei Y, et al. A cooperative game approach
case of Pico and Faial islands. Azores. Energy 2019;182:502–10.
for coordinating multi-microgrid operation within distribution systems. Appl
[4] Fernández Prieto L, Rodríguez Rodríguez G, Schallenberg Rodríguez J. Wave
Energy 2018;222:383–95.
energy to power a desalination plant in the north of Gran Canaria Island: Wave
[31] Zhao Z, Guo J, Lai CS, Xiao H, Zhou K, Lai LL. Distributed Model Predictive Control
resource, socioeconomic and environmental assessment. J Environ Manage 2019;
Strategy for Islands Multi-Microgrids Based on Non-Cooperative Game. IEEE Trans
231:546–51.
Ind Inf 2020. https://doi.org/10.1109/TII.2020.3013102.
[5] Lorafe Lozano, Edward M. Querikiol, Michael Lochinvar S. Abundo, Luzvisminda
[32] Sui Q, Zhang R, Wu C, Wei F, Lin X, Li Z. Stochastic scheduling of an electric vessel-
M. Bellotindos. Techno-economic analysis of a cost-effective power generation
based energy management system in pelagic clustering islands. Appl Energy 2020;
system for off-grid island communities: A case study of Gilutongan Island, Cordova,
259:114155.
Cebu, Philippines. Renew Energy 2019; 140:905-911.
[33] Sui Q, Wei F, Wu C, Lin X, Li Z, Wang Z. Day-ahead Energy Management for Pelagic
[6] Wijayatunga Priyantha, George Len, Lopez Antonio, Aguado Jose A. Integrating
Island Microgrid Groups Considering Noninteger-hour Energy Transmission. IEEE
Clean Energy in Small Island Power Systems: Maldives Experience. Energy
Trans Smart Grid 2020. https://doi.org/10.1109/TSG.2020.2994236.
Procedia 2016;103:274–9.
[7] Kuang Y, Zhang Y, Zhou B, Li C, Cao Y, Li L, et al. A review of renewable energy
utilization in islands. Renew Sustain Energy Rev 2016;59:504–13.

9
C. Wu et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106573

[34] Hu M, Wang Y, Xiao J, Lin X. Multi-energy management with hierarchical Xiangning Lin is currently a Professor at Huazhong University
distributed multi-scale strategy for pelagic islanded microgrid clusters. Energy of Science and Technology (HUST). He received the M.Sc. and
2019;185:910–21. Ph.D. degrees in electrical engineering from HUST, Wuhan,
[35] Mostafa MH, Aleem SHEA, Ali SG, Abdelaziz AY, Ribeiro PF, Ali ZM. Robust energy China. His research interests are modern signal processing and
management and economic analysis of microgrids considering different battery power system protective relaying
characteristics. IEEE Access 2020;8:54751–75.
[36] Moghaddas-Tafreshi Seyed Masoud, Jafari Morteza, Mohseni Soheil, Kelly Scott.
Optimal operation of an energy hub considering the uncertainty associated with
the power consumption of plug-in hybrid electric vehicles using information gap
decision theory. Int J Elec. Power 2019;112:92–108.
[37] Xie Yunyun, Chen Xi, Qiuwei Wu, Zhou Qian. Second-order conic programming
model for load restoration considering uncertainty of load increment based on
information gap decision theory. Int J Elec Power 2019;105:151–8.
[38] Nejdawi IM, Clements KA, Davis PW. An efficient interior point method for
sequential quadratic programming based optimal power flow. IEEE Trans Power
Syst 2000;15(4):1179–83.
[39] Allen. J. Wood, Bruce F. Wollenberg. Power Generation, Operation, and Control. Zhixun Wang received the B.Sc. degree in electrical engi­
2nd Edition; New Jersey: John Wiley & Sons, Inc; 1996. neering from Huazhong University of Science and Technology
[40] Sun Y, Li Z, Shahidehpour M, Ai B. Battery-Based Energy Storage Transportation (HUST) in 2016. He is a Ph.D. candidate at HUST. His re­
for Enhancing Power System Economics and Security. IEEE Trans Smart Grid 2015; searches mainly focus on optimal power system/microgrid
6(5):2395–402. scheduling and protective relay.
[41] Fan Hong, Yuan Qianqian, Cheng Haozhong. Multi-objective stochastic optimal
operation of a grid-connected microgrid considering an energy storage system.
Appl Sci 2018;8(12):2560.
[42] Cao X, Wang J, Zeng B. A Chance Constrained Information-Gap Decision Model for
Multi-Period Microgrid Planning. IEEE Trans Power Syst 2018;33(3):2684–95.

Chuantao Wu received the B.Sc. degree in electrical engi­


neering from Huazhong University of Science and Technology
(HUST) in 2018. He is a Ph.D. candidate at HUST. His research
interests are microgrid planning and scheduling as well as
Zhengtian Li received the B.Sc. degree from Wuhan University
optimal power system.
in 2002 and the Ph.D. degree from the Huazhong University of
Science and Technology (HUST) in 2011. Currently, he is an
Associate Professor at HUST. His research interests are digital
protection relaying.

Quan Sui received the B.Sc. degree in electrical engineering


from Huazhong University of Science and Technology (HUST)
in 2017. He is a Ph.D. candidate at HUST. His research interests
are microgrid planning and scheduling as well as optimal power
system.

10

You might also like