You are on page 1of 12

Applied Energy 235 (2019) 1277–1288

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Scenario-based investment planning of isolated multi-energy microgrids T


considering electricity, heating and cooling demand

Ali Ehsana,b, Qiang Yanga,
a
College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China
b
Department of Electrical Engineering, COMSATS University Islamabad, Sahiwal Campus 57000, Pakistan

H I GH L IG H T S

• AThescenario-based investment planning model is proposed for the isolated microgrids.


• The supply and demand of the electrical, heating and cooling energy are considered.
• The heuristic moment matching method is employed to generate the scenarios.
• solution minimizes the overall microgrid costs and Carbon dioxide emissions.

A R T I C LE I N FO A B S T R A C T

Keywords: Multi-energy microgrids provide a flexible solution for the utilization of the distributed energy resources in order
Multi-energy to meet the electrical, heating and cooling energy demands in the off-grid communities. However, the planning
Isolated of the multi-energy microgrids is a non-trivial problem due to the complex energy flows between the sources and
Microgrid the loads pertaining to the electrical, heating and cooling energy, along with the intermittency of the renewable
Distributed energy resource
distributed generation. This work proposes a scenario-based stochastic multi-energy microgrid investment
Heuristic moment matching
Uncertainty
planning model that aims to minimize the investment and operation costs as well as the Carbon dioxide emis-
sions by determining the optimal distributed energy resource mix, siting and sizing in the isolated microgrids.
The proposed planning model employs the power flow and heat transfer equations to explicitly model the energy
flows between electrical, heating and cooling energy sources and loads. Moreover, an uncertainty matrix is
employed to tackle the operational uncertainties associated with the wind and photovoltaic generation, and the
electrical, heating and cooling loads. The uncertainty matrix is modeled using the heuristic moment matching
method that effectively captures the stochastic moments and correlation among the historical scenarios. The
numerical results obtained from the case-study in the 19-bus microgrid test system confirm that the proposed
methodology provides significant reductions in the investment and operation costs as well as the Carbon dioxide
emissions. Finally, the superiority of the proposed planning solution is also validated using the deterministic
planning solution as the comparison benchmark.

1. Introduction access to the main grid, e.g., the islands, the remote villages and
communities, and the secluded military bases and industrial sites [5,6].
The rapid deployment of the microgrids is expediting the integra- Moreover, microgrids can relieve the control burden on the main
tion of the renewable and distributed energy resources (DERs) as well grid by employing a decentralized coordination of the distributed en-
as the distributed energy storage systems in the modern power systems ergy resources [7,8]. Microgrids not only secure a highly reliable en-
[1,2]. Microgrids usually operate in either the isolated (i.e. autono- ergy supply but also ensure resiliency against the natural calamities,
mous) mode or the grid-connected mode [3]. The isolated multi-energy particularly for the critical facilities [9]. The high share of the renew-
microgrids have recently been investigated technically, emerging as an able distributed generation (DG) in the microgrids mitigates the op-
economically profitable solution [4] for the provision of electrical, erational costs and the CO2 emissions, in addition to reducing the fossil-
heating and cooling energy demands in the remote areas with limited fuel dependency [9]. The exploitation of the recovered heat from the


Corresponding author.
E-mail addresses: aliehsan@zju.edu.cn (A. Ehsan), qyang@zju.edu.cn (Q. Yang).

https://doi.org/10.1016/j.apenergy.2018.11.058
Received 25 July 2018; Received in revised form 14 November 2018; Accepted 16 November 2018
Available online 30 November 2018
0306-2619/ © 2018 Elsevier Ltd. All rights reserved.
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

Nomenclature rij,xij resistance and reactance of branch i–j


v̲ ,v̄ lower and upper limit on bus voltage
Sets and indices
Variables and functions
c set of modular distributed energy resources: photovoltaic
(pv), gas boiler (bl), absorption chiller (ac), electric chiller bci binary variable for modular DER installed at ith bus
(ec) Ccost overall microgrid costs
d set of non-modular distributed energy resources: wind CCO2 overall microgrid CO2 emissions
turbine (wt), combined-heat-and-power (chp) Cis installed capacity of storage technology s at ith bus (MWh)
g set of all distributed energy resources( c∪ d) cris, h, dris, h charging and discharging rate of storage technology s at ith
l set of loads: electrical load (el), heating load (hl), cooling bus for scenario h (MW)
load (cl) Eis, h energy stored in storage technology s at ith bus for scenario
s set of storage technologies: electric storage (es), heat sto- h (MWh)
rage (hs) hij, h heat flow in pipe i–j for scenario h (MW)
h/ΩH index/set of generated scenarios iij, h current in branch i–j for scenario h
i/ΩB index/set of buses ndi number of non-modular DERs installed at ith bus
i− j/ΩL index/set of branches pij, h , qij, h active and reactive power flow in branch i–j for scenario h
(MW)
Parameters Pig, h, Qig, h active and reactive power output of gth DER at ith bus for
scenario h (MW)
α chp heat-to-power recovery ratio of CHP unit Pci installed capacity of modular DER at ith bus (MW)
Δt optimization time-step Pil, Qil active/reactive load at ith bus (MW)
ηsc , ηsd charging and discharging efficiency of storage technology s Pil, h, Qil, h active/reactive load at ith bus for scenario h (MW)
φs self-discharge losses for storage technology s (%/Δt) Pipv installed capacity of photovoltaic at ith bus (MW)
γij heat loss coefficient for heat transfer pipe i–j (%/m)
Pi,pvh , Qi,pvh active and reactive power output of photovoltaic at ith bus
λ i,i - j th
correlation factor of i bus and branch i–j for scenario h (MW)
μ ac, μ ec coefficient of performance for absorption chiller and Piu, h unsupplied loads at ith bus for scenario h (MW)
electric chiller Piwt installed capacity of wind turbine at ith bus (MW)
ar g annuity rate for gth DER technology Piwt wt
, h , Qi, h active and reactive power output of wind turbine at ith bus
cu penalty cost of unsupplied loads ($/MW) for scenario h (MW)
ccd capital cost of non-modular DER technology ($/MW) vi, h voltage at ith bus for scenario h
cr ¯s
¯ s, dr maximum charge and discharge rate of storage technology
Abbreviation
s (% of capacity)
ec g emissions rate of gth DER technology (kg/MWh)
CHP combined heat and power
fcc,vcc fixed capital cost ($) and variable capital cost ($/MW) of
CIU/ECIU cost of ignoring uncertainty/expected cost of ignoring
modular DER technology
uncertainty
E̲ s , Ēs minimum and maximum energy stored in storage tech-
DER distributed energy resource
nology s (MWh)
DG distributed generation
h¯ij maximum heat transfer capacity of pipe i–j (MW)
HMM heuristic moment matching
ī upper limit on branch current
ICE internal combustion engine
Nh ,Ny number of scenarios and hours of a year
MMIP multi-energy microgrid investment planning
oc g operation cost of gth DER technology ($/MWh)
PV photovoltaic
P¯di rated discrete capacity of non-modular DER at ith bus
WT wind turbine
(MW)

combined-heat-and-power (CHP) units improves the overall fuel utili- modelling of the uncertainties [15] and the energy flows in the multi-
zation efficiency, significantly improving the economic feasibility of the energy microgrids [16].
microgrids [10]. In the recent years, several research works have investigated the
Despite the several benefits of the renewable distributed generation, optimal planning and design of the DER technologies in the microgrids.
the intermittent power supply of the non-dispatchable wind turbines The authors in [17] have proposed a microgrid planning algorithm to
(WTs) and the photovoltaics (PVs) impose uncertainties on the micro- minimize the overall investment and operation costs. The two-stage
grid planning and operation [11]. These uncertainties, along with the robust planning model in [18] determines the optimal siting, sizing and
load variability result in several operational and control problems, such mix of the DGs in the microgrid, whilst minimizing the total costs and
as the protection degradation, transient instability, voltage rises and the CO2 emissions. In [19], the optimal siting and sizing of the DERs is
elevated fault levels [12]. When considering the planning and operation determined based on both the technical and economic criteria. The
of the multi-energy microgrids, the operational and control complexity viability of a residential energy supply system that comprises of several
is further aggravated by the energy flows between the various DERs and CHP units is investigated in [20], based on the optimal operational
the loads pertaining to the electrical, heating and cooling energy [5]. planning approach. In [21], the optimal operation strategy of an in-
The uncertainties associated with the renewable DG can be handled to dustrial energy supply network is investigated, which consists of a
some extent through the utilization of the distributed storage systems district heating network, a solar thermal plant with heat storage, sev-
that store energy during the low-demand periods, while supplying back eral CHP units, boilers and compression chillers. The optimal planning
the energy during the high-demand periods [13]. However, advanced and operation of a CHP-based energy network is examined in [22] to
planning and modelling techniques as well as innovative energy storage reduce the costs and the CO2 emissions in a remote community. How-
and energy management systems [14] are still required for the adequate ever, these works concentrated on a specific sub-problem without

1278
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

investigating the complete problem. The planning models in [17–19] electrical, heating and cooling energy demands. To the author’s best
determine the optimal siting, sizing and dispatch of DER technologies knowledge, the comprehensive microgrid planning model proposed in
taking the electrical energy flows into account, but without considering this work has not been reported in the existing literature. The main
the heating and cooling energy demands. The heating energy flows and technical contributions made in this work are summarized as follows:
demand are studied for the planning of district heating networks in (1) the uncertainties of the wind and photovoltaic generation as well as
[20,21]; however, the electrical loads are ignored. Although the elec- the electrical, heating and cooling energy demands are explicitly
trical loads are investigated for the planning of a district heating net- modelled using an uncertainty matrix based on the heuristic moment
work in [22], the electrical power flow equations are neglected. matching method; (2) the uncertainty matrix is incorporated to for-
A number of planning studies have investigated the modeling of mulate the scenario-based stochastic MMIP model for determining the
both electrical and heating networks pertaining to the multi-energy optimal mix, siting and sizing of the electrical, heating and cooling
microgrids. In [23], the optimal allocation of electrical and heating DER DERs in the isolated microgrids, whilst minimizing the investment and
mix is studied to meet the energy demand of a cluster of buildings; operation costs as well as the CO2 emissions; and (3) the performance of
however, the electrical energy is modeled as a commodity and the the proposed MMIP model-based solution is assessed through a case-
power flow constraints are ignored. The optimal planning and opera- study in the 19-bus microgrid test system, followed by a comparative
tion of an urban DER system is examined in [24] for the overall cost analysis against the deterministic planning solution.
reduction and reliable operation, but without taking the power flow The proposed planning approach effectively incorporates the power
constraints into account. In [25], the planning of the urban energy flow and heat transfer equations to model the complex energy flows
networks is investigated with the consideration of various city sizes and between the electrical, heating and cooling energy sources and loads.
technology scenarios, but power flow equations are ignored. In [26], The DER mix under study is comprised of the wind turbines, photo-
the integrated design of the heating, cooling and electricity networks is voltaics, battery storage units, gas-fired boilers, thermal storage tanks,
investigated, but the electrical power flows are not examined. The DER combined-heat-and-power units, electric chillers and absorption chil-
planning formulation in [27] has adopted linear power flow and heat lers. The technical results obtained from the case-study in the 19-bus
transfer equations; however, the cooling energy flows are not studied. microgrid demonstrate that the proposed uncertainty matrix provides
The modelling of electrical and heating networks is presented in [28] an adequate representation of the historical wind and photovoltaic
for the optimal planning of the CHP-based DERs, but the cooling energy generation as well as the historical electrical, heating and cooling en-
flows are not investigated. It should be highlighted that the afore- ergy demands. Moreover, the proposed MMIP model significantly re-
mentioned planning methodologies primarily concentrated on a spe- duces the overall microgrid costs and the CO2 emissions. The optimal
cific sub-problem without considering the complex multi-energy flows electricity, heating and cooling dispatch for two representative days,
in the microgrid. Although the previous works [9,16] have investigated i.e. a summer sunny day and a winter cloudy day demonstrates that the
the complex multi-energy flows in the microgrid, these planning for- proposed planning model effectively contemplates the complex energy
mulations have considered simplified typical-day profiles to model the flows in the multi-energy microgrid. The superiority of the proposed
energy demand, without taking into account the uncertainties of re- planning solution is further validated using the deterministic planning
newable generation and energy demand. solution as the comparison benchmark.
To deal with the uncertainties, more than a few methodologies have The remainder of the work is structured as follows: Section 2 pre-
been exploited in the existing literature, such as the stochastic opti- sents the modeling of the multi-energy microgrid system and the HMM-
mization [29], robust optimization [30], Monte Carlo simulation [31], based uncertainty matrix, along with the formulation of the proposed
Latin Hybercube Sampling technique [32], Taguchi’s orthogonal array MMIP optimization problem; Section 3 discusses the numerical results
testing [30] and probability statistical methods [33–35]. However, obtained from the case-study in the 19-bus test microgrid; and finally,
these methods can be computationally inefficient; for instance, the Section 4 presents the conclusive remarks and a few insights for the
Monte Carlo simulation in [31] generates excessive scenarios leading to future work.
computational complexity. The occurrence probability of the scenarios
based on Taguchi’s orthogonal array testing [30] is insufficient, 2. The scenario-based stochastic MMIP model
whereas selecting an appropriately-sized uncertainty set for the effi-
cient characterization of uncertainties is a not a trivial task in robust The framework of the scenario-based stochastic MMIP model is
optimization [30]. In comparison, the heuristic moment matching summarized in Fig. 1. Firstly, the uncertainty matrix comprising of the
(HMM) method [36] significantly improves the computational effi- representative scenarios of WT/PV generation and electrical/heating/
ciency by exploiting a reduced number of scenarios [37]. The HMM cooling demand is generated using the HMM method [36]. Then, the
method also outperforms the conventional moment matching method uncertainty matrix is incorporated to formulate the MMIP problem in
by reducing the computational complexity of high-dimensional discrete the YALMIP modelling language [41]. Finally, the solution of MMIP
variables. Although the HMM method has been implemented in [37] to problem is obtained in the CPLEX-MATLAB interface [42].
contemplate the stochastic behavior of wind turbine generation in
transmission network planning, the uncertainties of photovoltaic gen- 2.1. Multi-energy microgrid system model
eration and load demand are ignored. In addition, the uncertainties of
the wind and PV generation have been modelled using the HMM The multi-energy microgrid system model is shown in Fig. 2. The
method in [38–40]. In [38], the distributed generation planning demand at each bus is composed of the electrical loads, heating loads
method minimizes the power losses and the voltage deviation in the (i.e. water and space heating) and cooling loads. The electrical loads are
distribution system. The distributed generation investment planning supplied by a mix of CHP-enabled internal combustion engines (ICE),
model in [39] maximizes the net present value of the distribution wind turbines, photovoltaics and battery storage units. A mix of the gas-
network operator. The investment planning model proposed in [40], fired boilers, recovered heat from CHP-ICE and thermal storage tanks
utilizes the distributed generation and the storage arbitrage benefit to deliver the heating loads, whereas the cooling loads are delivered by
maximize the distribution network operator’s profit. However, the the combination of electric chillers and absorption chillers. As sug-
multi-energy sources as well as the electricity, heating and cooling gested in [9,16], the installed capacity of a DER technology is modeled
loads are not investigated in these works. using either a continuous variable or discrete variable. A continuous
To this end, this paper proposes a scenario-based multi-energy mi- variable is used when the DER technology is commercially available in
crogrid investment planning (MMIP) model to deal with the un- small-scale modules, referred to as the modular DERs in this study, i.e.
certainties of the wind and photovoltaic generation as well as the photovoltaics, boilers, electrical/heating storage units and electric/

1279
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

Start study, i.e. WT/PV generation and electrical/heating/cooling demand


are associated with uncertainties due to their diurnal, seasonal and
Input historical data for WT/PV generation and yearly variations. To deal with these uncertainties, a reasonable
electricity/heating/cooling demand number of scenarios should be considered because a complete set of
historical scenarios results in computational intractability [43]. In this
work, we adopt the HMM method [36] to model the uncertainty matrix
Calculate target moments and correlation matrix
consisting of a reduced number of scenarios characterized by a similar
degree of stochasticity as that of the historical scenarios. The HMM
Normalize target moments[eq. (1)] method employs the matrix and cubic transformations to capture the
Generation of uncertainty matrix

correlation and stochastic moments among the historical scenarios,


respectively [37]. The first four moments, i.e. expectation, standard
Randomly generate scenarios
deviation, skewness and kurtosis are typically considered sufficient for
(HMM method)

retaining the stochastic features of historical scenarios [36]. The pro-


Apply matrix transformation to satisfy target cess is described as follows:
correlation matrix [eq. (2)] Step 1: Firstly, the target moments and target correlation matrix (R)
N
of the historical hourly WT/PV generation and demand are determined,
Apply cubic transformation to satisfy target and normalized [36], as given in (1), where MiNT T
, k and Mi, k are the k
th

normalized moments[eq. (3)-(4)] normalized moment and k th target moment of ith column vector, re-
spectively. k = 1, 2, 3, 4 refer to the expectation, standard deviation,
error criteria is met? skewness and kurtosis, respectively, while i = 1, 2, 3, 4, 5 refer to WT
[eq. (5)-(6)] generation, PV generation, electrical load, heating load and cooling
load, respectively.
Y
⎧ MiNT
,1 = 0, MiNT
,2 = 1

Invert normalized scenarios to satisfy target Mi
T
,3 MiT,4
moments and output uncertainty matrix[eq. (7)] ⎨ MiNT
,3 = , MiNT
,4 = (M T ) 4
⎪ T
( Mi,2 ) 3
i,2
(1)

Integrate uncertainty matrix with deterministic where i = 1, 2, 3, 4, 5; k = 1, 2, 3, 4 .


MMIP problem formulation

decision variables [eq. (8)] Step 2: Nh scenarios of Nu uncertainty factors, i.e. WT gen-
eration(X 1) , PV generation(X 2) , electrical load (X 3) , heating load
Define objective function (X 4) and cooling load (X 5) are randomly generated to obtain the n-di-
(YALMIP)

[eq. (9)-(10)] mensional random matrix XNh × Nu [36].


Step 3: Matrix transformation [36], as given in (2), transforms
Define constraints for
X Nh × Nu into the n-dimensional matrix YNh × Nu to satisfy R , where L is the
electricity network [eq. (11)-(16)],
lower-triangle matrix ofR determined via Cholesky decomposition.
heating network [eq. (17)-(18)],
cooling network [eq. (19)] i
⎧Y = L × X = ∑ j = 1 Lij × Xi
storage [eq. (20)-(22)]
⎨ R = LLT (2)

Input test system data and simulation Step 4: Cubic transformation [36], as given in (3), transforms YNh × Nu
parameters into n-dimensional matrix of normalized scenarios Z Nh × Nu to satisfy
MiNT
, k . ai , bi , ci, di in (3) are calculated assuming that the moments of
Determine DER mix,
MMIP Solution

target scenarios(Mi, k (Zi )) are equal to MiT, k , as given in (4).


siting and sizing
(CPLEX)

Zi = ai + bi Yi + ci Yi2 + di Yi3 (3)


N
Constraints are Mi, k (Zi ) = MiT, k (4)
satisfied ?
Step 5: The solution converges when the moment error (εm) and
Y correlation error (εc ) , as given in (5) and (6), respectively, are less than
predefined thresholds, i.e. ε¯m = 5%, ε¯c = 5%. MikG in (5) is the k th mo-
Output optimal DER mix, ment of ith generated column vector Zi . In (6), RilG is correlation matrix
siting and sizing of generated scenarios and RilNT is target correlation matrix of historical
scenarios.
End Nh 4
εm = ∑ ⎛⎜|MiG1 − MiNT G NT NT ⎞
1 | + ∑ |Mik − Mik |/ Mik ⎟
Fig. 1. The framework of the proposed MMIP model. i=1 ⎝ k=2 ⎠ (5)

Nh Nw Nw
absorption chillers. It is worth mentioning that the adoption of con- 2
tinuous variables improves the computational efficiency. Discrete
εR = ∑ Nu (Nu − 1)
∑ ∑ (RilG − RilNT )2
i=1 i=1 i=1 (6)
variables are used for rest of the DERs, referred to as the non-modular
DERs in this study, i.e. CHP-internal combustion engines and wind Step 6: Finally, Z Nh × Nu is inverted to satisfy MiT, k ,
and obtain the
turbines. uncertainty matrixΩH = [Phwt , Phpv , Phel, Phhl, Phcl] containing scenarios of
WT generation (Phwt ) , PV generation (Phpv ) , electrical load (Phel ) , heating
2.2. Modeling of uncertainty matrix load (Phhl ) and cooling load (Phcl ) , as given in (7).
T
ΩH = Mi,2 × Zi + MNT
i,3 (7)
The input parameters of the MMIP optimization problem under

1280
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

Fuel
Boiler P bl

Internal Heat recovery αchp P chp


CHP combustion engine system
P chp
pv
Thermal storage tank P hs
Photovoltaic P Electric energy flow
Heat energy flow
Pwt P ac/μac Cooling energy flow
Wind turbine
ec
P /μec Electric Absorption
P es
Battery chiller chiller
P ec P ac

Electric load Cooling load Heating load


el cl
P P P ht
Fig. 2. Illustration of the multi-energy microgrid system model.

2.3. The optimization model ∑ (λ i,i - j pij, h ) + ∑ Pig, h + Piu, h + ηsd = es dris, h= es
i - j ∈ ΩL g ∈ [wt,pv,chp]

The proposed MMIP optimization model aims to minimize the in- Pig, h= ec cris, h= es
vestment and operation costs as well as the CO2 emissions by de- = Pil,=h el + + Pijloss
,h +
μ ec ηsc = es (11)
termining the optimal DER mix, siting and sizing, while considering the
uncertainties of WT/PV generation and electrical/heating/cooling de- ∑ (λ i,i - jqij, h) + ∑ Qig, h + Qiu, h = Qiloss l = el
, h + Qi, h
mand. To capture the aforementioned uncertainties, the uncertainty i - j ∈ ΩL g ∈ [wt , pv, chp] (12)
matrix is integrated with the deterministic decision
variables(Piwt , Pipv , Piel, Pihl, Picl ) , as given in (8). v j2, h = vi2, h − 2 (rij pij, h + xij qij, h) + (r 2ij + x 2ij) iij, h (13)

⎧ Pig, h∈ [wt , pv] = Pig ∈ [wt , pv]× Phg ∈ [wt , pv] vi2, h iij, h ⩾ pij2, h + qij2, h (14)
⎨ Pil,∈h [el, hl, cl] = Pil ∈ [el, hl, cl] × Phl∈ [el, hl, cl]
⎩ (8) v̲ ⩽ vi, h ⩽ v̄ (15)

|iij, h| ⩽ ī (16)
2.3.1. Objectives
The overall costs objective in (9) accounts for the annualized in- The heating network constraints (17)–(18) are adapted from [16]. The
vestment costs of non-modular and modular DERs (where annuity rate heat balance constraint (17) guarantees the nodal power balance be-
depends on interest rate and DER lifetime), the operation costs and the tween the total heat supply and demand, considering the net heat
costs of unsupplied loads. The CO2 emissions objective in (10) considers transfer from other buses, boiler heat, CHP recovered heat, thermal
the CO2 emissions from the operation of all DERs. As suggested in [16], storage discharging/charging, heating load, heat consumption of ab-
the operation costs and the CO2 emissions in (1) and (2), respectively, sorption chiller and heat transfer losses. Eq. (18) restricts the capacity
are scaled up from H-scenarios to 8760 h of a year using the weight of the heat piping network.
factor (w = Ny / Nh) , since the occurrence probabilities of the H-scenarios
Pig, h= ac
are uniform. ∑ (λ i,i−j hij, h) + Pig, h= bl + α chpPig, h= chp + ηsd= hs dris, h= hs = Pil,=j hl +
i− j∈ ΩL μ ac
Ccost = ∑i, c nid . P¯id . cc d. ar d + ∑i, d (fc c . bic + vc c . Pic ). ar c
cris, h= hs
Ny
+ N ∑i, g, h Pig, h. oc g +
Ny
∑i, h Piu, h. c u
+ ∑ γij hij, h +
h Nh (9) i − j ∈ ΩL ηsc= hs (17)

Ny 0 ⩽ hij, h ⩽ h¯ij (18)


CCO2 =
Nh
∑ Pig,h. ec g
i, g , h (10) The cooling network constraint (19) ensures that the demand at each
bus is balanced by the energy supplied from the absorption chillers and
electric chillers [16].
2.3.2. Constraints
The objectives in (9) and (10) are subject to the following con- ∑ Pig, h = Pil,=h cl
straints. The electrical network security constraints (11)–(16) adopt the g ∈ [ac, ec ] (19)
mixed-integer quadratically constrained programming based power
The constraints (20)–(22) for the dispatch operation of electrical and
flow model [44]. The constraint (11) enforces the nodal power balance
heat storage technologies are adapted from [16]. The energy stored in
between the total active power supply and demand, accounting for the
the storage units considering self-discharge losses is defined in (20) and
net injected power from other buses, outputs of WTs, PVs and CHPs,
bounded in (21). The charging and discharging rates of the storage
unsupplied loads, battery discharging/charging, electrical loads, elec-
units are constrained, as given in (22).
tric chiller consumption and power losses. Similarly, the constraint (12)
ensures the nodal power balance between the total reactive power Eis, h = (1 − φs) Eis, h − 1 + cris, h Δ t− dris, h Δt (20)
supply and demand. The constraints (13) and (14) define the bus vol-
tage and branch current, respectively, in terms of active and reactive E̲ s ⩽ Eis, h ⩽ Ēs (21)
power flows. The constraints (15) and (16) restrict the bus voltage and
cris, h ⩽ ¯ sCis,
cr dris, h ⩽ ¯ sCis
dr (22)
branch current, respectively.

1281
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

3. Case-study solution (Case II) in the 19-bus microgrid. The overall costs stabilize at
7.82 M$ when the number of scenarios reaches 80; thereafter, the
In this section, the effectiveness of the proposed MMIP model-based computation time increases without any variation in the overall costs.
solution is evaluated and validated through a case-study in the 19-bus Although the 70 scenarios also provide a close value of the overall costs
isolated microgrid. objective with less computation time, the 80 scenarios ensure the
convergence of the costs objective. Since the computational accuracy
3.1. Simulation setup takes precedence over the computational efficiency from the perspec-
tive of the long-term system planning [29,40], 80 scenarios are selected
To assess the performance of the proposed MMIP model, two test to obtain the planning solution in the 19-bus microgrid (see Section
cases with different objectives are considered as follows: Case I: the 3.3).
proposed MMIP model considering the minimization of overall costs
objective, as given above in Eq. (9); Case II: the proposed MMIP model
3.3. Numerical results
considering the minimization of the composite costs/emissions objec-
tive, as given above in Eqs. (9) and (10), respectively, where the costs
3.3.1. The proposed MMIP model based solution
and the emissions are allotted equal weights in the optimization ob-
The performance of the proposed MMIP solution using 80 scenarios
jective.
is evaluated through a case-study in the 19-bus test microgrid. The
The test system considered for the case-study is a 19-bus isolated
numerical results are summarized in Table 4, showing the optimal DER
microgrid [16] in which the electric power lines and the heat transfer
capacities and locations, the annualized investment costs, the annual
pipes are organized in a radial structure, as illustrated in Fig. 3. Table 1
costs of the operation and the unsupplied loads, and the annual CO2
shows the peak electricity load, the peak heating load and the peak
emissions.
cooling load at various buses in the 19-bus microgrid. All the buses are
In Case I, three CHP-ICE units with a total capacity of 6 MW are
considered as the candidate for the placement of DER units. An interest
installed, whereas the aggregate capacity of the installed gas boilers,
rate of 7% is considered. The investment parameters for the modular
the electric chillers and the absorption chillers is 0.2 MW, 2.6 MW and
DER technologies including the fixed and variable capital cost, the
1.7 MW, respectively. It should be noted that the wind and photovoltaic
operation cost, the technology lifetime, and the CO2 emissions rate are
units are not installed in Case I that only considers the costs objective.
adapted from [9,29], as given in Table 2. Moreover, the investment
This is due to the fact that these non-dispatchable renewable generators
parameters for the non-modular DER technologies including the rated
cannot ensure a fixed power output, resulting in cost-ineffectiveness
capacity, the capital cost, the operation cost, the CHP heat-to-power
compared to the dispatchable CHP units.
recovery ratio, the technology lifetime, and the CO2 emissions rate are
The optimal solution in the Case II that considers the equally-
extracted from [9,29], as given in Table 3.
weighted costs/emissions objective is quite different from the Case I.
The lithium-iron-phosphate battery model with deep discharge
Firstly, Case II allocates a significant capacity of the renewable gen-
technology is employed for electricity storage, which is specifically
erators to minimize the CO2 emissions. The aggregate capacity of the
designed for the photovoltaic and renewable installations, as suggested
installed wind and photovoltaic units is 2 MW and 1.2 MW, respec-
in [45,46]. Moreover, it should be noted that the CO2 emissions from
tively. Secondly, a 0.5 MW battery system is installed to store the sur-
wind and PV technologies refer to the life-cycle emissions, which are
plus generation of the non-dispatchable renewable generators. Thirdly,
produced when the wind turbines and PV modules are manufactured,
the aggregate CHP capacity is decreased from 6 MW (Case I) to 5 MW,
built, maintained and decommissioned. Although the life cycle CO2
whereas the installed boiler capacity increases to 1.6 MW to make up
emissions from wind and PV technologies are substantially lower than
for the reduced CHP capacity. Lastly, the aggregate capacity of the
the emissions from combustion-based natural gas and coal technologies;
electric chillers and the absorption chillers is the same as in the Case I.
still, the emissions from the wind turbines and PV units should be taken
It can be observed that the decrease in the CHP capacity from 6 MW
into account, as advised in [29,47].
(Case I) to 5 MW (Case II) resulted in several capacity additions in the
Case II, i.e. 2 MW of wind turbines, 1.2 MW of PVs, 0.5 MW of battery
3.2. The generation and selection of scenarios
and 1.6 MW of boiler capacity. This is because the non-dispatchable
wind and PV capacity cannot guarantee a fixed power output; hence,
For the generation of the scenario-based uncertainty matrix, the 3-
Case II requires the support of the battery storage and the additional
year historical hourly data of the wind and PV generation as well as the
boiler capacity to make up for the reduced CHP capacity. As a result of
electricity, heating and cooling demand is extracted from [48]. The
these capacity additions, although the investment costs elevate from
HMM method is employed to generate different number of scenarios,
1.29 M$ (Case I) to 2.09 M$ (Case II), the operation costs are reduced
ranging from 10 to 100. Fig. 4 demonstrates the effectiveness of the
from 5.98 M$ (Case I) to 5.73 M$ (Case II) due to the lower operation
HMM method by comparing the stochastic features of the generated
costs of the wind and PV units. In total, there is only a 3.4% increase in
scenarios with those of the historical scenarios. It is observed that the
the overall costs in Case II compared to Case I, i.e. from 7.56 M$ (Case I)
moment errors (i.e. pertaining to the expectation, the standard devia-
to 7.82 M$ (Case II). Moreover, the optimal solution in Case II sig-
tion, the skewness and the kurtosis) between the generated scenarios
nificantly reduces the CO2 emissions by 15.2% compared to Case I;
and the historical scenarios reduce significantly as the number of sce-
hence, leading to further cost reductions that depend on the considered
narios increases. This is because the greater number of scenarios can
social cost of the CO2 emissions.
cover a wider range of the historical data; hence, offering a better re-
presentation of the uncertainties, as suggested in [36]. From these re-
Electric power line
sults, it is confirmed that the uncertainty matrix effectively transforms 18 19 Heat transfer pipe
the historical scenarios into a reduced number of scenarios that possess
16 17
the similar degree of stochasticity as that of the historical scenarios.
The number of scenarios can be selected depending upon the opti- 1 12 13 14 15
mization problem under consideration, e.g., planning or operation. In 2 11
this work, the minimum number of scenarios that ensure the accuracy 9 10
of the proposed MMIP solution are determined by comparing the ob- 3 4 5 6 7 8
jective function value and the computation time. For example, Fig. 5
shows the overall costs objective and the computational time of MMIP Fig. 3. Topology of the 19-bus test microgrid.

1282
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

Table 1 Table 3
Peak electricity load, heating load and cooling load in the 19-bus test microgrid. Investment parameters for the non-modular DER technologies.
adapted from [9,29]
Bus number Energy Demand
Item Wind turbine CHP-ICE
Electricity load (kW) Heating load (kW) Cooling load (kW)
Rated capacity (MW) 1 1
2 127 263 276 Capital cost ($/MW) 2,640,000 3,074,000
3 191 96 100 Operation cost ($/MWh) 17 145
4 282 190 200 Heat-to-power recovery ratio – 1.019
5 304 382 401 Technology lifetime (years) 20 20
6 127 263 276 CO2 emissions rate (kg/MWh) 27.6 560
7 191 96 100
8 282 190 200
9 304 382 401
10 127 263 276
(i.e. 1–4 am and 9 pm–12 am), and supply electricity during high de-
11 191 96 100 mand (12–3 pm).
12 282 190 200 Fig. 8(a) and (b) show the optimal heating dispatch for the summer
13 304 382 401 sunny day and the winter cloudy day, respectively. The heat supply
14 127 263 276
comes from the gas-boilers and the CHP recovered heat, whereas the
15 191 96 100
16 282 190 200 heat demand is composed of the water and space heating loads as well
17 304 382 401 as the absorption chiller loads. It should be noted that the consideration
18 127 263 276 of the CHP recovered heat interconnects the dispatch of electricity and
19 191 96 100 heating in the multi-energy microgrid. The heat demand is low for the
summer sunny day, as shown in Fig. 8(a); consequently, the boiler heat
supply is not required, and the heat recovered from CHP is sufficient to
Fig. 6 shows a visual representation of the optimal locations (i.e. bus
meet the heat demand. The significantly high demand of the heating
numbers) and the capacities of the distributed energy resources based
energy during the early morning to the late evening (6 am–11 pm) is
on the optimal solution in Case II.
due to the absorption chillers that operate to provide the water and
space cooling. On the other hand, although the absorption chiller loads
are zero for the winter cloudy day (Fig. 8(b)), the heat demand is still
3.3.2. Optimal electricity, heating, and cooling dispatch
high compared to the summer day. Consequently, the heat from both
To demonstrate the optimal electricity, heating and cooling dispatch
the CHP and the boiler is delivered throughout the day to meet the high
based on the proposed MMIP solution (Case II), two representative days
heat demand.
are selected, i.e. a summer sunny day from the month of July and a
Fig. 9(a) and (b) show the optimal cooling dispatch for the summer
winter cloudy day from the month of January. Figs. 7–9 show the op-
sunny day and the winter cloudy day, respectively. The cooling demand
timal dispatch for the electricity, heating and cooling energy, respec-
is supplied by the electric chillers and the absorption chillers. The
tively, indicating that the proposed model effectively considers the
cooling demand is considerably high for the summer sunny day, as
energy flows between the electrical, heating and cooling DERs to meet
shown in Fig. 9(a); as a result, the electric chillers operate throughout
the multi-energy demands in the microgrid.
the day and the absorption chillers supply a part of the cooling load
Fig. 7(a) and (b) show the optimal electricity dispatch for the
from the early morning to the late evening (6 am–11 pm). On the other
summer sunny day and the winter cloudy day, respectively. The elec-
hand, the cooling demand is substantially low for the winter cloudy day
tricity is supplied by the mix of wind turbines, PVs, CHP-ICEs and
(Fig. 9(b)); consequently, only the electric chiller operation is sufficient
battery units, whereas the electricity demand is composed of the main
to meet the cooling demand and the absorption chiller operation is not
electricity loads and the electric chiller loads. For the summer sunny
required.
day (Fig. 7(a)), the maximum PV output occurs during the afternoon
(10 am–2 pm) and the wind generation peaks during the early morning
(1–5 am), while the CHP units operate throughout the day. The peak 3.3.3. Comparative analysis of the proposed MMIP solution against the
demand occurs during the evening (5–9 pm). The battery system stores deterministic planning solution
electricity during the early morning and late night (i.e. 1–4 am and The proposed MMIP solution utilizes a reduced number of scenarios
9 pm–12 am), while supplying electricity during high demand (i.e. 80 scenarios, see Section 3.2) with a similar degree of stochasticity
(2–5 pm). In the case of winter cloudy day (Fig. 7(b)), the PV output is as that of the historical wind/photovoltaic generation and electrical/
lower during the daytime and the sun-hours reduce significantly (7 heating/cooling demand. On the other hand, the deterministic planning
am–4 pm), compared to the summer day (5 am–8 pm). The wind gen- solution ignores the uncertainties by assuming perfect data for the
eration is greater throughout the day, making up for the reduced PV generation and demand [29]. In this section, the performance of the
generation. The total electricity demand is lower due to the reduced proposed MMIP solution (Case II) is further assessed through a com-
electric chiller loads; moreover, the peak demand shifts to afternoon parative analysis against the deterministic planning solution (Case II).
(2–5 pm). The CHP units again operate throughout the day, while the As shown in Fig. 10(a) and (b), the overall costs and the CO2 emissions,
battery system stores electricity during the early morning and late night respectively, obtained from the deterministic solution fluctuate subject

Table 2
Investment parameters for the modular DER technologies.
adapted from [9,29]
Item Photovoltaic Gas boiler Electric chiller Absorption chiller Battery storage

Fixed capital cost ($) 2500 6000 2300 250 500


Variable capital cost ($/MW) 2,500,000 45,000 230,000 250,000 400,000 $/MWh
Operation cost ($/MWh) 40 100 40 30 20
Technology lifetime (years) 20 10 10 20 15
CO2 emissions rate (kg/MWh) 58.4 560 – – –

1283
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

Fig. 4. Percentage moment errors between generated scenarios and historical scenarios.

Fig. 5. Selection of scenarios for the solution of MMIP problem (Case II).

Table 4
MMIP Solution: Optimal mix, siting and sizing of the DER technologies, the overall costs and the CO2 emissions.
Item Case I: Proposed MMIP model considering the Case II: Proposed MMIP model considering the minimization of
minimization of overall costs objective only composite costs/emissions objective

Optimal DER Mix, Siting and Wind-turbine 0 2 MW [1–bus 3,1–bus 13]


Sizing Photovoltaic 0 1.2 MW [0.4–bus 9, 0.8–bus 5]
CHP-ICE 6 MW [3–bus 1, 2–bus 2, 1–bus 4] 5 MW [3–bus 1, 1–bus 2, 1–bus 4]
Battery 0 0.5 MW [0.3–bus 1, 0.2–bus 2]
Gas boiler 0.2 MW [0.2–bus 1] 1.6 MW [1.2 MW-bus 1, 0.4 MW-bus 2)
Electric chiller 2.6 MW [1.3–bus 1, 0.8–bus 2, 0.5–bus 16] 2.6 MW [1.3–bus 1, 0.8–bus 2, 0.5–bus 16]
Absorption chiller 1.7 MW [0.8–bus 1, 0.5–bus 4, 0.4–bus 12] 1.7 MW [0.8–bus 1, 0.5–bus 4, 0.4–bus 12]

Overall costs (M$) Investment 1.29 2.09


Operation 5.98 5.73
Unsupplied loads 0 0
Total 7.56 7.82

CO2 emissions (tons) 18,563 15,729

0.3 0.5 Wind-turbine


18 19
Photovoltaic
1.2
1 Electric chiller
0.2 16 17
0.8 0.4 Battery
0.4
1.3 12 13 14 15 CHP-ICE
1
Gas boiler
2 11 0.4
Absorption chiller
1 9 10 Electric power line
3
Heat transfer pipe
0.8 3 4 5 6 7 8
0.5
0.8
1 1

Fig. 6. Visual representation of optimal DER siting and sizing in the 19-bus test microgrid (Case II).

to the scenario under consideration. As shown in Fig. 10(a), the overall more than that of the proposed MMIP solution. Moreover, as shown in
costs go as high as 9.47 M$ (scenario# 12) and as low as 8.16 M$ Fig. 10(b), the CO2 emissions go as high as 18,611 tons (scenario# 12)
(scenario# 32) for the deterministic solution, compared to that of and as low as 15,842 tons (scenario# 32) for the deterministic solution,
7.82 M$ (see Table 4) for the proposed solution. The mean of the compared to 15,729 tons (see Table 4) for the proposed solution. The
overall costs for the deterministic solution is 8.77 M$, which is 12.2% mean of the CO2 emissions is 17,331 tons, which is 10.2% more than

1284
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

(a)

(b)
Fig. 7. Optimal electricity dispatch based on the proposed MMIP solution in Case II (a) summer sunny day in July (b) winter cloudy day in January.

(a)

(b)
Fig. 8. Optimal heating dispatch based on the proposed MMIP solution in Case II (a) summer sunny day in July (b) winter cloudy day in January.

that of the proposed solution. Therefore, it is established that the pro- scenarios. When considering the overall costs, the maximum CIU is
posed MMIP solution leads to higher reductions in the overall costs and 1.61 M$ (scenario# 12), while the minimum CIU is 0.20 M$ (scenario#
the CO2 emissions, compared to the deterministic planning solution. 32), as shown in Fig. 10(a). The ECIU is 0.95 M$, which is 12.2% of the
The significance of considering uncertainties in the proposed MMIP overall costs obtained from the proposed solution (i.e. 7.82 M$) and
model can be assessed by employing the stochastic metric known as the cannot be ignored. In the case of CO2 emissions, the maximum CIU is
cost of ignoring uncertainty (CIU), as suggested in [29]. For a given 2882 tons (scenario# 12), while the minimum CIU is 113 tons (sce-
scenario, the CIU is computed by subtracting the deterministic solution nario# 32), as shown in Fig. 10(b). The ECIU is 1602 tons, which is
(i.e. for that scenario) from the MMIP solution. The expected cost of 10.2% of the CO2 emissions obtained from the proposed solution (i.e.
ignoring uncertainty (ECIU) is the weighted sum of CIU values across all 15,729 tons) and is not negligible. This shows that the cost of ignoring

1285
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

(a)

(b)
Fig. 9. Optimal cooling dispatch based on the proposed MMIP solution in Case II (a) summer sunny day in July (b) winter cloudy day in January.

(a)

(b)
Fig. 10. Proposed MMIP Solution vs. Deterministic Solution in Case II (a) overall costs (b) CO2 emissions.

uncertainty by the deterministic planning solution is significant; hence, generation and demand in the system planning formulations. This paper
confirming the superiority of the proposed stochastic MMIP model. presented a scenario-based investment planning model for the isolated
multi-energy microgrids considering the electricity, heating, and
4. Conclusion and future work cooling energy sources and loads. The proposed planning solution de-
termined the optimal distributed energy resource mix, siting and sizing
The multi-energy microgrids will become prevalent in the near fu- whilst minimizing the overall microgrid costs and Carbon dioxide
ture, calling for the adequate consideration of the various forms of the emissions. To contemplate the operational uncertainties of wind/

1286
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

photovoltaic generation and electrical/heating/cooling demand, an [9] Mashayekh S, Stadler M, Cardoso G, Heleno M. A mixed integer linear programming
uncertainty matrix was adopted. The uncertainty matrix was generated approach for optimal DER portfolio, sizing, and placement in multi-energy micro-
grids. Appl Energy 2017;187:154–68. https://doi.org/10.1016/j.apenergy.2016.11.
via the heuristic moment matching method which effectively captured 020.
the stochastic moments and correlation among the historical scenarios. [10] Piagi P, Lasseter RH. Autonomous control of microgrids. IEEE Power Eng Soc Gen
To demonstrate the effectiveness of the proposed planning model, a Meet Montreal 2006:1–6. https://doi.org/10.1109/PES.2006.1708993.
[11] Uzunoglu M, Onar OC, Alam MS. Modeling, control and simulation of a PV/FC/UC
case study was conducted using the 19-bus microgrid test system. The based hybrid power generation system for stand-alone applications. Renew Energy
numerical results demonstrated that the proposed uncertainty matrix 2009;34:509–20. https://doi.org/10.1016/j.renene.2008.06.009.
offered an accurate representation of the historical wind and photo- [12] Fang X, Yang Q, Dong W. Fuzzy decision based energy dispatch in offshore in-
dustrial microgrid with desalination process and multi-type DGs. Energy
voltaic generation, along with the historical electrical, heating and 2018;148:744–55. https://doi.org/10.1016/j.energy.2018.01.185.
cooling energy demands. The overall microgrid costs were minimized [13] Karavas C-S, Kyriakarakos G, Arvanitis KG, Papadakis G. A multi-agent decen-
by the proposed planning solution that considered the costs objective. tralized energy management system based on distributed intelligence for the design
and control of autonomous polygeneration microgrids. Energy Convers Manage
When the equally-weighted composite costs/emissions objective was
2015;103:166–79. https://doi.org/10.1016/j.enconman.2015.06.021.
considered, the proposed solution reduced the Carbon dioxide emis- [14] Kuznetsova E, Li Y-F, Ruiz C, Zio E. An integrated framework of agent-based
sions significantly with a slight increase in the overall costs. Moreover, modelling and robust optimization for microgrid energy management. Appl Energy
the optimal electricity, heating and cooling dispatch for two re- 2014;129:70–88. https://doi.org/10.1016/j.apenergy.2014.04.024.
[15] Ehsan A, Yang Q. Optimal integration and planning of renewable distributed gen-
presentative days, i.e. a summer sunny day and a winter cloudy day eration in the power distribution networks: a review of analytical techniques. Appl
demonstrated that the proposed planning model effectively con- Energy 2018;210:44–59. https://doi.org/10.1016/j.apenergy.2017.10.106.
templates the complex energy flows in the multi-energy microgrid. The [16] Mashayekh S, Stadler M, Cardoso G, Heleno M, Madathil SC, Nagarajan H, et al.
Security-constrained design of isolated multi-energy microgrids. IEEE Trans Power
superiority of the proposed planning solution was further validated Syst 2018;33:2452–62. https://doi.org/10.1109/TPWRS.2017.2748060.
using the deterministic planning solution as the comparison bench- [17] Khodaei A, Shahidehpour M. Microgrid-based co-optimization of generation and
mark. Finally, the significance of considering uncertainties in microgrid transmission planning in power systems. IEEE Trans Power Syst 2013;28:1582–90.
https://doi.org/10.1109/TPWRS.2012.2224676.
planning was confirmed using a stochastic metric known as the cost of [18] Wang Z, Chen B, Wang J, Kim J, Begovic MM. Robust optimization based optimal
ignoring uncertainty. DG placement in microgrids. IEEE Trans Smart Grid 2014;5:2173–82. https://doi.
The planning formulation proposed in this work provides an en- org/10.1109/TSG.2014.2321748.
[19] Kroposki B, Sen PK, Malmedal K. Optimum sizing and placement of distributed and
abling solution towards the realization of low-cost and low-carbon renewable energy sources in electric power distribution systems. IEEE Trans Ind
multi-energy networks. Based on the insights obtained from this work, Appl 2013;49:2741–52. https://doi.org/10.1109/TIA.2013.2262661.
several research directions and future challenges should be further [20] Wakui T, Kinoshita T, Yokoyama R. A mixed-integer linear programming approach
for cogeneration-based residential energy supply networks with power and heat
exploited. The proposed planning solution needs to be further validated
interchanges. Energy 2014;68:29–46. https://doi.org/10.1016/j.energy.2014.01.
in large-scale systems consisting of several interconnected microgrids; 110.
moreover, the utilization of proposed model in real-world applications [21] Buoro D, Pinamonti P, Reini M. Optimization of a distributed cogeneration system
needs to be assessed. Furthermore, the impact of electrical generator with solar district heating. Appl Energy 2014;124:298–308. https://doi.org/10.
1016/j.apenergy.2014.02.062.
contingencies on the supply security needs to be investigated by in- [22] Bracco S, Dentici G, Siri S. Economic and environmental optimization model for the
cluding the N-1 security constraints in the optimization model. The design and the operation of a combined heat and power distributed generation
thermal contingencies, such as the boiler outages should also be con- system in an urban area. Energy 2013;55:1014–24. https://doi.org/10.1016/j.
energy.2013.04.004.
sidered in the optimization model. Finally, the proposed planning so- [23] Omu A, Choudhary R, Boies A. Distributed energy resource system optimisation
lution needs to be extended to incorporate the allocation of electric using mixed integer linear programming. Energy Policy 2013;61:249–66. https://
vehicle charging stations. doi.org/10.1016/j.enpol.2013.05.009.
[24] Mehleri ED, Sarimveis H, Markatos NC, Papageorgiou LG. Optimal design and op-
eration of distributed energy systems: application to Greek residential sector.
Acknowledgements Renew Energy 2013;51:331–42. https://doi.org/10.1016/j.renene.2012.09.009.
[25] Keirstead J, Samsatli N, Shah N, Weber C. The impact of CHP (combined heat and
power) planning restrictions on the efficiency of urban energy systems. Energy
This work is supported in part by the Natural Science Foundation of
2012;41:93–103. https://doi.org/10.1016/j.energy.2011.06.011.
China (51777183), the Natural Science Foundation of Zhejiang [26] Yang Y, Zhang S, Xiao Y. Optimal design of distributed energy resource systems
Province (LZ15E070001) and the Natural Science Foundation of coupled with energy distribution networks. Energy 2015;85:433–48. https://doi.
org/10.1016/j.energy.2015.03.101.
Jiangsu Province (BK20161142).
[27] Morvaj B, Evins R, Carmeliet J. Optimization framework for distributed energy
systems with integrated electrical grid constraints. Appl Energy 2016;171:296–313.
References https://doi.org/10.1016/j.apenergy.2016.03.090.
[28] Basu AK, Bhattacharya A, Chowdhury S, Chowdhury SP. Planned Scheduling for
economic power sharing in a CHP-based micro-grid. IEEE Trans Power Syst
[1] Guerrero JM, Vasquez JC, Matas J, de Vicuna LG, Castilla M. Hierarchical control of 2012;27:30–8. https://doi.org/10.1109/TPWRS.2011.2162754.
droop-controlled AC and DC microgrids – a general approach toward standardiza- [29] Santos SF, Fitiwi DZ, Bizuayehu AW, Shafie-khah M, Asensio M, Contreras J, et al.
tion. IEEE Trans Ind Electron 2011;58:158–72. https://doi.org/10.1109/TIE.2010. Novel multi-stage stochastic DG investment planning with recourse. IEEE Trans
2066534. Sustain Energy 2016;8:164–78. https://doi.org/10.1109/TSTE.2016.2590460.
[2] Guerrero JM, Chandorkar M, Lee T-L, Loh PC. Advanced control architectures for [30] Xiang Y, Liu J, Liu Y. Robust energy management of microgrid with uncertain re-
intelligent microgrids—Part I: decentralized and hierarchical control. IEEE Trans newable generation and load. IEEE Trans Smart Grid 2015;7:1034–43. https://doi.
Ind Electron 2013;60:1254–62. https://doi.org/10.1109/TIE.2012.2194969. org/10.1109/TSG.2014.2385801.
[3] Lidula NWA, Rajapakse AD. Microgrids research: a review of experimental micro- [31] Li Y, Feng B, Li G, Qi J, Zhao D, Mu Y. Optimal distributed generation planning in
grids and test systems. Renew Sustain Energy Rev 2011;15:186–202. https://doi. active distribution networks considering integration of energy storage. Appl Energy
org/10.1016/j.rser.2010.09.041. 2018;210:1073–81. https://doi.org/10.1016/j.apenergy.2017.08.008.
[4] Karavas C-S, Arvanitis K, Papadakis G. A game theory approach to multi-agent [32] Wu T, Yang Q, Bao Z, Yan W. Coordinated energy dispatching in microgrid with
decentralized energy management of autonomous polygeneration microgrids. wind power generation and plug-in electric vehicles. IEEE Trans Smart Grid
Energies 2017;10:1756. https://doi.org/10.3390/en10111756. 2013;4:1453–63. https://doi.org/10.1109/TSG.2013.2268870.
[5] Stadler M, Cardoso G, Mashayekh S, Forget T, DeForest N, Agarwal A, et al. Value [33] Qin M, Chan KW, Chung CY, Luo X, Wu T. Optimal planning and operation of
streams in microgrids: a literature review. Appl Energy 2016;162:980–9. https:// energy storage systems in radial networks for wind power integration with reserve
doi.org/10.1016/j.apenergy.2015.10.081. support. IET Gener Transm Distrib 2016;10:2019–25. https://doi.org/10.1049/iet-
[6] Gu W, Wu Z, Bo R, Liu W, Zhou G, Chen W, et al. Modeling, planning and optimal gtd.2015.1039.
energy management of combined cooling, heating and power microgrid: a review. [34] Zheng Y, Hill DJ, Dong ZY. Multi-agent optimal allocation of energy storage systems
Int J Electr Power Energy Syst 2014;54:26–37. https://doi.org/10.1016/j.ijepes. in distribution systems. IEEE Trans Sustain Energy 2017;8:1715–25. https://doi.
2013.06.028. org/10.1109/TSTE.2017.2705838.
[7] Hatziargyriou N, Asano H, Iravani R, Marnay C. Microgrids. IEEE Power Energy [35] Xing H, Cheng H, Zhang Y, Zeng P. Active distribution network expansion planning
Mag 2007;5:78–94. https://doi.org/10.1109/MPAE.2007.376583. integrating dispersed energy storage systems. IET Gener Transm Distrib
[8] Katiraei F, Iravani R, Hatziargyriou N, Dimeas A. Microgrids management. IEEE 2016;10:638–44. https://doi.org/10.1049/iet-gtd.2015.0411.
Power Energy Mag 2008;6:54–65. https://doi.org/10.1109/MPE.2008.918702. [36] Høyland K, Kaut M, Wallace SW. A heuristic for moment-matching scenario

1287
A. Ehsan, Q. Yang Applied Energy 235 (2019) 1277–1288

generation. Comput Optim Appl 2003;24:169–85. https://doi.org/10.1023/ [43] Nick M, Cherkaoui R, Paolone M. Optimal allocation of dispersed energy storage
A:1021853807313. systems in active distribution networks for energy balance and grid support. IEEE
[37] Li J, Ye L, Zeng Y, Wei H. A scenario-based robust transmission network expansion Trans Power Syst 2014;29:2300–10. https://doi.org/10.1109/TPWRS.2014.
planning method for consideration of wind power uncertainties. CSEE J Power 2302020.
Energy Syst 2016;2:11–8. https://doi.org/10.17775/CSEEJPES.2016.00004. [44] Taylor JA, Hover FS. Convex models of distribution system reconfiguration. IEEE
[38] Ehsan A, Yang Q. Robust distribution system planning considering the uncertainties Trans Power Syst 2012;27:1407–13. https://doi.org/10.1109/TPWRS.2012.
of renewable distributed generation and electricity demand. IEEE Conf Energy 2184307.
Internet Energy Syst Integr Beijing 2017:1–6. https://doi.org/10.1109/ei2.2017. [45] Calzadilla TC, Macarulla AM, Borges CE, Vicario AA. Feasibility and simulation of a
8245622. solar photovoltaic installation in DC for standalone services buildings. DYNA Ing E
[39] Ehsan A, Yang Q, Cheng M. A scenario-based robust investment planning model for Ind 2018;93:24–30. https://doi.org/10.6036/8410.
multi-type distributed generation under uncertainties. IET Gener Transm Distrib [46] Calzadilla TC, Macarulla AM, Kamara-Esteban O, Borges CE. Analysis and assess-
2018. https://doi.org/10.1049/iet-gtd.2018.5602. ment of an off-grid services building through the usage of a DC photovoltaic mi-
[40] Ehsan A, Yang Q. Coordinated investment planning of distributed multi-type sto- crogrid. Sustain Cities Soc 2018;38:405–19. https://doi.org/10.1016/j.scs.2018.01.
chastic generation and battery storage in active distribution networks. IEEE Trans 010.
Sustain Energy 2018. [Online]. doi: 10.1109/tste.2018.2873370. [47] Nugent D, Sovacool BK. Assessing the lifecycle greenhouse gas emissions from solar
[41] Lofberg J. YALMIP: a toolbox for modeling and optimization in MATLAB. IEEE Int PV and wind energy: a critical meta-survey. Energy Policy 2014;65:229–44.
Conf Robot Autom, Taipei 2004:284–9. https://doi.org/10.1109/CACSD.2004. https://doi.org/10.1016/j.enpol.2013.10.048.
1393890. [48] U.S. Energy Information Administration. Sources & uses. [Online]. https://www.
[42] IBM ILOG CPLEX. CPLEX Optimizer [Online]. < https://www.ibm.com/analytics/ eia.gov (accessed October 18, 2018).
cplex-optimizer > (accessed October 15, 2018).

1288

You might also like