Electrical Power and Energy Systems 125 (2021) 106418

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

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

Sustainable microgrid design considering blockchain technology for real- T

time price-based demand response programs
Yu-Chung Tsaoa, , Vo-Van Thanha, Qiuwei Wub
⁎

a
Department of Industrial Management, National Taiwan University of Science and Technology, Taipei, Taiwan
b
Department of Electrical Engineering, Technical University of Denmark, Lyngby, Denmark

ARTICLE INFO ABSTRACT

Keywords: In recent years, rapidly increasing electric power demand and organizational concerns regarding social and
Sustainability environmental issues have resulted in significant attention being given to microgrids. However, sustainable
Microgrid microgrids that simultaneously address economic benefits and environmental and social issues have not been
Blockchain technology broadly explored by researches. This study addresses the sustainable microgrid design problem by leveraging
Demand response
blockchain technology to provide the real time-based demand response programs. Three sustainable objectives
Robust multi-objective optimization
Fuzzy programming
(economy, environment, and society) are formulated by a multi-objective mixed integer-linear programming
model. A robust fuzzy multi-objective optimization approach is proposed to determine the optimal number,
location, and capacity of renewable distributed generation units as well as the equilibrium supply and dynamic
pricing decisions under uncertain demand, capacity, and economic, environmental, and social parameters. The
proposed model and solution approach are then applied to a case study in Vietnam. We find that blockchain
technology-based sustainable microgrid can result in a 1.68% and 2.61% increase of profitability and consumer
satisfaction, respectively, and a 0.97% reduction of environmental impacts.

1. Introduction Leveraging blockchain technology to improve the effectiveness of SMGs

Microgrid that is effectively integrated with renewable distributed
generation (RDG) units is considered an efficient solution for reducing
environmental impacts and investment costs; however, they are still
vulnerable to uncertainties caused by the intermittent nature of re­
newable energy resources [1–3]. This has significantly affected the
sustainable properties of the microgrids. The concept of sustainability
strongly relies on three indispensable aspects: economy, environment,
and society [4]. One way to achieve sustainability in the energy sector
is providing demand response (DR) programs [5]. It is defined as
changes in normal energy consumption patterns of the end users in
response to changes in electricity prices [6–7].
In sustainable microgrids (SMGs), DR programs take place using
two-way communication and information technologies, and each in­
dependent system operator (aggregator) is responsible for providing
motivations (incentive bonus or day-ahead prices) for the consumers to
react in DR schemes. Price-based DR programs include time-of-use
pricing, real time pricing, critical-peak pricing, and peak-time rebates
[8]. The increasing application of novel technologies has led to actively
integrate the real-time price-based DR in the energy system operation
because its benefits in maintaining the sustainability of system.
is an essential application for modern power systems.
1.1. Sustainable aspects in microgrids
In the planning of power systems, uncertainties in demand, cost
structure, and environmental and social parameters have imposed
major obstacles in the process of designing and managing SMGs.
Several studies have been presented to incorporate various uncertain
resources into the SMG design problem, focusing on optimizing a single
objective (economy) or dual objective (economy and environment). For
example, a multi-objective robust optimization approach aims at
minimizing the greenhouse gas emissions, economic costs, and failure
probability of electricity supply chain networks under uncertainties in
renewable energy supply and forecasted demand [9]. Network sus­
tainability is created by using RDG units but DR programs have not
been considered in their study. In [10] a multi-objective optimization
approach was presented for minimizing environmental and economic
costs to design sustainable hybrid energy system. Uncertainty is mod­
elled using the method of moments, but DR programs have also been
not considered for social issues in their study. A two-stage stochastic
mixed-integer programming model have presented in [11–12] to design

⁎
Corresponding author.
E-mail address: yctsao@mail.ntust.edu.tw (Y.-C. Tsao).

https://doi.org/10.1016/j.ijepes.2020.106418
Received 17 March 2020; Received in revised form 20 June 2020; Accepted 31 July 2020
0142-0615/ © 2020 Elsevier Ltd. All rights reserved.
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

implemented based on the day-ahead planning. This means that the

Nomenclature
Demand response DR
Renewable distributed generation RDG
Robust fuzzy multi-objective optimization
programming RFMOOP
Robust stochastic optimization programing RSOP
Sustainable microgrid SMG
Sustainable microgrids SMGs
sustainable electrical grids. The authors have considered strategies to
create sustainability by hardening existing lines and building new lines,
and RDG units. However, concerns about the environment and society
were not discussed in their study. A recent study minimized the eco­
nomic, environmental, and social costs using a multiobjective robust
fuzzy stochastic programming approach to design a sustainable smart
grid [13]. Uncertainties and DR programs based on dynamic pricing are
considered in their model. Most of the studies pointed out there has not
considered the real-time price-based DR programs as well as its effects
on the sustainability of microgrids, except the study [13] that men­
tioned DR via day-ahead energy market.
1.2. Demand response programs in microgrids
In recent years, researchers and practitioners have expressed sig­
nificant interest in DR programs [14–20] due to its potential to relieve
stress for distribution network during peak times without adding more
capacity to power plants. In [14–15], a DR mechanism based on the
time-of-use price and real-time price has been explored in a smart
micro-grid environment. The proposed stochastic model with meta­
heuristic algorithm has significantly improved the operation cost of
considered microgrids. Time-of-use price and direct load control have
been combined in a DR program to minimize the eco-environmental
costs and to maximize the micro-grid’s reliability [16]. Microgrids with
dynamic price-based DR program were expected to optimize the flex­
ible and inflexible loads of customers [17]. A metaheuristic algorithm
was used to solve the DR optimization problem.
Table 1 summaries recent DR programs-related recent studies in
microgrids. To the best of our knowledge, the major limitations in the
existing literature are shown as follows: (i) DR programs have been
system operators will send price signals (e.g., time-of-use price and
dynamic price) to the customer in the previous night to perform DR for
the next day. For customers who are willing to participate, they will
have an active plan to react to the program. In the opposite case, cus­
tomers may easily ignore the program. This has happened in many real-
world situations in which customers’ electricity bills are not too high or
DR programs are not sufficiently motivated to encourage customers.
From a practical perspective, the above problem has resulted in the
inefficiency of DR programs and affected the sustainability (social and
environmental goals) of the microgrids; (ii) load control is an important
function in DR programs because it provides information visibility
about supply and demand balance in microgrids. DR only occurs when
the system is out of balance. Most previous studies did not intensively
considered this function. The authors in [16] mentioned the direct load
control in DR, but it was not formulated with mathematical equations
and did not described the operating mechanism explicitly.
1.3. Demand response via blockchain technology adoption
Today, with the development of technology, blockchain technology
can support to implement DR programs in the real-time. Blockchain
technology has a significant potential in managing and operating of
complex systems such as SMGs in a decentralized and transparent
manner [21]. Besides technical barriers, the actual implementation of
DR also faces regulatory and information security issues, for example
monitoring and accounting, especially in the case of large-scale, it will
be difficult to obtain low-cost if centralized management system is still
used [22]. Blockchain technology can not only solve the above pro­
blem, but also provide a transparent monitoring system for load control
in DR programs. Furthermore, the advantages of smart contract in the
blockchain technology (e.g., the cryptology method and the consensus
algorithm) can successfully overcome the obstacles of DR programs
based on the day-ahead planning and provide a real time load control
mechanism for all stakeholders in the SMG.
A number of studies have recognized the role of block technology in
enhancing the sustainability of DR programs in microgrids in recently
[23–27]. In [23], a decentralized blockchain mechanism has been
proposed for delivering transparent, secure, reliable, and timely energy
flexibility. The authors claimed that blockchain-based DR can be used
for the energy supply and demand balance at the grid level. A frame­
work for emerging blockchain technologies was introduced to

Table 1
A sumary of relevant researches.
Reference Objective Demand response Uncertainty decision Approach

Eco Env Soc Scheme Load control Technology platform

[9] ● ● ● No mention No mention Robust

[10] ● ● No mention No mention Simulation
[11] ● Day-ahead Information technology Deterministic
[12] ● No mention Control system ● Stochastic
[13] ● ● ● Day-ahead Information technology ● Fuzzy
[14] ● Day-ahead Intelligent monitoring Deterministic
[15] ● Day-ahead Control system ● Stochastic
[16] ● ● ● Day-ahead ● No mention Two-phase
[17] ● Dynamic No mention Simulation
[18] ● No mention Information technology ● Linear model
[19] ● ● Incentive No mention ● Stochastic
[20] ● ● No mention Control system ● Simulation
[21] ● Real time Public Blockchain Deterministic
[22] ● Real time Blockchain Game theory
[23] ● ● Real time Blockchain ● Simulation
[24] ● ● Real time Blockchain Game theory
[25] ● Real time Blockchain Game theory
[26] ● No mention Blockchain No mention
[27] ● Real time Blockchain No mention
This study ● ● ● Real time ● Public blockchain ● Robust fuzzy

2
Y.-C. Tsao, et al.
guarantee the seamless and secure implementation of DR programs in
microgrids [24]. Smart contracts in blockchain technology was devel­
oped in [25–26] to guarantee a seamless and efficient DR program for a
variety of consumers at local microgrids. Although a number of studies
have explored the potential of blockchain technology in implementing
DR programs, but it has just stopped at proposing the management
frameworks to integrate blockchain technology in existing grids. As can
be seen from Table 1, it lacks an effort to provide a mathematical model
and algorithm to SMG design with DR programs based on the block­
chain technology adoption under uncertain environments. Besides, the
benefits of blockchain technology on the sustainability of microgrid
have not been considered in the literature.
1.4. Motivation and contributions
Although microgrid design is currently of interest to many re­
searchers, the SMGs through real time-based DR programs by lever­
aging of blockchain technology has yet to be considered. From a
practical perspective, DR programs have not really attracted the at­
tention of consumer due to policy, technical, and regulatory barriers.
Besides, the system operators have not been provided information vis­
ibility about demand and generation loads to all network players to
implement DR programs in a timely manner. From a theoretical per­
spective, the DR programs based on day-ahead planning are only ef­
fective in the perfect environment about customer participation will­
ingness. Besides, the load control function in DR programs have not yet
intensively considered in the existing literature. Finally, it lacks a study
that consider all sustainable aspects of the microgrid with blockchain-
based DR programs under uncertain environments.
To address the importance of sustainability issues, this paper de­
velops a new SMG design model based on blockchain technology for
proving DR programs under uncertain environments considering eco­
nomic, environmental, and social goals. An approach based on robust
fuzzy multi-objective optimization programming (RFMOOP) is devel­
oped to determine the decisions related to the optimal number, loca­
tion, capacity, and technology of RDG units, electricity price, and the
amount of generated and transmitted electric power in a SMG. A real
time-based DR program is conducted by smart contracts in blockchain
technology. This model allows us to study the impact of such designs on
the cost efficiency, environmental impacts, and social issues related to
DR programs. It distinguishes contributions from the previous works in
the microgrid design field as follows:
• Practical: This study addresses important issues related to the sus­
tainability of microgrids. A multi-objective mathematical model is
proposed to consider three different aspects: economy, society, and
environment when designing SMGs. The installation and operation
costs of RDG units are used for measuring the economic objective,
the land use and CO2 emission costs are considered for measuring
the environmental objective, and customer-related social issues (i.e.,
consumer satisfaction in electricity bills) are used for measuring the
social objective. To overcome the disadvantages in the DR programs
based on the day-ahead planning, a simulation model for real time-
based DR with the blockchain technology adoption is developed.
The proposed model provides a visual load control tool for all net­
work players in the SMG through the information sharing me­
chanism in the smart contracts. Besides, the profit reward and
penalty cost of RDG units are also formulated in the smart contract
to ensure that the real time-based DR programs are implemented in
the most socially efficient for consumers.
• Theoretical: An RFMOOP model is developed based on the com­
bination of fuzzy possibilistic programming and robust program­
ming through weighted sum aggregation method to solve the de­
scribed multi-objective optimization problem. Compared to other
uncertain approaches that existed in the previous studies (i.e.,
3
Electrical Power and Energy Systems 125 (2021) 106418
stochastic programing, fuzzy programming), the proposed approach
successfully overcome different kind of uncertainty in data (i.e.,
uncertainty for lack of knowledge and epistemic uncertainty) as well
as the flexibility in objective functions and the elasticity in soft
constraints without increasing complex of the original model which
can lead to computational challenging. Moreover, the proposed
model is also capable to provide reliable solutions with optimality
and feasibility guarantee for conflicting multi-objective optimiza­
tion problems under uncertain environments.
The remainder of this paper is organized as follows: The formulation
is presented immediately after the description of the study problem in
Section 2. In Section 3, the solution approach is presented. In Section 4,
a real case of Vietnam’s electricity distribution network in the re­
sidential sector is presented. Section 5 presents the most important
conclusions of the present study.
2. Problem definition and formulation
In this paper, the proposed SMG structure consists of a group of RDG
units that act in isolation, which is shown in Fig. 1. RDG units with
large capacity levels are considered to establish at predetermined po­
tential locations. A hierarchical electricity market model for the re­
sidential sector with the interaction of home users, the microgrid op­
erator, and multiple aggregators is applied to operate the proposed
SMG. In this microgrid, aggregators aggregate electric power from
SMGs to provide residential areas. Since aggregators act as inter­
mediaries between the microgrid operator and the home users, they are
responsible for the network balance by using price-based DR programs,
while the microgrid operator is responsible for the technical validation
of the network (e.g., control load of RDG units). Residential areas are
composed of a group of smart houses equipped with automatic meter
reading systems to conduct price-based DR programs provided by ag­
gregators.
In microgrids, DR programs are provided to consumers based on
two main ways: incentives and prices to relieve stress for the network
during peak times. Price-based DR programs, such as time-of-use, real
time, and critical peak time have been widely applied in many com­
panies (e.g., Xcel Energy, Gulf Power, and Ottawa). Further information
about price-based DR programs can be found in [28]. In this study,
aggregators provide dynamic price-based DR programs to home users at
the previous night. Each aggregator has a different dynamic pricing
scheme to optimize his total profit of the SMG. In the next morning,
based on the price provided by aggregators, home users select an ag­
gregator to receive electricity and implement DR. This operational
model is called day-ahead dynamic pricing-based DR with the support
of automatic meter reading and on/off switch systems in the advanced
electricity markets, such as US, Australia, and Canada.
To address all aspects of the sustainability (economy, environment,
and society) in microgrids, the SMG design problem entails a multi-
objective mathematical model to determine decisions regarding the
number, location, and capacity of the RDG units, amount of electric
power generated and transferred in the network, day-ahead dynamic
electricity pricing, and DR from home users. We assume that energy
demands of all home users are satisfied. The potential locations and
generation technologies to install RDG units are predetermined. Energy
losses in power transmission and distribution are not significant. RDG
units operate stably during the time horizon.
The described research problem is formulated with two parts: (i) a
multi-objective mathematical model [Eqs. (1)–(16)] is developed to
design SMG with day-ahead dynamic price-based DR program; (ii) a
simulation model [Eqs. (17)–(22)] is built to integrate blockchain
technology into the SMG for implementing dynamic price-based DR
program with real time load control.
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

2.1. SMG design model with day-ahead dynamic price-based DR program Variables

In this section, a multi-objective mathematical model is developed

for SMG design problem. The following indices, parameters, and vari­
xig = 1
0 {
if a potential RDG unit i with technology g is established
otherwise
Lig capacity of RDG unit i with technology g
ables are used to formulate through the paper.
Sets and indices ligt amount of electric power generated ofRDG unit i
with technology g at time t
I set of potential locations for renewable distributed generation, i t
ligj amount of electric power deliveried from RDG unit i
= 1, ...,I
G set of renewable generator types, g = 1, ...,G with technology g to aggegator j at time t
J set of aggregators, j = 1, ...,J l tjd amount of electric power transmitted from aggregator j
D set of residential areas, d = 1, ...,D to demand area d at time t
T set of time periods, t = 1, ...,T pjt day-ahead dynamic electric power pricing provided by

Parameters aggegator jat time t

qdt response load of demand aread based on dynamic pricing
r discount rate [%](uncertainty)
lt lifetime of RDG units [year] Formulation
fcig capital investment of each capacity of RDG unit i Max W1 = dtj pjt
r 1
fcig N
j J d D t T 1 (1 + r ) lt i I g G
with technology g [m.u./mWh](uncertainty)
N number of days in a year [number] ( i I g G t T
fcig N +
1 t
g vcig lig )
t
certain percentage of fixed operation and maintance i I g G j J t T
vmigj ligj j J d D t T
vmjd ltjd (1)
cost [m.u./mWh]
Min W2 = eoLig + ep (1 lc ) ligt
g conversion factor of generation technology g [%] i I g G i I g G (2)
vcig generation variable cost of RDG unit i with technology g l tjd d
Max W3 = - l tjd pjt qdt 1
[m.u./mWh](uncertainty) j J d D t T d D t T d
qdt
vmigj delivery cost from RDG unit i with technology g to aggregator j
(3)
[m.u./mWh/km](uncertainty) The three objective functions are developed to consider simulta­
vmjd delivery cost from aggregator j to demand area d neously economic, environmental, and social aspects of the SMG. The
[m.u./mWh/km](uncertainty) first objective function aims to maximize the total profit, while the
second objective minimizes the created environmental impacts by es­
eo environmental cost of land use for opening RDG unit i
tablishing and operating of SMG. The third objective deals with the
with technology g [m.u./mWh](uncertainty) maximization of social benefits of consumers.
ep CO2 emissions handling cost for operating RDG unit i Eq. (1) presents the profit obtained by the difference between the
revenue and costs. The income of the network comes from the sale of
with technology g [m.u./mWh](uncertainty)
electric power by dynamic pricing, as shown by the first term. The total
lc coefficient of loss in electric power [%](uncertainty)
cost is composed of the installation cost of RDG units and the opera­
Qdt base electric power demand of each demand area [mW/h] tional costs of the SMG expressed in the remaining terms. The in­
(uncertainty) stallation cost of RGU units in the second term is over long period ac­
cording to their lifetime. To evaluate the effects of the RGU units
acceptance coefficient of demand area d
d
installation decision on the short-term social and environmental im­
in demand response [dimensionless] pacts, it is converted into the short-term cost in the planning time by
d coefficient between nominal demand and response demand of dividing the total installation cost based on the number of operating
days per year, according to Refs. [9–10].
demand area d [dimensionless]
Eq. (2) shows the environmental impacts under the costs form incurred
numg maximum number of RDG unit i with generation technology g by establishing and operating of RDG units with large capacities. The first
opened in area [number] term measures the environmental cost related to the land use for opening
minimum utilization rate of RDG unit i the SMG. To open the large capacity RGU units, it requires a large area of
ig
planned land. This affects agricultural and residential land in some project
with generation technology g [%](uncertainty)
areas. In addition, if the forest land is planned to develop renewable energy
LigL minimum capacity of RDG unit i with generation technology g projects, it will lead to affect climate change, soil erosion, and flooding. For
[mWh] example, large-scale microgrid projects in Vietnam has affected 27,600 ha
of agricultural land in 2018 [29]. The impacts on land use are quantified
LigU maximum capacity of RDG unit i with generation technology g
based on the installed capacity at the RGU units in this study as the en­
[mWh] vironmental impacts from the opening of the SMG. The second term cal­
ed price elasticity of demand [dimensionless](uncertainty) culates the CO2 emissions handling cost caused by the operating of the
pfixed fixed electric power price (policy price)[m.u./mWh] SMG. The CO2 emissions is assumed to be dependent on generation tech­
ei maximum limit of environmental budget [m.u] nologies g (e.g., biomass, solar, wind) and on the amount of electric power
generated and delivered to residential areas from the SMG.

4
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

Fig. 1. Underlying structure of proposed SMG.

In the longer term, mandatory DR programs for all consumers could established to ensure that the amount of generated power is equal to the
come at a cost to society [30–31]. Therefore, equation (3) aims to amount of transmitted power to the aggregators, and that it is larger
maximize consumer satisfaction from the social perspective of the than the demand response load qdt of the consumer at every instant t.
adoption of a DR program. The consumers’ electricity bill is expressed Equation (11) estimates the demand response load of each residential
in the first term, while the difference between the nominal user demand area d based on the day-ahead dynamic pricing pj provided by ag­
ddt and the user demand in response to dynamic pricing l tjd is calculated gregator j and nominal user demand Qdt of each residential area. Qdt is an
by the last term. d indicates the acceptance of a residential area of a DR uncertain parameter in the form of triangular fuzzy numbers and j is an
program. effect coefficient for the price of each aggregator j.
S. t Constraints (12) and (13) are coefficients about consumer satisfac­
tion based on the price elasticity of demand ed and current fixed price
x ig 1, i pfixed from the government. Constraint (14) ensures that the environ­
g G (4) mental costs of SMG do not exceed the allowable budget. Constraints
(15) and (16) guarantee the binary nature and nonnegativity of the
xig numg , g
(5) decision variables. In section 2.2, a simulation model for real time-
i I
based DR with visible load control using blockchain technology is
ig Lig ligt Lig , i, g presented.
t T (6)
2.2. Simulation model for real time-based DR with visible load control using
LigL x ig Lig LigU x ig , i, g (7) public blockchain
l tjd x ig Lig , i, g, t (8)
Blockchain technology has been recognized for its potential benefits
ligt = t
ligj , g, t to transactive energy systems or peer-to-peer energy trading markets.
i I j J (9) However, in this study, another potential of block technology related to
provide a real time load control mechanism to all the network players
t
ligj qdt , j, t (10) when implementing the DR program in SMGs. A simulation model with
public blockchain technology is built in this section. The blockchain
qdt = Qdt j pj (11) technology has advantages about the decentralized network, the cryp­
tology method, and the consensus algorithm that support real time load
1
= +1 control in DR programs. Thus DR programs can be implemented in the
d
ed (12)
real time with blockchain technology in an in efficient manner.
pfixed Most DR programs in previous studies were based on day-ahead
=
d
d (13) planning (the electricity price cannot be adjusted during the next day)
and the load control was not displayed to all the network players. This
eoLig + ep (1 lc ) ligt ei has influenced the system balancing role of aggregators. With block­
i I g G i I g G (14) chain technology, DR events take place in real-time and the electricity
price might be adjusted each hour based on demand signals from cus­
x ig {0, 1} (15) tomers. In addition, it ensures a safe and transparent method to share
pjt , ligt , l tjd , Ddt 0 all relevant information (e.g., capacity profile of RDG units, demand of
(16)
home users) to all the network players to support real time load control
Inequality constraints (4) and (5) indicate that the number of RDG from multiple sides of the SMG.
units with a certain technology cannot exceed a certain number based
on local policies. Constraint (6) states that power production amount 2.2.1. Public blockchain technology
cannot exceed the installed capacity, where ig is the minimum utili­ In our study, public blockchain technology is integrated as an ad­
zation rate of the RDG capacity. If an DRG is built, its capacity must lie vanced information technology to enhance the sustainability of mi­
between the specified lower and upper bounds, following constraint crogrids based on a disruptive innovative approach to DR programs. In
(7). Constraint (8) indicates that if an DRG is not opened, the corre­ traditional centralized management approach, DR programs are im­
sponding electric power should be zero. Constraints (9) and (10) were plemented by central authority. This has led to challenges in the

5
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

processing time of DR events and the private information security is­

sues. The private information security problem in DR programs has
been highlighted as the main cause of cyber-attacks in the modern
grids. Public blockchain technology allows anyone in a decentralized
network to verify or add data to the blockchain and energy transac­
tions, which could be authenticated by all nodes. This helps to promote
faster energy transactions in the real-time manner [32]. In the public
blockchain-based SMG, energy data is registered and stored in the local
blockchain using private-key cryptography. That will be replicated and
shared to all nodes in the network for validation using public-key
cryptography. Finally, DR program is conducted through a smart con­
tract with a consensus algorithm. The framework to integrate the
blockchain in an SMG is shown in Fig. 2.
As depicted in Fig. 2, the stakeholders in a SMG are only responsible
for physical flows (electric power), while the information flows
(transactive energy data) are conducted and controlled by four tech­
Fig. 3. Smart contract for DR program in SMG with public blockchain tech­
nologies in the blockchain. (1) Data storage is considered as a dis­
nology.
tributed ledger in the blockchain. This mainly involves the database of
each block under digital assets. (2) The cryptology method is a tool to
achieve security and privacy protection. It helps to avoid loops in the 2.2.2. Smart contract for managing the SMG balance via DR program
network by preventing nodes from forwarding a transaction if it was Using blockchain technology in the DR mechanism in an SMG is
already registered. (3) The smart contract is a code that implements conducted by a smart contract which is shown in Fig. 3. Smart contracts
business rules based on the verification and consent of all objects in the may describe the behavior of RDG units during the DR period to ensure
network. (4) A consensus algorithm is used to let everyone in the net­ the grid-level balance between the energy supply and demand. These
work to agree a certain smart contract and ensure its consistency. To contracts are registered in the blockchain when consumers were noti­
achieve consensus among all network players including consumers, fied about dynamic electricity pricing and triggered by a new energy
RDG units, and aggregator, a proof of stake (PoS) protocol in Ethereum transaction from consumers.
is applied to validate DR events in an SMG. In the PoS, the participant A control signal (DRsignal) is sent to the blockchain system from
(e.g., RDG units) will create the next block based on how much energy consumers as equation (17)
they have generated in the network. It is called “staked”. The stake is
adt pjt
determined by the number of energy coins that an RDG unit receives a DRsignal (t ) = ddt + ,t [tstart , tend]
specific case of DR events from consumers. All network players could bdt (17)
take the role of DR events validator and could be the miner of the next
where ddtis the forecast energy demand of a consumer d, and the
valid block. Each validator should own some stake (energy coins) in the
second term in the equation represents the energy demand adjustment
SMG. In our proposed model, the total reward for an DRG unit during a
based on the dynamic pricing.
DR event could be used as a guarantee of the block’s validity. Based on
DRsignal of consumers is registered in the data storage. The smart con­
the above advantages, PoS has the ability to achieve a high level of
tract checks the monitored energy consumption data of the consumers and
reliability in the DR programs by mining the next valid block based on
the energy profile of RDG units to detect any significant deviations and
the amount of stake and validating associated events in the previous
notifies the aggregators. The deviations are determined as
block based on the amount of guarantee of each DRG unit in the SMG.
The following section details the implementation of a smart contract tend
(t ) = ligt (t ) DRsignal (t )
based on the PoS consensus algorithm on the Ethereum platform. tstart (18)

Fig. 2. Blockchain application to the real-time price-based DR program in SMG.

6
Y.-C. Tsao, et al.
If Δ(t) is close to 0, no significant deviation is registered, and the DR
program is not conducted. A positive value [Δ(t) + ] indicates that the
RDG units have not reduced the generated energy as requested, while a
negative value [Δ(t)-] indicates that the RDG units have decreased the
generated energy.
In case of significant positive or negative deviation (over 10% of
DRsignal), the smart contract calculates the associated penalties for the
RDG units. Otherwise, the RDG units receive added revenue considering
how much the energy has been adapted during the DR time. The rev­
enues and penalties are established by the aggregators considering the
average electricity price in the grid. In our approach, the total penalty
for not sufficiently meeting the demand is calculated based on the
penalty rate and the deviation registered for an RDG as Eq. (19).
RDGpenalty = prate | ± (t )|
To determine how much an RDG has adapted its energy profile to
DRsignal, and using its forecast energy demand as a reference, we used the
adaptability power curve (APC) metric of the European Union as follows:
tend
tstart
|ligt (t ) ddt |
APC = tend
ddt
tstart
where APC ranges from 0 to 1, with 0 meaning that the RDG has not
adapted its energy to DRsignal, and 1 meaning that the energy was
adapted and its deviation from DRsignal is zero. The total reward for a
RDG during a DR program is calculated by the aggregators using the
following formulation:
RDGreward = APC × rrate × ligt (t )
where rrate is the revenue rate for each kW of energy shifted as es­
tablished by the aggregators.
The smart contract implements the rules for balancing the energy of the
entire grid and enforcing its resilience without a need for third-party in­
tervention (aggregators). The smart contract starts a new DR program and
communicates the associated DRsignal, RDGpenalty, and RDGreward to the in­
terested RDG units. The smart contract helps to balance the overall energy
production and consumption at the grid level as Eq. (22):
= l t (t ) DRsignal (t ), t [tstart , tend]
SG
i I g G ig
If the result is close to 0, there is a balance between energy pro­
duction and consumption at the grid level. Otherwise, a positive value
represents a surplus of energy in the grid, while a negative value shows
a deficit of energy in the grid. If such imbalances are determined, the
smart contract begins a new DR program that allows other DRGs to
solve them by increasing or decreasing their generated energy and, as a
result, rebalancing the state.
By running the simulation model (17)–(22) on the Solidity language
for implementing the smart contract, the DR decisions of an SMG with
blockchain technology are determined.
3. Proposed solution approach
Microgrid is a complex system because it consists of a group of RDG
units that are characterized by a high degree of the intermittent nature from
renewable energy sources. In addition, the aspects of environment and so­
ciety are taken into account for enhancing the sustainability of microgrids,
so a high degree of uncertainty in input data have posed considerable
challenges to solution approaches. As a result, deterministic methods are
not able to provide a precise analysis of SMG designing. Although there are
many uncertain programming methods that have been applied in the mi­
crogrid design-related previous studies (e.g., stochastic model, scenario
model, and, simulation model in Table 1), these methods have not suc­
cessfully overcome many different types of uncertainty in the real-life si­
tuations in which there is a lack of knowledge and historical data to model
stochastic models and scenarios for expressing uncertain parameters.
Electrical Power and Energy Systems 125 (2021) 106418
Fuzzy programming provides a framework to deal with a variety of
uncertainties as well as the flexibility in objective functions and the
elasticity in soft constraints without increasing complex of the original
model to lead to computational challenging [33]. However, to guar­
antee reliable solutions to conflicting multi-objective optimization
problems in an uncertain decision environment, a combination of fuzzy
programming and robust optimization should be considered. In this
study, an RFMOOP model is proposed for the SMG design problem in
uncertain environments based on the integration of fuzzy possibilistic
programming in [34] into robust optimization programming [35].
Fuzzy possibilistic programming is one of two main branches of
fuzzy programming including fuzzy flexible programming and fuzzy
possibilistic programming, it applies the strong mathematical concepts
such as expected interval and expected value of fuzzy numbers to cope
(19)
with imprecise input data or epistemic uncertainty as well as flexibility
in goals and constraints [36–38]. Numerous methods have been based
on this approach to cope with supply chain network design problems
under situations of uncertainty [39–40]. To the best of our knowledge,
the first time, fuzzy probabilistic programming is applied to adjust the
objective functions and constraints for effectively handling uncertain
conditions in the field of power systems design.
(20)
Robust optimization could help to achieve optimal solutions based
on risk-averse methods in decision-making process. The mature litera­
ture about robust optimization provides reliable solution methods with
optimality guarantee [41–43]. Pishvaee et al. [44] put a significant step
forward in extending of the robust optimization theory into fuzzy
programming. Some subsequent efforts in this line of robust fuzzy
programming were done by many scholars in the supply chain network
(21) design field [44–46]. The combination between robust optimization
and fuzzy programming can successfully deal with the complex system
features under uncertainties. The proposed RFMOOP approach is su­
perior to the other uncertain programming methods as follows:
(i) Providing reliable solution methods with optimality and feasibility
guarantee for optimization problems under uncertain environment.
(ii) Effectively handling different kind of uncertainty in data, the flex­
ibility in objective functions, and the elasticity in soft constraints
without increasing computationally challenging.
(22)
The process of constructing the RFMOOP model and solving pro­
blem are based on the compact form of the SMG design problem which
can be stated as follows:
MaxW1 = fx + cy (23)
MinW2 = gy (24)
MaxW3 = hy (25)
S. t
Ay d (26)
Bx = 0 (27)
Dy Fx (28)
x {0, 1}, y 0 (29)
where x and y represent the binary and other decision variables,
respectively. A, B, and D are parameter matrices known deterministi­
cally, while f, c, g, h, d, and F are fuzzy variables related to the opening
costs, operating cost, CO2 emissions, social cost, energy demand, and
capacity of the RDG units. Three objective functions W1,2,3 represent
the economic, environmental, and social issues in the SMG.
According to the robust optimization approach, a robust objective
function must include three components, namely (1) the expected value
of the goals, (2) the feasibility robustness, and (3) the optimality ro­
bustness. We apply the mathematical concepts, such as the expected
interval and expected value of fuzzy numbers in [37] to formulate the
7
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

first term of the robust objective function. For a triangular fuzzy uncertain parameters of the soft constraints, and the parameters β and α
number with three prominent points, for example f = (f1, f2, f3) with its denote the minimum satisfaction level of the possibilistic constraints for
membership function in range [0, 1], as described in Eq. (30). Ac­ capacity and demand, respectively. The penalty cost is determined
cording to the defuzzification process in [37], the expected value of f through the supply contracts, and these values can be adjusted based on
can be defined in Eq. (31). It is noted that the same equations can be the agreement between the stakeholders in the network or the gov­
used for all others fuzzy parameters in the model including c, g, h, d, ernment regulations.
and F. When capacity coefficient matrix F is tainted with epistemic un­
certainty, the RFMOOP model results in a nonlinear programming
x f1
gf (x ) = f2 f1
iff1 x f2 model in Eq. (36). To escape from the complexity of a nonlinear model,
a new auxiliary variable u = β.x is added into the model. Accordingly,
0 ifx = f2
µ f (x ) = Eq. (36) is reformulated into Eq. (39) in a linear form. Also, the three
f3 x
h f (x ) = f3 f2
iff2 x f3 added constraints with a sufficiently large number M are shown in Eqs.
(40)–(42), where u is equal to zero if x = 0, and u is equal to β if x = 1.
1 ifx f1 orx f3 (30) The RFMOOP model could be converted to the following linear model.
1 1 Min OF = [ 1 [EV (f ) x + EV (c ) y] + 2 [EV (g ) y ] + 3 [EV (h) y]]+
0 f
g 1 (x ) dx + 0
hf 1 (x ) dx f1 + 2f2 + f3
EV (f ) = = [µ [(f3 x + c3 y ) [EV (f ) x + EV (c ) y] + g3 y EV (g ) y + h3 y EV (h) y]]+
2 4 (31) d2 + d3 d1 + d2 F 1 + F2 F2 + F3
d2 ( ) (1 )( ) + u( ) + (x u )( ) F1 x
2 2 2 2
Considering the optimality robustness, some common methods have
been applied in the literature of the robust optimization, such as the (38)
variance of objective function [35,47] or the upper bound of constraints S. t .
determined by decision-makers [48]. However, these methods have
posed computational efforts for solving the problem in the non-linear F1 + F2 F2 + F3
Dy u + (x u)
programming model. To escape the obstacle, we apply the fuzzy 2 2 (39)
ranking method to control the optimality robustness. This method can
u Mx (40)
support different kinds of fuzzy number, such as triangular and trape­
zoidal. In addition, it does not rise the number of objective functions u M (x 1) + (41)
and inequality constraints. For triangular fuzzy number f = (f1, f2, f3)
with three prominent points (see Fig. 4), the fuzzy ranking method u (42)
controls the optimality robustness of uncertainty parameters by rea­
Constraints (34, 35, and 37) remain unchanged
lizing the difference between the maximum value of f (f3) and its ex­
SMG design is a multi-objective programming problem. The weight
pected value [EV(f)]. Eq. (32) shows the fuzzy ranking method for fuzzy
method is applied to aggregate three objective functions in order to
number f . It is applied similarly to other fuzzy parameters including c,
transform the original model into the RFMOOP model. The steps to
g, and h in the model.
solve the RFMOOP model can be summarized in the form of an algo­
R (f ) = f3 EV (f ) (32) rithm, as follows:

A penalty cost function for each unit of violation of constraints have Algorithm. Step 1: Identify all uncertain parameters and decision variables
been used in robust objective functions to take into account the feasi­ in the original model as Eqs (1) – (16).
bility robustness. In multi-objective optimization problems, the penalty Step 2: Convert the original model into the RFMOOP model by using Eqs
functions will be a balance or compromise between competing goals. (33) – (42).
Accordingly, the robust objective function of the proposed RFMOOP Step 3: Determine the value of the coefficients (i.e., πi, ρ, γ, β, and α)
model is presented as follows. based on the opinion of experts.
Step 4: Solve the RFMOOP model to obtain the optimal solutions for the
Min OF = [ 1 [EV (f ) x + EV (c ) y ] + 2 [EV (g ) y ] + 3 [EV (h) y ]]+
feasibility constraints. If the decision-maker is satisfied with the current
[ [(f3 x + c3 y ) [EV (f ) x + EV (c ) y ] + g3 y EV (g ) y + h3 y EV (h) y]]+
d2 + d3 d1 + d2 F 1 + F2 F2 + F3
solution, stop. Otherwise, provide another compromise solution by changing
d2 (
2
) (1 )(
2
) + x (
2
) + (1 )(
2
) F1 the values of the coefficients, and proceed to step 3.
(33)

d2 + d3 d1 + d2 4. Case study
S. t . Ay (1 )
2 2 (34)
4.1. Overview of case study
Bx = 0 (35)
To deliver sufficient electric power for a residential area that in­
F1 + F2 F2 + F3 cludes 300 smart houses in Vietnam’s Ho Chi Minh City, the Electricity
Dy + (1 ) x
2 2 (36) of Vietnam Company has been considering the establishment of RDG
x {0, 1}, y 0 (37)

The first term OF measures the average performance of the con­

cerned system based on the expected value of the fuzzy numbers. We
use a weighted sum aggregation function to combine three objective
functions (W1, W2, and W3), where πi (i = 1, 2, 3) is the weight of the ith
3
objective function ( i i = 1). The second term adjusts the degree of
optimality robustness by minimizing the maximum deviation of the
uncertain objective function. The parameter indicates the importance
of this term against the other terms. The last term controls the feasi­
bility robustness using an infeasibility penalty cost function. The
parameters ρ and γ represent the penalty cost of each violation unit in Fig. 4. Triangular fuzzy number f.

8
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

Table 2
Capacity and RDG units-related costs.
Potential locations Capacity (mW) under trapezoidal fuzzy number form Capital investment ($/mW) Generation variable cost ($/mW)

B S W

CC (250, 300, 400) (0.85, 1.14, 1.75) (0.25, 0.75, 1.00)

NB (550, 650, 700) (1.05, 1.75, 1.95) (0.50, 1.15, 1.50)
BC (350, 500, 650) (0.95, 1.25, 1.75) (0.25, 0.50, 1.20)
BT (200, 250, 350) (1.25, 1.75, 2.05) (0.85, 1.25, 1.45)
GV (450, 550, 700) (0.95, 1.25, 1.65) (0.25, 0.75, 1.00)
CG (450, 500, 600) (1.35, 1.75, 1.95) (0.75, 0.95, 1.25)
TD (200, 400, 550) (0.75, 0.95, 1.35) (0.35, 0.65, 1.05)
Q2 (300, 400, 550) (1.35, 1.75, 1.95) (0.75, 0.95, 1.25)
Q7 (500, 700, 800) (1.05, 1.75, 2.25) (0.55, 0.95, 1.55)
Q9 (350, 500, 650) (1.14, 1.75, 2.05) (0.95, 1.25, 1.75)

Table 3 aggregators that buy electric power from the RDG units at a fixed price
Environmental impacts-related data. to sell to consumers using the proposed dynamic pricing schemes.
Environmental impacts Generation
DR programs are conducted in each smart building. In a traditional
SMG, consumers will receive day-ahead competitive dynamic pricing
B S W information (24 h) from aggregators at midnight in order to conduct the
demand response the next morning. In an SMG with blockchain tech­
Land issues (6.895, 8.578, (5.245, 7.895, (7.895, 9.875,
($/kWh) 9.784) 8.475) 10.48)
nology, receiving information and conducting the DR of consumers
Noise (2.145, 3.645, (1.345, 2.456, (3.145, 5.645, occur in near real time of using smart contract.
($/kWh) 4.567) 3.978) 7.145)
CO2 emissions (0.025, 0.175, (0.034, 0.044, (0.003, 0.004,
($/kWh) 0.235) 0.054) 0.005)
4.2. Data setting

In this case study, the forecast electric power demand of residential

units with a five-year lifetime for a local market in the Thu Duc pro­
areas is obtained by referring to [49]. The capital investment and ca­
vince. Three renewable energy technologies [solar power (S), wind
pacity of each RDG unit is referred by [50] while some costs for op­
power (W), and biomass power (B)] are installed based on real condi­
erations and maintenance along with the CO2 emission level of power
tions of 10 considered potential locations including Cu Chi (CC), Nha Be
generation and distribution are referred to [51]. However, in an un­
(NB), Binh Chanh (BC), Binh Tan (BT), Go Vap (GV), Can Gio (CG), Thu
certain environment, they had to be modified into triangular fuzzy
Duc (TD), Quan 2 (Q2), Quan 7 (Q7), and Quan 9 (Q9). There are three
numbers based on expert opinion. Table 2 shows the information of the

Fig. 5. Network structure and optimal amount energy of SMG.

9
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

Fig. 6. Dynamic pricing and market share of each aggregator.

capacity, capital investment, and generation variable cost of these RDG
units.
For data related to environmental impact, we assumed that estab­
lishment of RDG units would use a certain land area. This would effect
on the habitats, such as soil erosion and noise. In addition, the CO2
emissions caused by electric power transmission and distribution would
depend on the length of lines. Table 3 shows the environmental im­
pacts-related data used in the case study, which are identified according
to [51].
For the social term in our model, we set the price elasticity of

10
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

Fig. 7. Code excerpt of smart contract for balance the energy supply and demand through DR events.

Table 4
Objective functions value in SMG and SMG with blockchain technology.
Weight of objectives SMG Simulation SMG with blockchain technology

6 6 6 2
W1 (10 ) W2 (10 ) W3 (10 ) OF (10 ) W1 (106) W2 (106) W3 (106) OF (102)

π1 = 0.5 2.04 5.24 0.27 9.07 2.32 4.87 1.43 24.4

π2 = 0.3
π3 = 0.2
π1 = 0.2 0.41 4.46 0.57 8.93 2.16 3.77 1.63 25.5
π2 = 0.5
π3 = 0.3
π1 = 0.3 1.86 3.91 1.11 5.63 2.76 4.57 2.04 21.9
π2 = 0.2
π3 = 0.5
Total 4.31 13.6 1.96 23.6 7.24 13.2 5.10 71.8

demand (ed ) to be −0.41. It means that when the electricity price in­
creases by 1%, the demand for electricity will decrease by 0.41%.
Coefficients for the price of each aggregator ( j ) and competition degree
among them ( j' ) are set to be 0.2 and 0.5, respectively. In Vietnam, a
number of renewable energy projects have been launched in recently at
a suggested fixed price ( pfixed ) to be $ 0.12/kWh, while the highest price
( pmax ) would be $1.65/ kWh in the deregulated market. Two parameters
adt andbdt are set randomly in range [0; 1] for considered residential
areas. These parameters are evaluated by the experts at Electricity of
Vietnam Company.
The penalty cost for each unit of violation of constraints regarding
power demand and the capacity of RDG units are γ = ρ = 3.14. All
remaining parameters, such as the weight of the objective functions (πi)
and the satisfaction levels of flexible constraints (α and β) are used to

11
Y.-C. Tsao, et al. Electrical Power and Energy Systems 125 (2021) 106418

is the same with low prices from 1:00 to 9:00, increasing from 10:00 to
17:00, and finally decreasing from 18:00 to 24:00. The highest prices
are $1.40, $1.62, and $1.62 for aggregators 1, 2, and 3, respectively,
occurring at 19:00, whereas the electricity consumption is largest from
10:00 to 17:00. It indents to reduce the peak load while minimizing the
CO2 emissions but still maintaining consumer satisfaction by a devia­
tion in electricity price and demand response.

4.3.2. Real-time price-based DR simulation results in SMG with blockchain

technology
In this section, we compare the balance of the energy supply and
demand in an DR program of an SMG and SMG with blockchain tech­
nology by using simulation approach. To validate the DR through smart
contracts in blockchain technology, a simulation based prototype is
Fig. 8. Actual versus forecast energy demand using DR based on blockchain. implemented on the Ethereum platform [52] using Solidity language.
The coin used to pay for DR events is the Ethereum coin. Also, each
consumer and RGU unit have their own coins which are obtained
through implementing DR events in the network. Fig. 7 shows a code
excerpt from the smart contract for balance the energy supply and
demand through DR events in a SMG with blockchain. To control the
balance between supply and demand in an SMG, a new smart contract
can be performed by the function of deplopSMGContract (see line 11 in
Fig. 7). All energy profit of RGU units must keep in the SMG’s registry
(line 15 in Fig. 7). Whenever an DR signal is registered by consumers,
the energy profile of a RGU unit and the balance state of SMG are
updated accordingly (see lines 7 and 8 in Fig. 7) and a new DR event is
generated in line 22 in Fig. 7. SMG responses to a new DR event from a
consumer based on the energy balancing process. If the energy profile
of RGU unit is inadequate to balance the grid, the DR event is auto­
Fig. 9. Energy deviations Δ (t) in smart contracts. matically reverted and RGU unit will be charged a penalty cost as Eq.
(20) (line 32 in Fig. 7). If the energy profile of a RGU unit is appropriate
to adapt a new DR event, the DR event is registered in the system and
Table 5
Comparing the value of the three objective functions and the computational
the smart contract will be performed in line 37 in Fig. 7.
time under two approaches. The simulation results in 300 mins are summarized in Table 4 and
displayed in Figs. 8 and 9. As seen in Fig. 8, the grid electric power
Weight of The proposed RFMOOP model The RSOP model
achieves a balance between the forecast demand and response demand
objectives
W1 W2 W3 CPU W1 W2 W3 CPU of consumers in near real time based on the blockchain by a deviation
(106) (106) (106) time (s) (106) (106) (106) time (s) of 8.7%, while this deviation is 13.9% with the DR program based on
day-ahead dynamic pricing in the SMG.
π1 = 0.5 2.32 4.87 1.43 105 2.29 4.89 1.38 228
The energy deviations from using blockchain technology with smart
π2 = 0.3
π3 = 0.2 contracts are shown in Fig. 9. There are five times that RDG units in­
π1 = 0.2 2.16 3.77 1.63 103 2.09 3.79 1.59 219 creased the energy generated by a matching demand (e.g., 4:00, 10:00,
π2 = 0.5 13:00, 16:00, and 18:00). The smart contract effectively manages the
π3 = 0.3 grid energy balance by using DR events in a timely manner. Moreover,
π1 = 0.3 2.76 4.57 2.04 112 2.69 4.61 2.05 217
π2 = 0.2
it is able to efficiently correct the deviations [Δ(t)] by assigning new DR
π3 = 0.5 signals to interested RDG units by considering their energy profiles.
Average 2.41 4.40 1.70 106.67 2.36 4.43 1.67 221.33 Table 4 details the values of the economic, environmental, and so­
cial objectives for two approaches: (1) SMG with day-ahead dynamic
pricing-based DR and (2) SMG with real-time price-based DR with
perform a robustness analysis of the proposed model. The RFMOOP blockchain technology. By comparing the values of the objective
model of the case study is solved by using MATLAB 2013 optimization functions, it can be claimed that SMG with DR based on blockchain
software running on a dual-core 3.40-GHz computer with 8.0 GB of technology achieves a profit and consumer satisfaction increases by
RAM, while real-time price-based DR with the public blockchain tech­ 1.68% and 2.61%, respectively, while it reduces the environmental
nology in SMG is simulated using MATLAB. The significant results are impacts to 0.97%. The proposed model accurately reflects the nature of
presented and discussed below. the DR in the SMGs toward maximizing consumer satisfaction. DR helps
to reduce the environmental impacts that were acknowledged in many
4.3. Result analysis and discussion previous studies. However, the blockchain technology application in
DR programs for environmental impact reduction and economic and
4.3.1. Network design decisions social benefit increase is investigated for the first time in our study. The
The results of solving the model including day-ahead electricity expected performance of the SMG with blockchain technology (OF
pricing, network structure, and optimal amount energy of SMG are value) is higher than that for the SMG by 3.04 times. This reveals that
visualized in Figs. 5 and 6. As can be seen in Fig. 5, the capacity of the blockchain technology with a smart contract helps to enhance the
solar power installed accounts for 42%, biomass power constitutes sustainability of a microgrid by adjusting the grid balance state quickly.
31%, and wind power represents 27%, corresponding to the seven RDG We also present sensitivity analyses for the weights of the objectives
units opened. As depicted in Fig. 6, each aggregator has a different day- to show the behavior of the proposed model. As shown in Table 4, the
ahead pricing scheme to promote DR programs, but the overall pattern social objective is sensitive to weight changes, while the environmental

12
Y.-C. Tsao, et al.
objective does not have a clear trend. It is obvious that there is a conflict
between the economic and social objectives in the proposed model.
4.3.3. Performance assessment of the proposed RFMOOP model
In this section, the performance of the proposed RFMOOP model is
evaluated by comparing the results about the value of the objective
functions and the computational time with the robust stochastic opti­
mization programing (RSOP) model. Stochastic programming has been
widely applied in previous studies to overcome the uncertainties in the
power system planning models (see Table 1). However, this method has
revealed some disadvantages in terms of the computational time due to
a large number of scenarios considered to express the uncertainties. In
addition, the requirement for historical data to analyze the probability
distributions of the uncertainty parameters has made the method less
effective in some real-life situations where the relevant historical data is
insufficient.
The steps to develop the RSOP can be found in [53]. The weighted
sum aggregation method is also applied to combine three objective
functions (W1, W2, and W3) in the RSOP model. The comparative results
are shown in Table 5. It is clear that the RFMOOP model has a better
performance in improving the sustainable goals (economy, environ­
ment, and society) of microgrid and the computational time under
uncertain conditions.
5. Conclusion
This study solves the SMG design problem with consideration of
blockchain technology to support real-time price-based DR programs
uncertain environments. An RFMOOP model was introduced to si­
multaneously determine network design decisions, equilibrium supply,
and day-ahead pricing decisions while a simulation model was devel­
oped to explore the benefits of blockchain technology on the sustain­
ability of microgrids through real-time price-based demand response.
The proposed model was applied to a real case study in Vietnam’s
Ho Chi Minh City. The results indicate that SMG with blockchain
technology increases profitability and consumer satisfaction by 1.68%
and 2.61%, respectively, while it reduces the environmental impacts to
0.97%. In addition, the expected performance of the SMG with block­
chain technology is 3.04 times higher than in SMG without blockchain
technology. It reveals that the blockchain technology helps to enhance
the sustainability of microgrids with an ability to quickly balance the
supply and demand in the grid. Besides, the proposed model has a
better performance in improving the sustainable goals (economy, en­
vironment, and society) of microgrid and the computational time under
uncertain conditions compared to the robust stochastic programming
approach.
Future research can address the planning of SMGs in the local
marketplace, where peer-to peer trades are conducted with the support
of blockchain technology. In addition, smart contract design in block­
chain should be considered in SMGs to ensure a fair deal for various
stakeholders in the network. Finally, the application-related work of
algorithms, such as metaheuristic approaches in uncertain environ­
ments, should be conducted in the power field.
CRediT authorship contribution statement
Yu-Chung Tsao: Conceptualization, Methodology, Supervision. Vo-
Van Thanh: Methodology, Software. Qiuwei Wu: Supervision.
Declaration of Competing Interest
None.
13
Electrical Power and Energy Systems 125 (2021) 106418
Acknowledgement
This paper is supported in part by the Ministry of Science and
Technology in Taiwan under grant 108-2636-E-011 -004.
References
[1] Shakouri GH, Kazemi A. Multi-objective cost-load optimization for demand side
management of a residential area in smart gird. Sustain Cities Soc 2017;32:171–80.
[2] Farhangi H. The path of the smart gird. IEEE Power Energy Magaz 2010;8(1):18–28.
[3] Eising JW, Onna TV, Alkemade T. Towards smart grids: identifying the risks that
arise from the integration of energy and transport supply chains. Appl Energy
2014;123:448–55.
[4] Hall J, Matos S, Silvesre B. Understanding why firms should invest in sustainable
supply chain: a complexity approach. Int J Prod Res 2012;50(5):1332–48.
[5] Shariatzadeh F, Mandal P, Srivastave AK. Demand response for sustainable energy
systems: A review, application and implementation strategy. Renew Sustain Energy
Rev 2015;45:343–50.
[6] U.S Dept. of Energy. Benefits of Demand Response in Electricity Markets and
Recommendations for Achieving Them; 2016.
[7] Khajavi P, Monsef H. Load profile reformation through demand response programs
using smart grid. In: IEEE conference on modern electric power systems (MEPS’10);
2010. p. 20–2.
[8] Newsham GR, Bowker BG. The effect of utility time-varying pricing and load control
strategies on residential summer peak electricity use: a review. Energy Policy
2010;38:3289–96.
[9] Jabbarzadeh A, Fahiminia B, Rastegar S. Green and resilient design of electricity
supply chain networks: a multi-objective robust optimization approach. IEEE Trans
Eng Manage 2017;99:1–21.
[10] Abdullah MA, Muttaqi KM, Agalgaonkar AP. Sustainable energy system design with
distributed renewable resources considering economic, environmental and un­
certainty aspects. Renew Energy 2015;78:165–72.
[11] Yamangil E, Bent R, Backhaus S. Designing resilient electricity distribution grids,
arXiv preprint arXiv:1409. 4477; 2014.
[12] Ma S, Su L, Wang Z, Qiu F, Guo G. Resilience enhancement of distribution grids
against extreme weather events. IEEE Trans Power Syst 2018;33(5):4842–53.
[13] Tsao YC, Thanh VV, Lu JC. Multiobjective robust fuzzy stochastic approach for
sustainable smart grid design. Energy 2019;176:929–39.
[14] Wang Y, Huang Y, Wang Y, Zeng M, Li F, Wang Y, et al. Energy management of
smart micro-grid with response loads and distributed generation considering de­
mand response. J Clean Prod 2018;197:1069–83.
[15] Nikmehr N, Ravadanegh SN, Khodaei A. Probabilistic optimal scheduling of net­
worked microgrids considering time-based demand response programs under un­
certainty. Appl Energy 2017;198:267–76.
[16] Varasteh F, Nazar MS, Heidari A, Shafie-khah M, Catalao JPS. Distributed energy
resource and network expansion planning of a CCHP based active microgrid con­
sidering demand response programs. Energy 2019;172:79–105.
[17] Hassan MAS, Chen M, Lin H, Ahmed MH, Khan MZ, Chughtai GR. Optimization
modeling for dynamic price based demand response in microgrids. J Clean Prod
2019;222:231–41.
[18] Bahmani R, Karimi H, Jadid S. Stochastic electricity market model in network
microgrids considering demand response programs and renewable energy sources.
Electr Power Energy Syst 2020;117:205606.
[19] Khalili T, Nojavan S, Zare K. Optimal performance of microgrid in the presence of
demand response exchange: A stochastic multi-objective model. Comput Electr Eng
2019;74:429–50.
[20] Haddain H, Noroozian R. Multi-microgrid-based operation of active distribution
networks considering demand response programs. IEEE Trans Sustain Energy
2019;10(4):1084–812.
[21] Aitzhan NZ, Svetinovic D. Security and privacy in decentralized energy trading
through multi-signatures, blockchain and anonymous messaging streams. IEEE
Trans Depend Secure Comput 2016;15(5):840–52.
[22] Zhe W, Zhang Z. Research on distribution network data fusion considering re­
newable energy. In: Processing of the 2nd International Conference on Power and
Renewable Energy (ICPRE), China; 2017. p. 20–23.
[23] Pop C, Cioara T, Antal M, Anghel I, Salomie I, Bertoncini M. Blockchain based
decentralized management of demand response programs in smart energy grids;
2018. http//doi: 10.3390/s18010162.
[24] Noor S, Yang W, van Dam KH, Wang X. Energy demand side management within
micro-grid networks enhanced by blockchain. Appl Energy 2018;228:1385–98.
[25] Wang X, Yang W, Noor S, Chen C, Guo M, van Dam KH. Blockchain-based smart
contract for energy demand management. Energy Proc 2019;158:2719–24.
[26] Yang XD, Zhang YB, Lu JJ, Zhao B, Huang FT, Qi J, et al. Blockchain-based auto­
mated demand response method for energy storage system in an energy local net­
work. Proc CESS 2017;37:3703–16.
[27] Bin L, Chao L. Research in implementation of automatic demand response system
based on blockchain theory. Proc CESS 2017;37:3691–702.
[28] Albadi MH, Saadany EFEI. A summary of demand response in electricity markets.
Electr Power Syst Res 2008;78:1989–96.
[29] Y, LT. Combining agricultural production with solar energy exploitation in
Vietnam. Online (assessed 19 Nov 2019. < http://icon.com.vn/vn-s83-156948-
618/Ket-hop-san-xuat-nong-nghiep-voi-khai-thac-nang-luong-mat-troi-o-Viet-Nam.
aspx > .
[30] Darby ST, McKenna E. Social implications of residential demand response in cool
Y.-C. Tsao, et al.
temperate climates. Energy Policy 2012;49:759–69.
[31] da Silva HB, Santiago LP. On the trade-off between real-time pricing and the social
acceptability costs of demand response. Renew Sustain Energy Rev
2018;81:1513–21.
[32] Sharma TK. Public vs Private blockchain: A comprehensive comparison. Online
(assesses 7 Aug 2019). < https://www.blockchain-council.org/blockchain/public-
vs-private-blockchain-a-comprehensive-comparison/ > .
[33] Dubois D, Fargier H, Fortemps P. Fuzzy scheduling: modelling flexible constraints
vs. coping with incomplete knowledge. Eur J Oper Res 2003;147:231–52.
[34] Zadel LA. Fuzzy sets as a basis for a theory of possibility. Fuzzy Sets Syst
1978;1:3–28.
[35] Mulvey MJ, Vanderbei RJ, Zenios S. Robust optimization of large-scale systems.
Oper Res 1995;43(2):264–81.
[36] Inuiguchi M, Ramik J. Possibilistic linear programming: a brief review of fuzzy
mathematical programming and a comparison with stochastic programming in
portfolio selection problem. Fuzzy Sets Syst 2000;111:3–28.
[37] Mula J, Poler R, Garcia JP. MPR with flexible constraints: a fuzzy mathematical
programming approach. Fuzzy Sets Syst 2006;157:7–97.
[38] Mousavi SM, Vahdani B, Moghaddam RT, Hashemi H. Location of cross-docking
centers and vehicle routing scheduling under uncertainty: A fuzzy possibilistic–­
stochastic programming model. Appl Math Model 2014;38:2249–64.
[39] Subulan K, Tasan AS, Baykasoğlu A. Designing an environmentally conscious tire
closed-loop supply chain network with multiple recovery options using interactive
fuzzy goal programming. Appl Math Model 2015; 39: 2661-02.
[40] Tsao YC, Thanh VV, Lu JC, Yu V. Designing sustainable supply chain network under
uncertain environments: Fuzzy multi-objective programming. J Clean Prod
2018;174:1550–65.
[41] Beyer Hans-G, Sendhoff B. Robust optimization – A comprehensive survey. Comput
Methods Appl Mech Eng 2007;196:3190–218.
[42] Gabrel V, Murat C, Thiele A. Recent advances in robust optimization: An overview.
14
Electrical Power and Energy Systems 125 (2021) 106418
Eur J Oper Res 2014;235(3):471–83.
[43] Shang K, Feng Z, Ke L, Chan FTS. Comprehensive pareto efficiency in robust
counterpart optimization. Comput Chem Eng 2016;94:75–91.
[44] Pishvaee MS, Razmi J, Torabi SA. Robust possibilistic programming for socially
responsible supply chain network design: a new approach. Fuzzy Sets Syst
2012;206:1–20.
[45] Ehsan D, Mohsen SN, Maghsoud A, Armin J. Hybrid robust, stochastic and possi­
bilistic programming for closed-loop supply chain network design. Comput Ind Eng
2018;123:220–31.
[46] Bairamzadeh S, Pishvaee MS, Mehrabad MS. A multi-objective robust possibilistic
programming approach to sustainable bioethanol supply chain design under mul­
tiple uncertainties. Ind Eng Chem Res 2015;55(1):237–56.
[47] Govindan K, Fattahi M. Investigating risk and robustness measures for supply chain
network design under demand uncertainty: a case study of glass supply chain. Int J
Prod Econ 2015;183:680–99.
[48] Vercher E, Bermuder JD, Segura JV. Fuzzy portfolio optimization under downside
risk measure. Fuzzy Sets Syst 2007;158(7):769–82.
[49] Vu NHM, Khanh NTP, Cuong VV, Binh PTT. Forecast on Vietnam Electricity
Consumption to 2030. In: International Conference on Electrical Engineering and
Informatics (ICELTICs 2017); 2017. p. 72-7.
[50] Risto T, Aija K. Comparison of electricity generation costs. Research Report.
Lappeenranta University of technology; 2008.
[51] Health and environmental impacts of electricity generation systems: procedures for
comparative assessment. International Atomic Energy Agency, Vienna, Technical
Reports Series 1999; 394 1999.
[52] Wood, G. ETHEREUM. A secure decentralized generalized transaction ledger.
Online (assessed July 18, 2017). http://gavwood.com/paper.pdf.
[53] Liu C, Lee C, Chen H, Mehrotra S. Stochastic robust mathematical programming
model for power system optimization. IEEE Trans Power Syst 2016;31(1):821–2.