Professional Documents
Culture Documents
1 s2.0 S2210670722004322 Main
1 s2.0 S2210670722004322 Main
A R T I C L E I N F O A B S T R A C T
Keywords: This manuscript proposes an intelligent supply and demand management system in a complete network of
Smart building electricity production and consumption. A micro smart grid (MSG), which includes a solar cell, a wind turbine, a
Micro smart grid diesel generator, and battery storage system capable of trading energy with the smart gride (SG), connected to
Multi-objective optimization, Fuzzy inference
smart buildings with different types of loads is modelled. Different types of intelligent fuzzy controllers for
system
distributed management were proposed and optimized via the non-dominated sorting genetic algorithm-II
(NSGAII), which is a multi-objective optimization method. Maximum user comfort, the amount of renewable
energy employment, minimum total power consumption cost, total energy consumption at peak time, and MSG
loss of power supply probability are the five objective functions of the optimization process. Various un
certainties of the real world have also been considered. The most crucial distinguishing feature of this proposed
method is the design of controllers to manage the demand and supply of electricity without the need for daily
optimization. Comparison experiments with other methods presented in the field of electricity supply and de
mand management are conducted to show the superiority of this method in terms of optimality of the results, low
processing volume required to implement real controllers, and its resilience to changing conditions.
* Corresponding author.
E-mail address: aliyehkazemi@ut.ac.ir (A. Kazemi).
https://doi.org/10.1016/j.scs.2022.104119
Received 19 February 2022; Received in revised form 11 July 2022; Accepted 7 August 2022
Available online 13 August 2022
2210-6707/© 2022 Elsevier Ltd. All rights reserved.
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
utilized, considering geographical territories, applications, and power deterministic wind power forecasting, a model was developed to fore
requirements. In the meantime, consumers are also ineffective and have cast a probabilistic interval for the wind power. Akbari et al. (2020)
bi-directional communication with the MSG network. They manage proposed a hybrid robust-stochastic optimization model for smart home
energy consumption according to the network status and pre-defined energy management. They concentrated on the uncertainties of energy
legislations, using intelligent measuring and control equipment prices and photovoltaic (PV) generation in a smart home. Nizami et al.
(Hafez and Bhattacharya, 2012). Since the MSG networks perform better (2020) used a framework for bi-directional electricity trading of demand
consumption management and employ renewable energy resources, response and transactive energy in a residential system. They used a
they have reduced the cost of electricity generation while providing stochastic bi-level minimization model regarding price uncertainty and
better environmental adaptation. On the other hand, as such networks proposed a scheduling and bidding strategy.
have more consistent and precise interaction with the consumers, they Yu et al. (2020) considered market price uncertainty in the energy
are more adapted to their requirements (Osmani and Zhang, 2014). management of a wind-PV-storage grid. They provided a robust opti
With the emersion of MSGs providing localized interactions between mization technique for a large electricity consumer. Zhang et al. (2020)
consumers and the networks, the consumers have become significant assumed the demand for energy as an uncertain parameter. They pre
contributors to the networks, and thus, smart buildings were manifested. dicted the residential energy demand considering the behavior of energy
A smart building can control its resources and systems according to pre- users and their interaction with buildings and available technologies.
defined schedules and the exterior status of the building to achieve Agent-based modeling was adopted to generate uncertain demand sce
maximum energy optimization (Stimmel, 2015). narios. The uncertain demand scenarios were used to input an estab
The cases addressed in this article are several smart facilities used as lished stochastic mixed-integer linear programming (MILP) model.
residential buildings, hotels, and commercial establishments, with a Moghaddas-Tafreshi et al. (2019) proposed a day-ahead forecasting
central smart management system that manages all electrical and me model to estimate a micro-grid network’s electrical and thermal loads. A
chanical facilities. This smart control system is responsible for managing day-ahead forecast was also used to determine electricity generation
different energy resources according to the requirements defined by the from wind turbines. Due to the uncertainty associated with forecasts, a
operator. It is assumed that these facilities are in close vicinity and in Monte Carlo simulation was used to estimate thermal loads, electrical
tegrated into one MSG. In this MSG, various types of generators are loads, and wind power generation.
integrated. Moreover, the MSG network is interconnected with the SG
network and can exchange bi-directional electricity. In addition, we are 2.2. System Components
confronted with uncertainty on both sides of the network. The proposed
method in this study is based on employing controllers capable of allo Some articles only focused on smart buildings and suggested solu
cating power and choosing the appropriate resource according to tions such as user satisfaction, demand management, and cost reduction.
different criteria. The contribution of the research is threefold: (Oprea et al., 2019). Other articles have chosen an independent MSG as
First, we define a complete system including different types of smart the problem. In most of these articles, different electricity generators are
buildings, which are connected to an MSG, and bidirectional trade op located in a small geographic area to supply electricity to a place far
portunity with an SG is possible. from the infrastructure of the electricity supply network. These include
Second, we consider uncertainty on both the consumer and pro PV generators, wind generators, fuel cells, diesel generators, gas
duction sides, which better simulates the reality. microturbines, battery storage systems, and hybrid systems like com
Third, unlike the studies that directly utilize the power allocated to bined heat and power plants (CHP) (Sedighzadeha et al., 2019).
the loads, we develop fuzzy controllers for managing different parts of Some researchers considered an MSG system connected to an SG
this system. system. The MSG can buy or sell electricity from SG (Motalleb et al.,
In the following, the related literature and the theoretical back 2018). Smart buildings connected to an MSG were a problem for some
ground of power management in smart buildings and SG optimization research. The smart building was assumed to be capable of controlling or
are reviewed in Section 2. Then, mathematical models for smart build modifying the energy consumption of an MSG, including various gen
ings and MSG are defined in detail (Section 3). Next, the objective erators such as solar, wind, and fossil (Harkouss et al., 2017).
functions of the optimization problem are described in Section 4. The
uncertainties considered in this study, including the ones on the con 2.3. Problem-solving Methodology
sumer and the generator side, are explained in Section 5. In Section 6,
experimental results are presented, and the proposed method is Many articles have been published to address issues such as demand
compared with the power-based strategy in two different modeling management, user welfare, cost reduction, environmental pollutants
scenarios with one and three smart buildings, respectively. Eventually, a reduction, reliability enhancement, and other noticeable problems in
summary and conclusion of the study are represented in Section 7. smart buildings, MSG, and SG. We categorized these studies into
different groups.
2. Theoretical Background
2.3.1. Mixed-integer Linear Programming
We categorized the previous papers from different aspects: uncer Most of the studies in power management in smart buildings have
tainty consideration, system components included in the modeling, and employed MILP optimization algorithms. For example, Gomes and
problem-solving methodology. Melicio (2021) designed a support management system for the man
agement and operation planning of a microgrid by the new electricity
2.1. Uncertainty Consideration market agent, the microgrid aggregator. The aggregator performed the
management of microturbines, wind and PV systems, energy storage,
Since there are uncertainties in measurements such as weather, user electric vehicles, and energy usage, aiming to have the best market
demand, and electricity prices in the real world, some researchers have participation. This model had a formulation based on a stochastic MILP
focused on this issue. Sun and Wang (2022) presented a multi-objective problem that depended on knowledge of the stochastic processes that
optimization dispatching method to optimize the output power of describe the uncertain parameters. A set of plausible scenarios computed
distributed generators of a micro-grid considering uncertainty in wind by Kernel Density Estimation sets the characterization of the random
power forecasting, aiming to minimize the operational cost and variables. However, as commonly happen, a scenario reduction was
pollutant emission. An adaptive confidence interval was proposed to necessary to avoid the need to have significant computational re
deal with uncertainty in wind power forecasting. In addition, based on quirements due to the high degree of uncertainty. The scenario
2
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
reduction carried out was a two-tier procedure, following a K-means objective functions. Peng et al. (2022), proposed a micro multi-strategy
clustering technique and a fast backward scenario reduction method. multi-objective artificial bee colonyartificial bee colony (ABC) algo
De Oliveira et al. (2011) proposed various mathematical models to rithm to solve microgrid energy optimization problem. Based on the
perform energy management in a residential building and considered consumer psychology principle and game theory, Lu and Zhang (2021)
comfort and the cost of energy as the two functions to be optimized using constructed a non-cooperative Stackelberg model to study the demand
MILP. Parisio et al. (2011) optimized a smart home connected to an MSG response characteristics of multi-type users. The model quantified the
to provide the required energy demand, consider the operational con impact of grid load fluctuation on the benefits of the power company
straints, and reduce the total energy expenses. Kriett et al. (2012) pro and users’ satisfaction with electricity consumption.
posed an MILP-based model to minimize the expenditures in a Among the earlier research, Harkouss et al. (2017) solved a
residential micro-network connected to an SG network. Huber et al. multi-objective optimization problem to design a net-zero energy
(2013) used an MILP algorithm to reduce the cost of electricity in a building in a cold-weather region in Lebanon. To achieve optimization,
residential complex equipped with small CHP units and PV cells. they applied the non-dominated sorting genetic algorithm-II (NSGA-II)
Zhang et al. (2013) emphasized that most energy-consuming tasks optimization method. Shakouri and Kazemi (2017) proposed an energy
are not time-sensitive; instead, they can be carried out consecutively in management framework for minimizing electrical peak load and elec
different buildings with careful planning. Zhang et al. (2014) proposed a tricity cost simultaneously. They defined daily energy requirements and
mathematical formulation to optimize the energy distribution between consumer preferences as constraints and used the goal programing
smart homes with MSG, which was optimized using lexicographic technique to solve the multi-objective MILP problem.
minimax and MILP. Zakariazadeh et al. (2014) proposed a random en Safaei et al. (2015) proposed a life-cycle model for a smart building
ergy planning method for MSGs that considered all scenarios for con connected to an MSG. The model included various energy resources
sumer demand response. In their proposed approach, various types of consisting of PV solar panels, thermal solar cells, conventional boilers,
consumers, such as residential, commercial, and industrial were and electric generators. They employed a multi-objective optimization
modeled. method. The objective was to minimize the sum of non-renewable en
ergy used in the building, greenhouse gas (GHG) emissions, acidifica
2.3.2. Dynamic Programming tion, eutrophication, and the required energy costs together. This
Some studies employed dynamic programming (DP) to optimize multi-objective linear programming (MOLP) problem was solved while
energy consumption in smart homes and power utilization in SGs (Reka satisfying the power demand of the building. Ma et al. (2016) proposed a
and Ramesh, 2016). For instance, Nguyen et al. (2009) used a DP to multi-objective model for an SG connected to several smart buildings.,
optimize an MSG model with renewable energy resources. Their model They employed a method based on non-cooperative game theory to
included a CHP, a PV array, an air turbine, a fuel cell, and a heat load optimize the proposed multi-objective model. In this study, the existence
(boiler). Mitra et al. (2005) optimized a microgrid model concerning the of the Nash equilibrium point in the presented game model was proven.
expenditures and reliability constraints using a DP-based methodology. An iterative solution method was applied to find the equilibrium point.
Nguyen and Crow (2012) employed a dynamic optimization method Ferrari et al. (2017) proposed a modeling method for prosumers (i.e.,
based on DP and the simplex method to perform power planning in an the consumers who are also capable of producing their energy demand).
MSG. Park et al. (2019) proposed optimal energy management of a direct They used a genetic algorithm (GA)-based multi-objective optimization
current (DC) microgrid system using DP to reduce the complexity of the to optimize the production and load for these customers to minimize the
problem. Jafari et al. (2019) used a predictive 2D dynamic programming expenses while providing their demand. Huang et al. (2019) formulated
for microgrid management in grid-connected and off-grid operation a demand response problem as a mixed integer non-linear programming
modes. (MINLP) problem leading to a nondeterministic polynomial time hard
(NP-hard) problem. A global optimization approach that combined a
2.3.3. Multi-objective Optimization PSO algorithm with sequential quadratic programming (SQP) local
In most the articles whose concern is optimal power management in optimizer was proposed to solve this problem.
smart buildings and SG optimization, only one objective function is Jamshidimonfared et al. (2019) proposed a hybrid price-based de
witnessed, and it generally is the cost of energy. However, MSG- mand response (HPDR) strategy. This strategy was implemented in the
connected smart homes have other significant objectives, such as com day-ahead scheduling of a residential microgrid. Liang et al. (2019)
fort, the proportion of utilizing renewable energy, and indoor air qual presented a multi-objective power management problem considering
ity. Therefore, many works have adopted multi-objective strategies. demand response in a microgrid. The multi-objective problem consisted
(Khalid et al., 2018). For instance, Ullah and Hafeez (2021) developed of four conflicting objective functions: the average efficiency function of
an energy optimization model to optimize the smart microgrid’s per the diesel generation unit, the emission of microgrid, the dissatisfaction
formance by reducing operating costs and pollution emissions and caused by demand response, and the total profit function. A distributed
maximizing availability using renewable energy sources (RES). To neuro-dynamic algorithm was proposed for each objective optimization
resolve uncertainty and non-linearity of RES, a hybrid scheme of de problem.
mand response programs and incline block tariff with the participation Guliashki et al. (2019) presented a methodology for energy efficiency
of industrial, commercial, and residential consumers was introduced. An optimization for buildings connected in microgrids. The initial invest
energy optimization strategy based on multi-objective wind-driven ment costs for the building and the energy costs were optimized while
optimization and multi-objective genetic algorithms was utilized to the environmental pollution was minimized simultaneously. A
optimize the objective functions. Das and Mandal (2021) proposed a bi-criterion optimization problem was formulated, and a multi-objective
framework for optimal microgrid scheduling, which minimized the cost genetic algorithm solved it.
of generating units and emissions. The particle swarm optimization Qin et al. (2019) selected an integrated energy system combined with
(PSO) technique was employed to solve this problem. Weighting factors the power grid, natural gas pipeline, district heating network, and
were used for optimization in a multi-objective framework, where costs renewable energy generation to enhance the integration of renewable
and emissions were minimized simultaneously. energy and smooth the load demand profile. They constructed a
Hemeid et al. (2021) proposed a hybrid microgrid renewable energy multi-objective robust optimization model for the integrated energy
system that included PV and wind energy systems, battery banks, and a system based on minimizing the fuel cost, the wind power curtailment,
conventional diesel generator to meet the energy requirements in a and the variance of peak-valley electrical load on the end-user side, as
remote area located in a city in Egypt. Optimization of electricity cost, the objection functions. a PSO algorithm was utilized to solve the
renewable factor, and loss of power supply probability formed the main mentioned model.
3
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
4
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
5
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
The model for time-shiftable loads consides PTSL is the input, while from SG are specified transiently. Moreover, the price of electric power
the output is the work done. from diesel generators is dependent on the price of fuel.
The model for permanent loads input is the fixed amount of power However, the cost of electricity from solar cells and wind turbines is
required, and it cannot be modified. The output is also the work done. different. Indeed, if the total annual maintenance expenses for these
According to Fig. 2, two power outputs and one input power are resources are considered constant, the cost of electrical energy produced
considered for the building. PDesired for the building is as follows. by them depends on their production capability. The production ratio
for these resources is unstable and depends on the intensity of the sun’s
PDesired = PDesired + PDesired + PDesired + PPermanent + PTSL (6)
L T A light and the intensity of the wind. The pricing model considered for
such resources is as follows.
where PDesired
L is the desired power level for the lighting intensity system
and can be determined according to the lighting system model, outside Price =
PProduced
(8)
lighting intensity level, and the desired lighting. Also, PDesired
T is the RMC
desired power level for the temperature control system and can be In the above equation, Price is the cost of energy generated by the
computed using the temperature control model, external temperature resource in kilowatt-hours (kWh), PProduced is the power generated by the
level, and the desired temperature. In addition, PDesired
A is the desired resource each hour (in kilowatts), and RMC is the maintenance cost per
power level for the air-conditioning system. This variable can be hour and is calculated based on the annual maintenance cost.
determined according to the air-conditioning model, outdoor air quality, The following two factors are influential in calculating the cost of the
and the desired air quality. power attainable from the battery.
The smart home internal controller assigns an amount of power to
each load via a trade-off between the cost of each electrical energy unit 1- The maintenance expenditures for the battery
and the desired conditions, which can be either smaller or equal to its 2- The resources that are used for charging the battery
desired power. The overall PDemand value is determined as follows.
To determine the electricity cost obtained from the battery, we
PDemand = PL + PT + PA + PPermanent + PTSL (7)
consider the prices of the resources effective in charging the battery.
PProvided is the amount of power provided by the MSG resource This cost will be zero if the battery is empty. Hence, there are two
management system to the building. Based on the MSG production different states for the battery.
capability or if purchasing from the SG would be economically advan
tageous, this value can be less than or equal to the PDemand. 1- While charging the battery, the total power cost, which is attainable
Noted that, it is economical to buy electricity from SG when the cost from the battery, is updated according to the following equation.
of energy production within the MSG network is higher than the price of
TotalCost(t + 1) = TotalCost(t) + PCharging ∗ SourcePrice (9)
energy in the SG network, or the amount of consumption is more than
the amount of electricity generated by MSG.
where TotalCost represents the total expenses for the power utilized to
3.2. Fuzzy Resource Ranking System charge the battery. PCharging is the power consumed to charge the battery
at time (hour) t, and SourcePrice is the price for each power unit.
The fuzzy system for ranking the energy resources in MSG is designed
as shown in Fig. 2. This system ranks generators or other energy re
2- Similarly, when the battery is discharged, the total power cost is
sources for momentary utilization by taking all the transient prices as according to the following equation.
input and employing rules that resulted from minimizing all fitness
functions. In Table 3, descriptions of this figure are provided. A Mam TotalCost(t + 1) = TotalCost(t) − PDischarging ∗ SourcePrice (10)
dani fuzzy system is employed for ranking the resources. In this system,
the transient price for each resource is the input. Moreover, the system
assigns an instantaneous score to each resource while considering all
target functions. This score, which ranges in [0, 1], specifies the priority where PDischarging is the power drained from the battery at time t.
of operating each resource at any time to provide the power demand. Finally, TotalCost and the maintenance costs for the battery define
The fuzzy system contains five inputs that determine the transient the transient power price for each unit of available electrical energy that
power price of each resource, while the outputs are the scores assigned can be calculated according to the following equation.
to each of them. Similar to the lighting system’s fuzzy controller, the TotalCost
membership functions for the variables are also defined as Gaussian PriceBattery = + RMC (11)
PTotal
constant, with uniform centers in the defining range of each variable.
The price and the amount of electrical energy that can be purchased where PTotal represents the total power in the battery, and RMC is the
hourly maintenance cost. The hourly maintenance cost is calculated
Table 3 based on the annual maintenance cost.
Description of variables and parameters used in the fuzzy ranking system model. Finally, the output from the fuzzy ranking system for energy gener
Symbol Explanation ators within the MSG is a number in the range of [0,1] for each available
resource. Hence, all energy resources can be ranked using these scores.
PriceSolar Solar resource power price: the price for production of electrical power
by the solar cell
PriceWind Wind resource power price: the price for production of electrical power 3.3. Modeling the MSG Resource Management System
the by wind turbine
PriceDiesel Diesel resource power price: the price for production of electrical power
In this system, the decision on exploiting resources is carried out
by a diesel generator
PriceSG SG power price: the price for purchasing electrical power from SG considering all transient prices and the problem objectives. Table 4
PriceBattery Battery power price: the price for electrical power attainable from the provides a further description of the variables in Fig. 1. Using the
battery rankings generated by the corresponding fuzzy system for the electricity
Ranking Resource ranking: transient ranking of the resources or energy production resources, this system determines the share of production for
generators for operation
each resource, purchase/sale from/to the SG network, and the charge/
6
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
EB (t + 1) = EB (t)(1 − σ ) − DP/ηBC (13) On one side, generating electricity in an MSG network is costly, and
in some cases, it will be necessary to purchase electricity from the SG
where, EB represents the battery energy, and ηBC and ηBD demonstrate network. On the other hand, there are some cases where solar and wind
the battery charge and discharge rates, respectively. These numbers are generators produce excessive electricity, or their production costs are so
typically considered to be 90% and 85%. σ is the battery leak rate, which high that it would be more efficient to sell their production. Based on
is generally considered 0.2%. SP and DP are respectively surplus and these situations, the building’s total electrical energy consumption cost
deficit power. In addition, the constraint for these equations is as Eq. can be determined as follows.
(14). ∑( )
TotalPowerCost = PProduced + PBought − PSold (16)
EB,min ≤ EB (t) ≤ EB,max (14) time
As can be seen from the equation, the total cost of generating power
where EB,max and EB,min are the maximum and minimum amounts of
is derived by adding the cost for power production (Pproduced ) to the power
energy stored in the battery, respectively.
purchased from the SG (Pbought ), minus the power sold to the SG (Psold ).
What is evident is that the objective is reducing the cost and thus
3.4. Modelling the Electricity Price Estimator
minimizing the objective function via properly adjusting the rules of the
fuzzy systems of the model.
The electricity price estimator system is demonstrated in Fig. 2. This
system estimates the ultimate electricity unit cost in smart building
management controllers, using the specified ranking of the resources 4.3. Total Energy Consumption at Peak Time (FTPP)
and the overall power requirements. It is worth noting that since PDesired
is generally more or equal to the PDemand, the electricity unit cost esti This objective function is the total power consumed in the building at
mates the actual energy cost determined in the next step. This estimation peak times. The goal is to minimize the function and thus reduce
serves merely to generate an auxiliary variable to be utilized in the smart building consumption at peak times.
building management system.
FTPP = PL + PT + PA + PPermanent + PTSL (17)
7
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
4.4. MSG’s Loss of Power Supply Probability (LPSP) variables, the model can simulate the uncertainty existing in nature.
According to the model considered for the building, uncertainty in the
Reliability is one of the significant objectives in MSG networks, above variables results in uncertainty in the consumption powers PL, PT,
which aims to reduce the probability of power supply loss. This objective and PA. It is worth saying that, depending on the geographical area, the
function is provided according to the following equation (Harkouss average values, and standard deviations for all variables with uncer
et al., 2017). tainty can be explicitly considered for each season or month and,
∑ ( ) therefore, obtain dedicated controllers for any given period of the year.
PProvided − PSolar − PWind + PDiesel + PBattery + PSG
LPSP = time ∑ (18)
time (PProvided ) 6. Experimental Results
PSolar, PWind, PDiesel, PBattery, and PSG indicate the amount of power each
resource produces or exchanges with the SG or the battery. This objec The work process, in general can be summarized in the following
tive function demonstrates the probability of power loss in the MSG steps.
system. Here, the purpose is to minimize the LPSP.
Step 1: Form the structure of fuzzy controllers in different sections.
4.5. The Percentage of Employing Renewable Energy (RE) Step 2: Define the objective functions, constraints, and other
formulas.
One primary objective for operating such a system is to minimize the Step 3: Set up and run a multi-objective optimization algorithm
amount of pollution and GHGs. Using renewable energy resources based on NSGAII.
instead of fossil-based energy resources will cause fewer air pollutants. Step 4: Adjust the parameters of the fuzzy controllers based on the
In the following equation, this objective function is demonstrated (Ari results obtained from the optimization.
kiez et al., 2016): Step 5: Operate the independent system (no need for re-
∑ optimization).
(P + PWind )
RE(%) = time ∑ Solar × 100 (19)
time (PProvided ) We considered two different scenarios. The first model consists of
one smart residential building connected to an MSG; while the second
5. Dealing with Uncertainties one has three residential, commercial, and hotel buildings connected to
the same MSG.
We have applied various probability distribution functions (PDFs) to Various data have been used to run the optimization program. In
address effects of different types of uncertainties in the model. The un summary, essential data is listed as follows. The following information
certainties considered in this study are divided into two categories: the are used as current or average statistical profiles (Sedighizadeh et al.,
uncertainties on the generator side and the uncertainties on the con 2019).
sumer side.
• Instant price of energy that can be purchased from the SG
5.1. Uncertainties on the Generator-side • Light intensity outside the building and the amount of light desired
by the user
The uncertain variables on the generator side are solar radiation • Outdoor room temperature and user’s desired indoor temperature
intensity (SRI) and wind speed (WS). These two variables are considered • Outdoor air quality and indoor air quality desired by the user
randomly, with the Normal and Weibull probability distribution func • The intensity of light radiation to the surface of the solar cell
tions, respectively. Clearly, since the weather conditions significantly • Wind speed in the wind turbine environment
vary all the time, it is necessary to generate each of these PDFs for
specific time intervals. By specifying the above variables, the following variables can also be
The PDF for SRI were examined by researchers and found to be calculated.
normal (Zakariazadeh et al., 2014). No value is reported for hours when
the amount of sunlight is considered insignificant. The well-known • Optimal power value for lighting intensity system: This variable can
Normal PDF has the following definition, where μ and σ represent the be calculated using the lighting system model, outdoor lighting level
mean value and standard deviation of the normal function, respectively. and the desired brightness value identified by the user.
• Optimal power value for temperature control system: This variable
1 − (x− μ)2
y = f (x|μ, σ) = √̅̅̅̅̅e 2σ2 (20) can be calculated using the model of the temperature control system,
σ 2π
outdoor temperature, and the desired temperature valueidentified by
Since the light intensity signal is also non-stationary, it has a distinct the user.
PDF function each hour. For the wind speed, the Weibull PDF function is • Optimal power value for air condition system: This variable can be
usually employed with the following equation. calculated using the model of the ventilation system, outdoor air
( )b quality and air quality desired identified by the user.
− x
• Solar cell production capacity: Knowing the amount of light radia
(21)
a
fw (v) = ba− b vb− 1 e
tion on the surface of the solar cell and the model of this source, the
In this equation, a = vmean /0.9 and b = 2, where vmean is the hourly solar cell production capacity can also be calculated instantly.
average wind speed corresponding to each time interval days (Talari • Wind turbine production capacity: Knowing the amount of wind
et al., 2015). speed to the wind turbine and the model of this source, the wind
turbine production capacity can also be calculated instantly.
5.2. Consumer-side Uncertainties • Power price of solar cell and wind turbine: It is considered with the
price model and based on the amount of production of these re
The uncertainties considered on the consumer side are OL, OT, and sources, the current price of their power can also be calculated.
OA. These signals are non-stationary random variables. OL has the same • Diesel generator power price: By specifying the fuel price and diesel
distribution function as the intensity of light radiation on the solar cell’s generator consumption model, the price of produced power can be
surface, which was described in Section 5.1. For OT and OA , however, calculated.
the Weibull PDF function is employed. By employing the non-stationary
8
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
Fig. 3. The block diagram of the power-based method for one building.
9
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
Fig. 4. Sunlight and wind speed against the electrical power generated by solar cells and wind turbines.
Fig. 5. All the assigned electrical powers for the various loads inside the building, along with the overall power provided.
different in each building, three distinct BMS controllers should perform This paper aimed to use smart fuzzy controllers for indoor con
energy allocation. sumption and MSG resource management. In the previously carried out
After running the NSGAII algorithm and finding the optimal pa research in energy management, the decision variables were often the
rameters and consequently the appropriate rules of fuzzy systems, all the allocated power to loads of the building. At the same time, in this study,
vectors of the objective functions are normalized. Then the smaller the considered mathematical model includes an MSG with a solar cell,
vector is considered the best answer. wind turbine, diesel generator, and solar battery, capable of trading
To comprehensively compare the proposed algorithm with the energy with the SG. Moreover, on the consumer side, a smart building
power-based strategy, we solved the problem using both methods for a has power-shiftable loads consisting of lighting, air conditioning, cool
24-hours interval with similar conditions. The values for the objective ing/heating system, time-shiftable loads, and permanent loads.
functions of the proposed algorithm and the power-based method are Furthermore, uncertainty is regarded both on the consumer and pro
compared in Table 7. The objective functions are determined separately duction sides. In our model, several fuzzy controllers are used to manage
for each building, and the reported values represent the sum of these consumption and resources. In the smart building, four fuzzy controllers
values. As can be seen, the proposed algorithm demonstrates superior are responsible for controlling the power assigned to the loads. To
performance. elaborate, for power-shiftable loads, these controllers determine the
10
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
Fig. 6. A comparison between the proposed algorithm and the power-based approach for the indoor light, temperature, and air quality.
Fig. 7. The outdoor light intensity, temperature, air quality, and values for the three-building types, alongside their desired values.
11
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
12
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
19 Determine the rules of the fuzzy lighting controller system using Hafez, O., & Bhattacharya, K. (2012). Optimal planning and design of a renewable
energy based supply system for microgrids. Renewable Energy, 45, 7–15.
the input parameters.
Harkouss, F., Fardoun, F., & Biwole, P.-H. (2017). Multi-objective decision making
20 Determine the rules of the fuzzy temperature controller system optimization of a residential net zero energy building in cold climate. 2017 Sensors
using the input parameters. Networks Smart and Emerging Technologies.
21 Determine the rules of the fuzzy air condition controller system Hemeid, A. M., & Omer, A. S. (2021). Multi-objective multi-verse optimization of
renewable energy sources-based micro-grid system: Real case. Ain Shams Engineering
using the input parameters. Journal, 13(1).
22 Determine the rules of the fuzzy time-shiftable system using the Huang, Y., Wang, W., & Hou, B. (2019). A hybrid algorithm for mixed integer nonlinear
input parameters. programming in residential energy management. Journal of Cleaner Production, 226,
940–948.
23 Determine the rules of the fuzzy resource ranking system using Huber, M., Sänger, F., & Hamacher, T. (2013). Coordinating smart homes in microgrids:
the input parameters. A quantification of benefits. IEEE PES ISGT Europe 2013.
24 Determine the rules of the fuzzy battery charger system using the Jafari, M., Malekjamshidi, Z., & Zhu, J. (2019). A magnetically coupled multi-port, multi-
operation-mode micro-grid with a predictive dynamic programming-based energy
input parameters. management for residential applications. International Journal of Electrical Power &
25 Determine the initial values for the battery energy variables and Energy Systems, 104, 784–796.
the cost of energy stored in the battery. Jamshidimonfared, H., Ghasemi, A., Loni, A., & Marzband, M. (2019). A hybrid price-
based demand response program for the residential micro-grid. Energy, 185,
26 Calculate the power required for the building to establish optimal 274–285.
conditions for the user using the model of each load. Khalid, A., Javaid, N., Guizani, M., Alhussein, M., Aurangzeb, K., & Ilahi, M. (2018).
27 Score each resource using the controller and rank them. Towards dynamic coordination among home appliances using multi-objective
energy optimization for demand side management in smart buildings. Special Section
28 Supply the energy required according to the resources ranking to
on Energy Management Building, 6, 19509–19529.
estimate the electricity cost for building controllers (this step is Kriett, P. O., & Salani, M. (2012). Optimal control of a residential microgrid. Energy, 42
done only to estimate the price). (1), 321–330.
29 Allocate the power to the indoor lighting system using the Liang, X., He, X., & Huang, T. (2019). Distributed neuro-dynamic optimization for multi-
objective power management problem in micro-grid. Neurocomputing, 362, 51–59.
lighting system controller. Lu, Q., & Zhang, Y. (2021). A multi-objective optimization model considering users’
30 Allocate the power to the indoor temperature system using the satisfaction and multi-type demand response in dynamic electricity price. Energy,
lighting system controller. 240.
Ma, L., Liu, N., Wang, L., Zhang, J., Lei, J., Zeng, Z., Wang, C., & Cheng, M. (2016). Multi-
31 Allocate the power to the indoor air condition system using the party energy management for smart building cluster with PV systems using
lighting system controller. automatic demand response. Energy and Buildings, 121, 11–21.
32 Determine whether time-shiftable loads are on or off using their Mitra, J., Patra, S. B., & Ranade, S. J. (2005). A dynamic programming based method for
developing optimal microgrid architectures. In 15th power systems computational
dedicated controller. conference.
33 Calculate the power required inside the buildings. Moghaddas-Tafreshi, S. M., Mohseni, S., Karami, M. E., & Kelly, S. (2019). Optimal
34 Calculate the provided energy costs for the buildings. energy management of a grid-connected multiple energy carrier micro-grid. Applied
Thermal Engineering, 152, 796–806.
35 Charge the battery using the solar cell energy overflow; the fuzzy Motalleb, M., Annaswamy, A., & Ghorbani, R. (2018). A real-time demand response
controller decides how much energy is spent to charge the battery market through a repeated incomplete-information game. Energy, 143, 424–438.
and how much to sell to the SG. Nguyen, D. H., & Funabashi, T. (2019). Decentralized control design for user comfort and
energy saving in multi-zone buildings. Energy Procedia, 156, 172–176.
36 Charge the battery based on extra power generation.
Nguyen, M. Y., Yoon, Y. T., & Choi, N. H. (2009). Dynamic programming formulation of
37 Calculate the objective functions. micro-grid operation with heat and electricity constraints. In Transmission &
38 Determine the parameters needed to create fuzzy system rules. distribution conference & exposition: Asia and Pacific.
39 Determine the number of generations, population size, mutation Nguyen, T. A., & Crow, M. L. (2012). Optimization in energy and power management for
renewable-diesel microgrids using dynamic programming algorithm. In 2012 IEEE
rate, and crossover rate for the optimization. international conference on cyber technology in automation, control, and intelligent
40 Run NSGAII. systems.
41 Find the best Pareto answers. Nizami, M. S. H., Hossain, M. J., Ruhul Amin, B. M., & Fernandez, E. (2020). A residential
energy management system with bi-level optimization-based bidding strategy for
day-ahead bi-directional electricity trading. Applied Energy, 261.
Oprea, S. V., Bâra, A., Ifrim, G. A., & Coroianu, L. (2019). Day-ahead electricity
References consumption optimization algorithms for smart homes. Computers & Industrial
Engineering, 135, 382–401.
Osmani, A., & Zhang, J. (2014). Optimal grid design and logistic planning for wind and
Akbari-Dibavar, A., Nojavan, S., Mohammadi-Ivatloo, B., & Zare, K. (2020). Smart home
biomass based renewable electricity supply chains under uncertainties. Energy, 70,
energy management using hybrid robust-stochastic Optimization. Computers &
514–528.
Industrial Engineering, 143.
Pachauri, S., van Ruijven, B. J., Nagai, Y., Riahi, K., van Vuuren, D. P., Brew-
Arikiez, M., Grasso, F., Kowalski, D., & Zito, M. (2016). Heuristic algorithm for
Hammond, A., & Nakicenovic, N. (2013). Pathways to achieve universal household
minimizing the electricity cost of air conditioners on a smart grid. In 2016 IEEE
access to modern energy by 2030. Environmental Research Letters, 8(2), Article
International Energy Conference.
024015.
Das, G., & Mandal, M. D. K. K. (2021). Multi-objective optimization of hybrid renewable
Parisio, A., & Glielmo, L. (2011). A mixed integer linear formulation for microgrid
energy system by using novel autonomic soft computing techniques. Computers and
economic scheduling. In 2011 IEEE international conference on smart grid
Electrical Engineering, 94.
communications.
De Oliveira, G., Jacomino, M., Ha, D. L., & Ploix, S. (2011). Optimal power control for
Park, K., Lee, W., & Won, D. (2019). Optimal energy management of DC microgrid
smart homes. IFAC Proceedings Volumes, 44(1), 9579–9586.
system using dynamic programming. IFAC-Papers OnLine, 52(4), 194–199.
Deb, K., Pratap, A., Agarwal, S., & Meyarivan, T. (2002). A fast and elitist multi-objective
Peng, H., Wang, C., & Han, Y. (2022). Micro multi-strategy multi-objective artificial bee
genetic algorithm: NSGA-II. IEEE transactions on evolutionary computation, 6(2),
colony algorithm for microgrid energy optimization. Future Generation Computer
182–197.
Systems, 131, 59–74.
Ferrari, L., Esposito, F., Becciani, M., Ferrara, G., Magnani, S., Andreini, M.,
Qin, C., Yan, Q., & He, G. (2019). Integrated energy systems planning with electricity,
Bellissima, A., Cantù, M., Petretto, G., & Pentolini, M. (2017). Development of an
heat and gas using particle swarm optimization. Energy, 188.
optimization algorithm for the energy management of an industrial Smart User.
Reka, S. S., & Ramesh, V. (2016). Demand side management scheme in smart grid with
Applied energy, 208, 1468–1486.
cloud computing approach using stochastic dynamic programming. Perspectives in
Fossati, J. P., Galarza, A., Martín-Villate, A., & Fontan, L. (2015). A method for optimal
Science, 8, 169–171.
sizing energy storage systems for microgrids. Renewable Energy, 77, 539–549.
Safaei, A., Freire, F., & Antunes, C. H. (2015). A life cycle multi-objective economic and
Gellings, C. W. (2009). The smart grid: Enabling energy efficiency and demand response. The
environmental assessment of distributed generation in buildings. Energy Conversion
Fairmont Press, Inc.
and Management, 97, 420–427.
Gomes, I. L. R., Melicio, R., & Mendes, V. M. F. (2021). A novel microgrid support
Sedighizadeh, M., Esmaili, M., Jamshidi, A., & Ghaderi, M.-H. (2019). Stochastic multi-
management system based on stochastic mixed-integer linear programming. Energy,
objective economic-environmental energy and reserve scheduling of microgrids
223.
considering battery energy storage system. International Journal of Electrical Power &
Guliashki, V. G., Marinova, G. I., & Groumpos, P. P. (2019). Multi-Objective
Energy Systems, 106, 1–16.
Optimization Approach for Energy Efficiency in Microgrids. IFAC-PapersOnLine, 52
(25), 477–482.
13
Y. Mohammadi et al. Sustainable Cities and Society 86 (2022) 104119
Shakouri, G., & Kazemi, H. A. (2017). Multi-objective cost-load optimization for demand Yu, D., Zhang, T., He, G., Nojavan, S., Jermsittiparsert, K., & Ghadimi, N. (2020). Energy
side management of a residential area in smart grids. Sustainable Cities and Society, management of wind-PV-storage-grid based large electricity consumer using robust
32, 171–180. optimization technique. Journal of Energy Storage, 27.
Siano, P. (2014). Demand response and smart grids—A survey. Renewable and Sustainable Zakariazadeh, A., Jadid, S., & Siano, P. (2014). Smart microgrid energy and reserve
Energy Reviews, 30, 461–478. scheduling with demand response using stochastic optimization. International Journal
Stimmel, C. L. (2015). Building smart cities: Analytics, ICT, and design thinking. Auerbach of Electrical Power & Energy Systems, 63, 523–533.
Publications. Zareipour, H., Bhattacharya, K., & Canizares, C. (2004). Distributed generation: Current
Sun, S., & Wang, C. (2022). Multi-objective optimization dispatching of a micro-grid status and challenges. In Annual North American power symposium.
considering uncertainty in wind power forecasting. Energy Reports, 9, 2859–2874. Zhang, D., Liu, S., & Papageorgiou, L. G. (2014). Fair cost distribution among smart
Talari, S., Yazdaninejad, M., & Haghifam, M.-R. (2015). Stochastic-based scheduling of homes with microgrid. Energy Conversion and Management, 80, 498–508.
the microgrid operation including wind turbines, photovoltaic cells, energy storages Zhang, D., Shah, N., & Papageorgiou, L. G. (2013). Efficient energy consumption and
and responsive loads. IET Generation, Transmission & Distribution, 9(12), 1498–1509. operation management in a smart building with microgrid. Energy Conversion and
Ullah, K., & Hafeez, G. (2021). A multi-objective energy optimization in smart grid with Management, 74, 209–222.
high penetration of renewable energy sources. Applied Energy, 299. Zhang, Z., Jing, R., Lin, J., Wang, X., van Dam, K. H., Wang, M., Meng, C., Xie, S., &
Wang, Z., Yang, R., & Wang, L. (2011). Intelligent multi-agent control for integrated Zhao, Y. (2020). Combining agent-based residential demand modeling with design
building and micro-grid systems. optimization for integrated energy systems planning and operation. Applied Energy,
Wolsink, M. (2012). The research agenda on social acceptance of distributed generation 263.
in smart grids: Renewable as common pool resources. Renewable and Sustainable
Energy Reviews, 16(1), 822–835.
14