Content uploaded by Nadeem Javaid
Author content
All content in this area was uploaded by Nadeem Javaid on Aug 16, 2017
Content may be subject to copyright.
Towards optimal sizing and integration of large
scale penetration of renewable energy sources
By
Sardar Mehboob Hussain
CIIT/FA15-RCS-025/ISB
MS Thesis
In
Computer Science
COMSATS Institute of Information Technology
Islamabad - Pakistan
Spring, 2017
COMSATS Institute of Information Technology
Towards optimal sizing and integration of large
scale penetration of renewable energy sources
A Thesis Presented to
COMSATS Institute of Information Technology, Islamabad
In partial fulllment
of the requirement for the degree of
MS (Computer Science)
By
Sardar Mehboob Hussain
CIIT/FA15-RCS-025/ISB
Spring, 2017
ii
Towards optimal sizing and integration of large
scale penetration of renewable energy sources
A Post Graduate Thesis submitted to the Department of Computer Science as
partial fullment of the requirement for the award of Degree of MS (Computer Sci-
ence).
Name Registration Number
Sardar Mehboob Hussain CIIT/FA15-RCS-025/ISB
Supervisor:
Dr. Nadeem Javaid
Associate Professor,
Department of Computer Science,
COMSATS Institute of Information Technology (CIIT),
Islamabad Campus.
iii
Final Approval
This thesis titled
Towards optimal sizing and integration of large
scale penetration of renewable energy sources
By
Sardar Mehboob Hussain
CIIT/FA15-RCS-025/ISB
has been approved
For the COMSATS Institute of Information Technology, Islamabad
External Examiner:
Dr. Waseem Shahzad
Associate Professor and HoD, Dept. of Computer Science,
FAST National University, Islamabad
Supervisor:
Dr. Nadeem Javaid
Associate Professor, Dept. of Computer Science,
COMSATS Institute of Information Technology, Islamabad
HoD:
Dr. Majid Iqbal Khan
Associate Professor, Dept. of Computer Science,
COMSATS Institute of Information Technology, Islamabad
iv
Declaration
I Sardar Mehboob Hussain (Registration No. CIIT/FA15-RCS-025/ISB) hereby de-
clare that I have produced the work presented in this thesis, during the scheduled
period of study. I also declare that I have not taken any material from any source
except referred to wherever due that amount of plagiarism is within acceptable range.
If a violation of HEC rules on research has occurred in this thesis, I shall be liable to
punishable action under the plagiarism rules of the HEC.
Date: June, 2017 Sardar Mehboob Hussain
CIIT/FA15-RCS-025/ISB
v
Certicate
It is certied that Sardar Mehboob Hussain (Registration No. CIIT/FA12-RCS-025/ISB)
has carried out all the work related to this thesis under my supervision at the De-
partment of Electrical Engineering, COMSATS Institute of Information Technology,
Islamabad and the work fulls the requirement for award of MS degree.
Date: June, 2017
Supervisor:
Dr. Nadeem Javaid
Associate Professor, Department of Computer Science
Head of Department:
Dr. Majid Iqbal Khan
Department of Computer Science
vi
Contents
Dedication ix
Acknowledgements x
Abstract xi
List of Publications xii
List of Figures xiii
1 Introduction 1
1.1 ProblemDescription ........................... 4
1.2 Thesisorganization............................ 7
2 Literature Review 8
3 Proposed Solution 13
3.1 SystemModel............................... 14
3.1.1 Power Output of Renewable Energy Sources . . . . . . . . . . 18
3.1.2 Load Categorization . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.3 Energy Consumption Model . . . . . . . . . . . . . . . . . . . 22
3.1.4 Capacity Constraints . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.5 Thermal Storage Constraints . . . . . . . . . . . . . . . . . . . 23
3.1.6 Electric Storage Constraints . . . . . . . . . . . . . . . . . . . 24
3.1.7 EnergyBalance.......................... 24
vii
3.1.8 Start and End Time Horizon . . . . . . . . . . . . . . . . . . . 25
3.1.9 PowerDemand .......................... 25
3.1.10 Peak to Average Ratio . . . . . . . . . . . . . . . . . . . . . . 26
3.1.11WaitingTime........................... 26
3.1.12 Objective Function . . . . . . . . . . . . . . . . . . . . . . . . 26
4 Heuristic Techniques 28
4.0.1 Teaching Learning-Based Optimization . . . . . . . . . . . . . 29
4.0.2 Enhanced Dierential Evolution . . . . . . . . . . . . . . . . . 30
4.0.3 Enhanced Dierential Teaching Learning Algorithm . . . . . . 31
4.1 Feasible Region of Objective Function . . . . . . . . . . . . . . . . . 32
5 Simulations and Discussion 35
5.1 Simulations and Discussion . . . . . . . . . . . . . . . . . . . . . . . . 36
5.1.1 Electricity Demand . . . . . . . . . . . . . . . . . . . . . . . . 37
5.1.2 ElectricityCost.......................... 40
5.1.3 Peak to Average Ratio . . . . . . . . . . . . . . . . . . . . . . 44
5.1.4 UserDiscomfort.......................... 46
5.1.5 HeatDemand........................... 47
5.1.6 Execution Time and Performance Tradeos . . . . . . . . . . 48
5.2 Life Cycle Energy Analysis . . . . . . . . . . . . . . . . . . . . . . . 49
6 Conclusion 51
6.1 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . 52
viii
DEDICATION
𝒟
edicated
to my elder brother Matloob Hussain, my cousin brother Sardar Sajid
Imtiaz, my family, teachers and Google.com
ix
ACKNOWLEDGEMENTS
Alhamdulillah, all praises to Allah Almighty, the most Merciful and the most Gra-
cious, for the strengths and His blessing in completing this thesis.
Working on thesis was indeed a challenging task that demanded immense input
and application. I feel proud to express my deepest sense of gratitude and apprecia-
tion to my supervisor Dr. Nadeem Javaid for his kind help, advice, inspired guidance,
unlimited support, sympathetic attitude and sincere personal involvement through-
out the study.
I would never have been able to reach this stage but for the prayers and great
support of my family. I am also thankful to my parents who always give me lots of
encouragement. Thanks and best wishes for all those who have made this learning
experience so wonderful for me.
x
ABSTRACT
Energy Management in Microgrids Using Swarm Intelligence
Optimization Techniques
Today’s buildings are responsible for about 40% of total energy consumption and
30–40% of carbon emissions, which are key concerns for the sustainable development
of any society. The excessive usage of grid energy raises sustainability issues in the
face of global changes, such as climate change, population, economic growths, etc.
Traditionally, the power systems that deliver this commodity are fuel operated and
lead towards high carbon emissions and global warming. To overcome these issues,
the recent concept of the nearly zero energy building (nZEB) has attracted numerous
researchers and industry for the construction and management of the new genera-
tion buildings. In this regard, this paper proposes various demand side management
(DSM) programs using the genetic algorithm (GA), teaching learning-based optimiza-
tion (TLBO), the enhanced dierential evolution (EDE) algorithm and the proposed
enhanced dierential teaching learning algorithm (EDTLA) to manage energy and
comfort, while taking the human preferences into consideration. Power consumption
patterns of shiftable home appliances are modied in response to the real-time price
signal in order to get monetary benets. To further improve the cost and user discom-
fort objectives along with reduced carbon emission, renewable energy sources (RESs)
are also integrated into the microgrid (MG). The proposed model is implemented in
a smart residential complex of multiple homes under a real-time pricing environment.
We gure out two feasible regions: one for electricity cost and the other for user
discomfort. The proposed model aims to deal with the stochastic nature of RESs
while introducing the battery storage system (BSS). The main objectives of this pa-
per include: (1) integration of RESs; (2) minimization of the electricity bill (cost) and
discomfort; and (3) minimizing the peak to average ratio (PAR) and carbon emission.
Additionally, we also analyze the tradeo between two conicting objectives, like elec-
tricity cost and user discomfort. Simulation results validate both the implemented
and proposed techniques.
xi
Conference Proceedings
Dr. Nadeem Javaid, Associate Professor (Supervisor)
Sardar Mehboob Hussain
1
Sardar Mehboob Hussain
, Ayesha Zafar, Rabiya Khalid, Samia Abid, Umar
Qasim, Zahoor Ali Khan and Nadeem Javaid,“An ecient scheduling of elec-
trical appliance in micro grid based on heuristic techniques", The 11th Inter-
national Conference on Complex, Intelligent, and Software Intensive Systems
(CISIS), 2017.
Download
2 Samia Abid, Rabiya Khalid, Ayesha Zafar,
Sardar Mehboob Hussain
, Has-
san Rahim, and Nadeem Javaid. “An Optimized Priority Enabled Energy Man-
agement System for Smart Homes." In Advanced Information Networking and
Applications (AINA), 2017 IEEE 31st International Conference on, pp. 1035-
1041. IEEE, 2017.
Download
3 Naseem, Mudassar, Samia Abid, Rabia Khalid, Ghulam Hafeez,
Sardar Mahboob Hussain
,
and Nadeem Javaid.“Towards Heuristic Algorithms: GA, WDO, BPSO, and
BFOA for Home Energy Management in Smart Grid." In International Confer-
ence on Broadband and Wireless Computing, Communication and Applications,
pp. 267-278. Springer International Publishing, 2016.
Download
4 Zafar, Ayesha, Samia Abid, Rabiya Khalid,
Sardar Mehboob Hussain
, Has-
san Rahim, and Nadeem Javaid.“A meta-heuristic home energy management
system."
Download
xii
List of Figures
1-1 Graphical representation of nearly zero energy building (nZEB) bal-
anceconcept. ............................... 3
1-2 Smart residential complex. . . . . . . . . . . . . . . . . . . . . . . . . 7
3-1 Architecture of the proposed demand side management (DSM). . . . 14
3-2 Battery charge and discharge levels . . . . . . . . . . . . . . . . . . . 17
3-3 Ambient temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3-4 Estimated solar radiations . . . . . . . . . . . . . . . . . . . . . . . . 18
3-5 Power ow of renewable energy sources (RESs) in the proposed model. 20
4-1 Feasible region of objective function. . . . . . . . . . . . . . . . . . . 33
4-2 Feasible region of tradeo. . . . . . . . . . . . . . . . . . . . . . . . . 34
5-1 Real time porice signal . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5-2 Hourly electricity demand. Teaching learning-based optimization (TLBO);
enhanced dierential evolution (EDE); and enhanced dierential teach-
ing learning algorithm (EDTLA). . . . . . . . . . . . . . . . . . . . . 38
5-3 Hourly electricity demand with RESs and battery storage system (BSS). 40
5-4 Estimated energy generation. . . . . . . . . . . . . . . . . . . . . . . 40
5-5 Hourly electricity cost. . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5-6 Daily electricity bill. . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5-7 Hourly electricity bill with RESs and BSS. . . . . . . . . . . . . . . . 43
5-8 Daily electricity bill with RESs and BSS. . . . . . . . . . . . . . . . . 43
5-9 Peak to average ratio (PAR). . . . . . . . . . . . . . . . . . . . . . . 45
xiii
Chapter 1
Introduction
1
Chapter 1
According to the European commission’s report [1], buildings consume about 40%
of overall energy and are responsible for 30–40% of carbon emissions. As energy,
water, land and other resources are required for the construction, maintenance, control
and demolition of all buildings [2, 3], so the carbon emissions and wastes due to
building construction and maintenance cannot be neglected. Due to all of these
reasons, scientist and researchers began to start using passive techniques for building
construction and active techniques for control and management while taking into
account human needs regarding comfort and green environment.
In addition, the European “Energy Performance of Buildings Directive (EPBD)”
released in 2010 and “Energy Eciency Directive (EED)” released in 2012 led the
member countries of the European Union to readjust their legislation regarding build-
ing energy management for sustainable development of society. According to this,
from the year 2019, all new buildings will be nearly zero energy buildings (nZEBs),
and by the end of year 2020, all buildings will have to maintain nZEB [4,5]; where the
term zero energy building (ZEB) is dened as: “an energy ecient buildings where
annual energy delivered to home/residential sector is less than or equal to the total
energy generation from on-side or standalone renewable energy sources (RESs)” [6].
However, the concept of ZEB can be characterized as: (i) grid-connected ZEB; and
(ii) stand-alone or autonomous ZEBs. The stand-alone ZEBs are further separated
into three categories [7]; Figure 1-1.
1. nZEB: a ZEB connected to grid having a nearly zero energy balance. This
means that the energy consumption in any building or sector is slightly greater
than the total renewable energy.
2. net zero energy building (NZEB): a ZEB connected to grid having zero energy
balance. The total energy consumption and generation are almost equal.
3. positive energy building (PEB): has a positive energy balance. The energy
consumption in PEB is less than the energy generation from renewable sources
where surplus energy is sold back to the grid.
2
Chapter 1
Energy
balance Reference
building
Net zero balance line
Weighted demand
[kWh, CO2, etc.]
Weighted supply
[kWh, CO2, etc.]
Net zero energy
Nearly net zero
energy
Energy efficiency
Figure 1-1: Graphical representation of nearly zero energy building (nZEB) balance
concept.
In traditional energy generation systems, fossil fuels dominate as the power gener-
ation sources and are responsible for greenhouse gas (GHG) emissions. The challenge
is not only to reduce GHG emissions, but also to increase electricity generation in view
of socio-economic aspects of generation. RESs are considered as future replacement
with zero carbon emission and low price electricity producers. RESs are intermittent,
uncertain and random in nature; they do not produce a xed amount of energy and
are heavily dependent on weather, season and area. Integration of advanced informa-
tion and communication technology (ICT) into the traditional power grid forms the
smart grid (SG) [8].
An emerging type of distributed generation network called the microgrid (MG) is
perceived as a medium voltage or low voltage power system with small distributed
generation, few controllable distributed loads and on-site energy storage. One of the
key underpinnings of MG, which allows it to dominate traditional grids, is the usage
of RESs at the consumer level. Improvisation of the power grid has led to many
challenges and issues, e.g., protection, selectivity, security, adaptability, scalability,
reliability and many more [9]. Hybrid RESs with a battery storage system (BSS) are
intensively discussed in the literature and, therefore, are widely accepted in order to
3
Chapter 1
cater to the uncertainty of RESs. Two major parties are involved in the operation
of MG, i.e., consumers and the utility. Demand side management (DSM) oers de-
mand response (DR) programs to the residential consumers in order to change daily
electricity usage pattern in response to some incentives. These incentives are usually
monetary rebates.
Major objectives of SG include reduced electricity bill, reduced peak to average
ratio (PAR), maximized user comfort and balanced power consumption [10]. In [11],
the authors use integer linear programming (ILP) to reduce the daily electricity bill of
residential consumers. An approximate dynamic programming (DP) is used in [12] to
reduce the electricity burden and PAR on the main power grid. An energy sale mecha-
nism among dierent MGs is proposed using the game theoretic approach. A particle
swarm optimization (PSO)-based DR program is discussed in [13] to curtail PAR
and minimize daily consumption cost in the presence of RESs. In [14], two heuris-
tic techniques, i.e., teaching learning-based optimization (TLBO) and the dierential
evolution (DE) algorithm are used to reduce cost and increase the comfort level of
users.
In this paper, we design an energy management model (EMM) in nZEB using
genetic algorithm (GA), TLBO, enhanced dierential evolution (EDE) and our novel
proposed EDTLA. Our main focus is RE integration and local distributed energy
resources (DERs) scheduling in order to meet electricity and heat demands while
reducing carbon emissions. We also compute the feasible regions for cost and user
discomfort. A tradeo between cost and comfort is also shown under the four dierent
techniques. For this study, the time interval and time slot are used interchangeably.
Similarly, electrical loads and electric tasks can be referred to as home appliances.
1.1 Problem Description
In SG, electricity bill minimization, power consumption minimization, PAR reduc-
tion, user comfort maximization and RESs integration are key challenges. Nu-
merous mathematical and heuristic-based strategies have been proposed to deal with
4
1.1. Problem Description Chapter 1
these optimization problems. Predominantly, user comfort is ignored to minimize
the inevitably growing electricity bill problem. Existing DSM techniques target the
electricity bill reduction while neglecting either PAR or the user comfort level. The
randomness of the power usage pattern aects the optimal energy consumption sched-
ule and user cost minimization at the consumer level. The authors in [28] proposed
an MILP-based energy consumption model to reduce electricity consumption cost
and GHG emissions. However, integration of RESs, PAR and user comfort are not
tackled in the proposed model. However, [11] considered the electricity bill minimiza-
tion along with user comfort maximization. The authors use ILP to solve the convex
nature of the optimization problem, and a tradeo between cost and discomfort is
also computed. Although the scheduling strategy reduces the energy expense, RESs
integration can further decrease the electricity bill, carbon emissions and PAR. Resi-
dential load scheduling and power trading among dierent homes in the presence of
RESs is discussed in [12]. Stochastic optimization is used for the residential schedul-
ing, whereas a game theoretic approach is adopted for power trading among multiple
homes. PAR and user comfort level are not considered, which are crucial parameters
of SG. No proper mechanism is provided for power ow from one user to another. A
GA-based DR program for HEMS is proposed in [31]. The scope is limited to only
one home, and no RESs are incorporated. Furthermore, user comfort level is dis-
turbed. Additionally, mathematical methods also require long computational time.
These proposed techniques are limited to be applicable on a single home and may not
result in the optimal solution when extended to a large scale.
Therefore, in this paper, several smart homes in a smart residential complex are
considered, as depicted in Figure 1-2. The building has its own MG as the local en-
ergy provider. The smart building demands electricity and heat, prior to scheduling
daily appliances available in each smart home and later to maintain the inner temper-
ature of the building. DERs are also available in the smart building; however, some
resources only cater to the heat demand, while others deal with electricity demand.
The electricity demand is fullled by the energy generated by PV, WT and energy
imported from the main grid. A storage system is incorporated to store energy in
5
Chapter 1
order to use later whenever required. We compute feasible regions, and heuristic
techniques are applied to validate that the obtained solution lies within the bounded
region. Additionally, a feasible region of tradeo between electricity cost and delay is
also obtained to show an equilibrium between cost and discomfort. Three heuristic-
based techniques, i.e., GA, TLBO and EDE, are employed on the aforementioned
scenario, and a novel EDTLA is proposed in this study to minimize the total elec-
tricity cost and PAR. The newly-proposed EDTLA is a hybrid of EDE and TLBO.
TLBO sometimes gets stuck in local minima, so we increase the diversity of the search
by integrating mutation and crossover steps of EDE in TLBO. The procedural steps
of the novel hybrid algorithm are also provided in Algorithm
??
. A tradeo between
cost and user comfort is also analyzed. Moreover, the usage of local DERs contributes
to lowering the harmful carbon emissions. The comprehensive problem can be stated
as hereunder:
Provided are: (a) the scheduling time window; (b) the earliest starting and latest
nishing time horizons; (c) the number of loads and respective power rating; (d) the
length of the operation time interval; (e) the total heat demand of smart building;
(f) the specications of DERs; (g) the RTP signal and natural gas price; (h) the
maintenance cost; (i) the minimum charge and maximum discharge limits; (j) the
capacity constraints of thermal and electrical storage; (k) the heat to power ratio.
Figure out: (a) the appliance schedule plan; (b) PAR; (c) the waiting time; (d)
the energy generation plan; (e) the energy storage plan; (f) the power purchased from
the grid; (g) the local energy harvesting and import from the main grid.
So as to nd: (a) the optimal consumption pattern; (b) the minimum electricity
bill; (c) meet electricity and heat demand; (d) the reduced PAR and discomfort; (e)
economic and environmentally-friendly generation.
6
1.2. Thesis organization Chapter 1
Figure 1-2: Smart residential complex.
1.2 Thesis organization
Remainder of the thesis is organised as follows. Section II comprises of recent related
work. Section III depicts the detail description of proposed model. Linear optimiza-
tions model is discussed in Section IV. Simulation ndings and results are inscribed in
Section V. Lastly, in Section VI, concluding remarks are presented followed by future
work.
7
Chapter 2
Literature Review
8
Chapter 2
In order to optimally schedule home tasks and DERs in residential MG, several
methods have been proposed in the literatures [9–30]. Some of the recent approaches
are discussed hereunder.
The analysis and sizing of RE in coordination with BSS is discussed in [15], and
a hybrid model is proposed using mixed ILP (MILP). One year of available weather
data is used to predict the weather prole of the next three years and then is used to
gure out the optimal size of the wind turbine (WT), photovoltaic (PV) and thermal
load prole for the residential building. The authors maximize the use of RE and
reduce the burden of high power demand at the grid.
In [16], the authors present a complete nZEB framework and propose various
methods regarding the implementation point of view. In another work [4], the au-
thors implement an NZEB concept and propose a fuzzy logic-based energy man-
agement system for lighting, shading and HVAC systems. The authors implement
various congurations, while taking into consideration walls, window geometry and
glass properties. Regarding implementation, it was found that for a given amount
of solar radiation, each room requires a diversied management system to maintain
balance between comfort and energy management. A GUI-based energy and com-
fort management system for ZEB is proposed in [7]. A multiagent system is used to
control distributed loads, while a particle swarm optimization algorithm is used to
manage comfort and energy in the residential sector.
An active controller is proposed in [13] to optimally integrate the heating and
cooling system in MG. The research improves the reliability of MG and minimizes
the cost of MG, the size of RE resources and imported energy from the grid. The
main purpose of this study is to minimize peak load and consumption cost.
In [17], the authors propose an SG equipped with 100% RESs to satisfy electricity
and heating demands. BSS is used to deal with the uctuating behavior of RESs.
Combined heat and power (CHP) plants and district heating and cooling systems
are introduced, which are responsible for providing heating and cooling loads to the
households and other commercial buildings.
9
Chapter 2
A cooperative interaction between the AD system (ADS) connected to multiple
grids and the energy system is formulated in [18], and a dynamic energy management
strategy is proposed. The rst interaction is between MG and ADNs, whereas,the
other is among dierent MGs. The authors propose a dynamic energy management
technique for cooperation between MG and ADSs that caters to the inuences of
the high penetration of RESs. The work in [19] considered real-time energy storage
management to increase the RE share in MG. The authors use an o-line algorithm
for optimization and proposed a novel sliding window-based on-line algorithm. The
main objectives of the research are to minimize the cost of power purchased from the
grid and to maximize the penetration of RESs in MG. The cost function is formulated
by the strictly convex function and solved using DP.
In [20], the joint operation of energy storage and load scheduling with RESs is
considered in the residential domain. Electricity demand, starting times of appliances,
the length of operation times of appliances and RE generation are considered as
random and stochastic. The stochastic nature of the problem is solved by modifying
the Lyapunov optimization technique. Regarding ZEB, an nZEB can be achieved by
integrating RESs, such as solar and wind. In [21], the authors consider Vietnam,
where solar energy is infrequently used in residential sectors. To promote energy
management along with the integration of solar energy, a solar panel of 15-kW capacity
is installed in the rooftop to compensate energy demand. However, prior to the
installation of the solar panels, it is required to estimate the energy obtained from
these panels. For this purpose, the PVSYST simulation tool has been used.
A home energy management system (HEMS) is proposed in [22,23], using dierent
heuristic techniques. GA, BPSO and ACO are used to design a HEMS scheduler,
which optimally schedules home appliances under large-scale penetration of RESs.
The main objectives are to reduce the daily electricity bill and PAR.
In [11], an energy control system is proposed in a smart home of the residential
domain. Dierent types of appliances are scheduled according to the given time frame.
The optimization problem is solved using LP. The major objective is to reduce the
electricity bill. A tradeo between cost and discomfort is also analyzed using the
10
Chapter 2
Taguchi loss function.
Day-ahead scheduling of all resources for optimal operation of MG is proposed
in [24]. The authors claim that one-day-ahead scheduling can avoid vulnerabilities and
ensures the consistent operation of MG. An agent-based modeling (ABM) technique
is used where each agent acts as a bus and provides information about losses and
other attacks.
The participation of dierent DSM strategies in HEMS is analyzed in [25]. The
major focus of this paper is to develop HEMS and DSM systems in order to reduce
the electricity bill and maximize RE usage. The use of dierent incentive-based
algorithmic techniques in DSM is analyzed, and their impact is elaborated.
The load scheduling and power trading problems in the residential area of MG with
a large share of RE are discussed in [12]. An approximate DP is used for appliance
scheduling, and a game theoretic approach is used for power trading among dierent
users. All users, having excess generation, participate in a gamble, and the rst
winner is prioritized to sell excess generation rst. In this way, every user reduces
power usage and tries to sell maximum energy, which generates revenue and lowers
the electricity bill.
In [26], electric vehicles (EVs) are integrated with MG in the presence of RESs.
The major focus is to reduce power losses and improve the stability of MG under the
large-scale integration of EVs. The DE algorithm is used to solve the multi-objective
nature of the problem.
In [27], the authors use the real time pricing (RTP) signal in DSM programs
to reduce the daily electricity bill and PAR. A new load scheduling learning (LSL)
algorithm is proposed, which schedules appliances after learning from a series of
actions. The change in load scheduling, power demand and pricing signal are modeled
as a Markov decision process, and their information is stored with respect to each
time slot.
An MG is formed with local DERs in which heat and electricity demand is provided
to consumers in [28]. The former is provided by local generators like the CHP and
boiler, whereas the latter is provided by WT, PV generation and energy import
11
Chapter 2
from the main grid. The authors aim at reducing the electricity expense and carbon
emissions.
12
Chapter 3
Proposed Solution
13
Chapter 3
3.1 System Model
DSM plays a vital role in ecient and reliable operation of SG. Adopting dierent
mechanisms, DSM benets the end users and utilities under two major functionali-
ties, i.e., ecient energy management and control over the end users’ activities. In
the residential sector, each home is equipped with advanced metering infrastructure
connected to a central controller in order to ensure stable and optimized energy con-
sumption decisions under two-way communication between utilities and consumers.
A conceptual diagram of the proposed DSM mechanism is illustrated in Figure 3-1.
This structure enables users to reduce the electricity bill and the utility to curtail
PAR for persistent operation of SG. All appliances request the central scheduler to
execute the job, and the scheduler makes the decision about the status of appliances
at a particular hour. The scheduler must respect the scheduling horizon provided
earlier by users.
Figure 3-1: Architecture of the proposed demand side management (DSM).
Our system model is composed of a smart building of 30 homes in an MG scenario.
Each home is equipped with 12 smart appliances, which are to be scheduled within
a given time window as shown in Table 3.1. All appliances must not start before
the earliest starting time and nish respective working hours prior to nishing the
time horizon. An assumption is made that all users are living with same power
14
3.1. System Model Chapter 3
consumption habits, and only once a day, an appliance is required to operate. The
start and end times of appliances are assumed to be provided by users, whereas other
parametric values are listed in Table 3.1 [32]. Figure 1-2 shows the block diagram of
the smart building and local generation sources.
Table 3.1: Parameters of the appliances.
Task Power
(kW)
Earliest
Start-
ing
Time
(h)
Latest
Finish-
ing
Time
(h)
Time
Window
Length
(h)
Duration
(h)
Dish
washer 1.5 9 17 8 2
Cloth
washer 1.5 9 12 3 1.5
Spin dryer 2.5 13 18 5 1
Cooker
hob 3 8 9 0.5 0.5
Cooker
oven 5 18 19 0.5 0.5
Microwave 1.7 8 9 0.5 0.5
Lighting 0.84 18 24 6 6
Laptop 0.1 18 24 6 2
Desktop 0.3 18 24 6 3
Cleaner 1.2 9 17 8 0.5
Fridge 0.3 0 24 - 24
Electric
car 3.5 18 8 14 3
A time interval of half an hour is considered because the minimum of one hour
15
Chapter 3
operation time for home appliances seems impractical. Some home appliances like the
coee maker and the toaster work for less than one hour a day. All appliances have
constant power consumption rates; however, the power consumption cost depends on
the number of time intervals an appliance runs and the price of electricity during
execution cycle.
In addition to the above, the smart residential building also requires heat demand
along with the ground area of 2500 m
2
, calculated using CHP sizer Version 2 software
(Oak Ridge, TN 37830, USA) [33]. No electricity from the utility is imported to
satisfy the heat demand; instead, the smart building has local DERs, which are to be
scheduled according to the heat demand curve. DERs and their respective capacities
are assumed to be known, which are listed in Table 3.2 [28]. The operation and
maintenance costs of DERs are based on natural gas and other specications are:
∙
a CHP production plant with a 1.2 heat to power ratio.
∙
a boiler of 120-kW capacity.
∙
one BSS with charge and discharge eciencies of 90%.
∙
a gas connection for the CHP and boiler to run.
∙
the total payable cost depends on the electricity price, the natural gas price and
the operation cost.
Table 3.2: Technical parameters of distributed energy resources (DERs).
Resource Capacity Eciency
(%)
Operation/Maintenance
Cost (%)
CHP 20 kW 40 2.7 cents/kWh
Boiler 120 kW 85 2.7 cents/kWh
Storage 10 kWh 90 0.5 cents/kWh
16
3.1. System Model Chapter 3
8:00 12:00 16:00 20:00 0:00 4:00 8:00
Time (hours)
0
100
200
300
400
500
600
Battery chanrge and discharge (kW)
Unscheduled
GA
TLBO
EDE
EDTLA
Figure 3-2: Battery charge and discharge levels
.
Total electricity demand is satised by local generation plus energy imported from
the main power grid. BSS is used to store excess electricity generated from RESs and
used later when high price hours at the grid or no RE is available. Battery charge
and discharge levels under all techniques are shown in Figure 3-2. RE generation
depends on installed capacity, ambient temperature and solar radiations. The prole
for ambient temperature and solar radiation is shown in Figures 3-3 and 3-4, respec-
tively, and obtained from Meteonorm 6.1 software (Oak Ridge, TN 37830, USA) for
the Islamabad region of Pakistan.
17
Chapter 3
8:00 12:00 16:00 20:00 0:00 4:00 8:00
Time (hours)
15
20
25
30
35
40
Temprature (°C)
Ambient temperature
Figure 3-3: Ambient temperature
.
8:00 12:00 16:00 20:00 0:00 4:00 8:00
Time (hours)
0
200
400
600
800
1000
1200
1400
Total irradiance (kW/m2)
Solar irradiance
Figure 3-4: Estimated solar radiations
.
3.1.1 Power Output of Renewable Energy Sources
Major RESs include solar and wind; however, PV is the least expensive source of
generation and requires one time investment. The planet we live on receives 174,000
terawatts (TW) of solar radiation [34]. The areas with insolation levels of 150–300
18
3.1.1. Power Output of Renewable Energy Sources Chapter 3
W/m
2
or 3.5–7.0 kWh/m
2
per day are the most populated areas in this world [35]. The
incoming solar energy (in the form of radiation) that reaches the surface of the Earth
is dened as insolation. Approximately 70% of the total radiations are absorbed, and
the rest is reected back to space. This is called the solar energy cascade, which
does not have a xed value around the dierent locations of the Earth, nor is it
constant over dierent periods of time. If we move to the north and south of the
Equator, the insolation shows a continuously varying trend, and its quantity keeps
decreasing towards the poles with respect to season. In March and September, the
isolations are at the highest level in the Northern Hemisphere, whilst the Southern
Hemisphere enjoys September and March [36]. Thus, the power output of RESs
depends on several parameters. Some of them are nature dependent, while others can
be adjusted accordingly. The former includes season (summer or winter), weather
(sunny or cloudy) and topographical constraints, whereas the latter comprises the
size and eciency of installed technologies. A block diagram of the power ow of
RESs in the proposed model is shown in Figure 3-5.
Power output from the PV unit can be measured as a function of solar irradiation
and ambient temperature. Both solar irradiance and temperature highly depend on
weather and season. A solar panel consists of several cells, which are coupled together
to produce power. In a similar work [37], the authors use renewable energy generation
and storage systems, where deterministic and stochastic methods have been used to
consider uncertainties.
The temperature of a cell can be given as [15]:
𝑇𝑐(𝑡) = 𝑇𝑎(𝑡) + 𝐼𝐺(𝑡)𝑁𝑂𝐶𝑇 −20
0.8
(3.1)
where
𝑇𝑐(𝑡)
belongs to the temperature of cell (
∘
C) at time
t
,
𝑇𝑎
denotes the current
temperature (
∘
C) at given location,
𝐼𝐺
is the global solar irradiance (kWh/m
2
), NOCT
is the nominal operating cell temperature (
∘
C), which can be dened as a level of
temperature reached under the following conditions: irradiance = 800 w/m
2
, air
temperature = 20
∘
C, wind velocity = 1 m/s and tilt angle of cell =
45∘
. Therefore,
19
Chapter 3
the output of a PV array can be measured as [38,39]:
𝑜𝑝𝑣
𝑡=𝑋𝑑
𝐼𝐺(𝑡)
𝐼𝑆1−𝐾𝑝
100(𝑇𝑐(𝑡)−𝑇𝑆 𝑇 𝐶 )
(3.2)
Figure 3-5: Power ow of renewable energy sources (RESs) in the proposed model.
where
𝑜𝑝𝑣
𝑡
is the output of the PV at time
t
,
𝑋𝑑
is the derating factor of the
PV array,
𝐼𝑆
is the standard irradiations (kWh/m
2
) and
𝐾𝑝
shows the temperature
coecient (
%/∘
C). The Meteonorm software is used to calculate the radiations on
a PV panel of tilt angle 30
∘
, which is considered as the optimal direction for PVs.
Generically, if the temperature remains within the limits of 16–36
∘
C throughout the
day and the area of the PV is 200–220 m
2
, then the output power on a typical sunny
day would be roughly 500 kW. Moreover, the partially sunny and cloudy days will
have 200–300 and 50–70 kW of generation, respectively, under similar conditions.
20
3.1.2. Load Categorization Chapter 3
The power output from a WT is modeled as a piecewise function of wind speed. We
denote
𝑣
as wind speed,
𝑣𝑟
as rated wind speed (i.e., where WT generates maximum
energy),
𝑣𝑐𝑖
the cut-in speed (i.e., minimum required speed to turn on WT) and
𝑣𝑐𝑜
the
cut-out speed (i.e., excessive speed, blades brought to test). Besides, the air density,
the size of WT obviously eects the total power output. A generic mathematical
expression is given below to nd output of WT [31]:
𝑜𝑤𝑡
𝑡=⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
0𝑖𝑓 𝑣 < 𝑣𝑐𝑖 𝑜𝑟 𝑣 > 𝑣𝑐𝑜
𝑣3−𝑣3
𝑐𝑖
𝑣3
𝑟−𝑣3
𝑐𝑖 𝑖𝑓 𝑣 > 𝑣𝑐𝑖 𝑎𝑛𝑑 𝑣 < 𝑣𝑟
1𝑖𝑓 𝑣 > 𝑣𝑟𝑎𝑛𝑑 𝑣 < 𝑣𝑐𝑜
(3.3)
The parametric values for
𝑣𝑐𝑖
,
𝑣𝑟
and
𝑣𝑐𝑜
are 3, 10 and 20, respectively. Users
usually do not follow a specic pattern to run daily appliances. Therefore, appliances
are categorized according to their electricity consumption pattern in the subsequent
section.
3.1.2 Load Categorization
In view of daily power consumption pattern, appliances are categorized into two
groups, i.e., exible loads and inexible loads, as described below.
1. Inexible appliances: This type of appliance is also referred to as xed or regular
appliances because of their constant power usage pattern and length of opera-
tion time. Typically, inexible loads include fridge, fan, light, etc., which are
considered to be required run loads and cannot be shifted to later hours. These
appliances usually do not participate in the DR, so they cannot contribute to
the optimization process in order to achieve lower electricity bill. Therefore,
regular loads execute their job on respective time slots and have no relation
with the appliance scheduler.
2. Flexible appliances: Flexible loads are also known as shiftable or burst loads.
Flexible appliances include the dish washer, washing machine, spin dryer, etc.
21
Chapter 3
The power consumption pattern of this type of appliance can be altered to later
hours in response to some incentives. Appliances are shifted to later hours due
to two main reasons: either appliances are preferred to alter the consumption
pattern from on-peak hours to o-peak hours or when the price for the grid is
high, appliances are shifted to low price hours for bill reduction.
3.1.3 Energy Consumption Model
Each consumer in the smart building has two sets of appliances, discussed earlier,
i.e.,
𝐹
and
𝐼
. The set of exible appliances
𝐹={𝑎1, 𝑎2, 𝑎3, . . . , 𝑎𝑓}
and the set
of inexible appliances
𝐼={𝑏1, 𝑏2, 𝑏3, . . . , 𝑏𝑖}
over a scheduling horizon of
𝑇=
{1,2,3,4,5,...,48}
. The hourly electricity demand of a single appliance is given
as:
𝐸𝑎,𝑡 =𝐸𝑎,𝑡1+𝐸𝑎,𝑡2+𝐸𝑎,𝑡3,...,+𝐸𝑎,𝑡48
(3.4)
where
𝐸𝑎, 𝑡1+𝐸𝑎, 𝑡2+𝐸𝑎, 𝑡3,...,+𝐸𝑎, 𝑡48
denote electricity demand of an appliance in
a respective time slot. The daily electricity consumption of both types of appliances
(exible and inexible) is given by Equations (3.5) and (3.6), respectively:
𝐸𝑎=
48
𝑡=1 𝐹
𝑓=1
𝐸𝑎
𝑡={𝐸𝑎
𝑡1+𝐸𝑎
𝑡2+. . . +𝐸𝑎
𝑡48} ∀ 𝑓𝜖𝐹
(3.5)
𝐸𝑏=
48
𝑡=1 𝐼
𝑖=1
𝐸𝑏
𝑡={𝐸𝑏
𝑡1+𝐸𝑏
𝑡2+. . . +𝐸𝑏
𝑡48} ∀ 𝑖𝜖𝐼
(3.6)
where
𝐸𝑎
𝑡1, 𝐸𝑎
𝑡2, . . . , 𝐸𝑎
𝑡48
denote the power consumption of exible appliances and
𝐸𝑏
𝑡1, 𝐸𝑏
𝑡2, . . . , 𝐸𝑏
𝑡48
represent the power consumption of inexible appliances at time
𝑡
. The total daily power consumption
𝐸𝑠𝑢𝑚
is given as:
𝐸𝑠𝑢𝑚 =
48
𝑡=1 𝐹
𝑓=1
𝐸𝑎
𝑡,𝑓𝜖𝐹 +
𝐼
𝑖=1
𝐸𝑏
𝑡,𝑖𝜖𝐼
(3.7)
22
3.1.4. Capacity Constraints Chapter 3
3.1.4 Capacity Constraints
The power outputs of considered resources must not exceed beyond the limits of
designed capacities:
𝑜𝑐
𝑡≤𝑃𝑐∀𝑡
(3.8)
𝑜𝑏
𝑡≤𝑃𝑏∀𝑡
(3.9)
𝑆𝑡𝑠
𝑡≤𝑃𝑡𝑠 ∀𝑡
(3.10)
𝑜𝑝𝑣
𝑡≤𝑃𝑝𝑣 ∀𝑡
(3.11)
𝑜𝑤𝑡
𝑡≤𝑃𝑤𝑡 ∀𝑡
(3.12)
𝑆𝑒
𝑡≤𝑃𝑒∀𝑡
(3.13)
where
𝑜𝑐
𝑡
,
𝑜𝑏
𝑡
,
𝑆𝑡𝑠
𝑡
,
𝑜𝑝𝑣
𝑡
,
𝑜𝑤𝑡
𝑡
and
𝑆𝑒
𝑡
are the outputs;
𝑃𝑐
,
𝑃𝑏
and
𝑃𝑡𝑠
are the capacities
of the CHP, boiler, thermal storage, PV, WT and electric storage respectively at
𝑡
.
3.1.5 Thermal Storage Constraints
The total heat in the storage at
𝑡
depends on heat stored at
𝑡−1
, heat charged and
heat discharged. Heat discharged is subject to being subtracted from the total heat
because it is the out going source and results in depletion of storage. Heat loss in the
process of charging and discharging, i.e., turn around eciency, is denoted by
𝜆𝑡𝑠
. In
order to store 10 kW in storage,
10 + 𝜆𝑡𝑠
kW must be provided, and in the case of
discharging 10 kW of power,
10 + 𝜆
kW must be discharged:
𝑆𝑡𝑠
𝑡=𝑆𝑡𝑠
𝑡−1+𝜆𝑡𝑠𝑐𝑡−𝑑𝑡/𝜆𝑡𝑠 ∀𝑡
(3.14)
where
𝑐𝑡
and
𝑑𝑡
are the charge and discharge rates of thermal storage at
𝑡
, respectively.
The charge and discharge rates must be within the designed limits of the charge and
discharge rates.
𝐶𝑡𝑠
and
𝐷𝑡𝑠
are the thermal storage charge and discharge limits:
23
Chapter 3
𝑐𝑡≤𝐶𝑡𝑠 ∀𝑡
(3.15)
𝑑𝑡≤𝐷𝑡𝑠 ∀𝑡
(3.16)
3.1.6 Electric Storage Constraints
The total electricity in the electric storage at
𝑡
depends on the electricity stored at
𝑡−1
, the electricity charged and discharged. Electricity discharged is subject to being
subtracted from the total electricity, because it is the out going source and, hence,
results in the depletion of storage. Electricity loss in the process of charging and
discharging, i.e., turn around eciency, is denoted by
𝜆𝑒
. The rest of the storage and
discharge process is the same as that for heat storage described above:
𝑆𝑒
𝑡=𝑆𝑒
𝑡−1+𝜆𝑒𝑔𝑡−ℎ𝑡/𝜆𝑒∀𝑡
(3.17)
where
𝑔𝑡
and
ℎ𝑡
are the charge and discharge rates of electrical storage at time
𝑡
,
respectively. The charge and discharge rates must be within the designed charge and
discharge limits.
𝐺𝑒
and
𝐻𝑒
represent electrical storage charge and discharge limits:
𝑔𝑡≤𝐺𝑒∀𝑡
(3.18)
ℎ𝑡≤𝐻𝑒∀𝑡
(3.19)
3.1.7 Energy Balance
The electricity demand is fullled by the local RESs, energy drawn from storage
minus the energy sent to electrical storage and direct connection of the grid, whereas
heat demand is met by the CHP generation, boiler units and heat retrieval from
storage minus heat saved in storage. The electricity consumption must not exceed
the electricity imported from main grid and the total generation of PV and WT:
24
3.1.8. Start and End Time Horizon Chapter 3
30
𝑖=1
12
𝑘=1
𝑇𝑖𝑘
𝜃=0
𝑃𝑐𝑐
𝑘𝜃 𝑆𝑖𝑘𝑡 =𝑜𝑝𝑣
𝑡+𝑜𝑤𝑡
𝑡+ℎ𝑡−𝑔𝑡+𝐸𝐼𝐺
𝑡∀𝑡
(3.20)
where
𝑇𝑖𝑘
is the processing time duration of appliance
𝑘
of home
𝑖
,
𝑃𝑐𝑐
𝑘𝜃
is the power
consumption capacity of appliance
𝑘
at time period
𝜃
,
𝑆𝑖𝑘𝑡
is a binary variable that
shows the status of appliance
𝑘
at
𝑡
,
𝑜𝑝𝑣
𝑡
is the power output from PV,
𝑜𝑤𝑡
𝑡
is the
power output from WT at
𝑡
,
ℎ𝑡
and
𝑔𝑡
are the electric storage discharge and charge
rates and
𝐸𝐼𝐺
𝑡
is the total energy imported from the main grid at
𝑡
. As discussed
earlier, the heat demand balance is shown in Equation (3.21), and
𝛿
is the heat to
power ratio of CHP:
𝐻𝑑
𝑡=𝛿𝑜𝑐
𝑡+𝑜𝑏
𝑡+𝑑𝑡−𝑐𝑡∀𝑡
(3.21)
3.1.8 Start and End Time Horizon
Each appliance has to complete its working hours within the given time frame; how-
ever, no appliance can start before the provided earliest starting time window, nor
can it complete later than nishing time window. Since each appliance has to n-
ish between the given time interval minus the operational time duration, a binary
variable
𝑆𝑖𝑘𝑡
is introduced that indicates whether an appliance has nidhed its job or
not.
𝑆𝑖𝑘𝑡 =⎧
⎨
⎩
1𝑖𝑓 𝑎𝑝𝑝𝑙𝑖𝑎𝑛𝑐𝑒 𝑖𝑠 𝑂𝑁
0𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
(3.22)
3.1.9 Power Demand
The maximum electricity demand from the main grid over a period of time is given
as:
𝐸𝑚𝑎𝑥 ≥𝐸𝐼𝐺
𝑡∀𝑡
(3.23)
𝐸𝑚𝑎𝑥
is the maximum power demand from the power station, and
𝐸𝐼𝐺
𝑡
is the
energy imported from the grid at
𝑡
.
25
Chapter 3
3.1.10 Peak to Average Ratio
The basic aim behind balancing the PAR is to maintain the equilibria of demand and
supply between utility and consumers. In our proposed system model, PAR is dened
as a ratio of peak load over average load in the given time frame and is symbolized
as
𝜓
. Mathematically, PAR can be written as:
𝜓=𝑚𝑎𝑥(𝐸𝑡)
1
𝑇48
𝑡=1 𝐸𝑡
(3.24)
3.1.11 Waiting Time
Inexible appliances are supposed to run with the highest priority and without any
delay, so these appliances do not have any concern with the waiting time. Flexible
tasks play a crucial role in the optimization by altering the power consumption be-
havior. Let
𝛼𝑎
and
𝛽𝑎
be the start and end times of exible appliance
𝑎
, such that
𝛼𝑎≤𝛽𝑎
within the given time window interval. In our model, we consider waiting
time as discomfort. The more the waiting time is, the lesser the comfort will be. We
denote
𝜉𝑎
as the working duration and
𝜎𝑎
as the actual start time of appliance
𝑎
.
𝜎𝑎
has a value no less than
𝛼𝑎
, but less than or equal to
𝛽𝑎−𝛼𝑎
given as:
𝜎𝑎𝜖[𝛼𝑎, 𝛽𝑎−𝜉𝑎]
(3.25)
3.1.12 Objective Function
The objective is to minimize the total power consumption cost of all appliances in
the smart residential building. The electricity cost depends on the pricing signal
announced by the utility company and the appliances’ power consumption pattern.
We have no control over the pricing signal; however, we minimize the cost by altering
the power consumption pattern of appliances:
𝑚𝑖𝑛
48
𝑡=1 30
𝑖=1
12
𝑘=1 (𝐸𝑠𝑢𝑚
𝑡−𝑜𝑝𝑣
𝑡−ℎ𝑡)×𝐸𝑝
𝑡+𝐶𝑝ℎ
𝑡
(3.26)
26
3.1.12. Objective Function Chapter 3
The objective function is subject to the constraints given in Equations (3.8)–
(3.16), (3.20) and (3.23), where
𝐸𝑠𝑢𝑚
𝑡
is the total power consumption at time
𝑡
,
𝐸𝑝
𝑡
is
the price of electricity at
𝑡
announced by the utility and
𝐶𝑝ℎ
𝑡
is the cost incurred in
order to satisfy heat demand.
27
Chapter 4
Heuristic Techniques
28
4.0.1. Teaching Learning-Based Optimization Chapter 4
Numerous heuristic, meta-heuristic and mathematical techniques have been used
for DSM in the residential sector. Reducing energy expense, reducing PAR, balancing
demand and supply, maximizing the user comfort level, stabilizing the grid and en-
suring the power quality are some of the objectives of using optimization techniques.
One of the key underpinnings of heuristic techniques is that they