Content uploaded by Nadeem Javaid
Author content
All content in this area was uploaded by Nadeem Javaid on Sep 16, 2019
Content may be subject to copyright.
Towards Energy Efficiency in Smart Buildings
Exploiting Dynamic Coordination among
Appliances and Homes
1
Presented by: Adia Khalid
PhD (Scholar)
Supervisor : Dr. Nadeem Javaid
Co-Supervisor : Dr. Manzoor Ilahi Tamimy
Agenda
Introduction
Related Work
Problem Statement
Problem formulation
Proposed Solution
System Model
Results
Conclusion and Future Directions
2
❑Electric power is one of the emergent need from
decades
❑30-40% energy is consumed in buildings throughout
the world [1]
❑According to International Energy Outlook 2013
❑Energy consumption will increase up to 56% in
2040 [2] [1]. Gul, M. S., & Patidar, S. (2015). Understanding the energy consumption and occupancy of a
multi-purpose academic building. Energy and Buildings, 87, 155-165.
[2]. Energy Information Administration "https://www.eia.gov/todayinenergy
/detail.cfm?id=12251"United States Department of Energy, Washington (last visited 17th December
2015).
[3]. Paul Chefurka, “http://www.paulchefurka.ca/EnergyGap.html”, (last visited 29th January 2017).
Introduction (1/3)
Energy demand increase graph [3]
Millions of Tones of Oil Equivalent (MTOE)
3
❑Increasing energy demand and historical crises
within traditional grid are overcome by:
❑Integrating two way communication
technology in traditional grid (Smart Grid)
❑Demand Side Management
❑Smart Grid
Includes a variety of operational and energy
measures such as:
❑Smart meters, smart appliances, renewable
energy resources and bidirectional
communication
Introduction (2/3)
4
Demand Side Management
❑Planning, employment and monitoring the utilities
strategies [4]
It is Planned :
❑To make changes in the consumer usage of electricity
❑Agendas of utilities under DSM are:
❑Load management using Demand Response
❑Time of Use (TOU), Real Time Price (RTP), Variable Peak
Price (VPP), Critical Peak Price (CPP), etc.
❑Encourage the customers to use less electricity during the
On-peak hours [5]
❑Generation on customer end using renewable resources [4] Demand Side Management (vol. 1, Overview of Key Issues), Final Rep. for RP2381-
4, prepared by Battelle-Columbus Division and Synergic Resources Corp., EA/EM-3597,
Electric Power Research institute, Palo Alto, CA,Aug. 1984.
[5] Gelazanskas, L., & Gamage, K. A. (2014). Demand side management in smart grid: A
review and proposals for future direction. Sustainable Cities and Society, 11, 22-30.
Introduction (3/3)
5
On-peak hours (where
electricity demand is high)
Off-peak hours (where demand
is low)
Related Work (1/2)
6
[6]. Elkazaz, M. H., Hoballah, A.,& Azmy, A. M. (2016). Artificial intelligent based optimization of automated
home energy management systems. International Transactions on Electrical Energy Systems.
[7]. Rasheed, M. B., Javaid, N., Ahmad, A., Khan, Z. A., Qasim, U., & Alrajeh, N. (2015). An Efficient Power
Scheduling Scheme for Residential Load Management in Smart Homes. Applied Sciences, 5(4), 1134-1163.
[8]. Adika, C. O., & Wang, L. (2014). Smart charging and appliance scheduling approaches to DSM.
International Journal of Electrical Power \& Energy Systems, 57,232-240.
Technique(s)
Objective(s)
Price
Tariff
Result
Limitation(s)
Genetic Algorithm
(GA)
[6]
❑
Cost minimization
❑Allow two way energy flow
(renewable resources)
RTP
32
.73% electricity
cost
reduction
Ignored
Peak Average
Ratio
(PAR)
and
distributed
generation's
maintenance
and
installation
cost
Wind Driven
Optimization (WDO)
by mapping problem
as min
-max regret-
based knapsack [7]
❑
Cost minimization
❑
User comfort
❑Appliance operational
waiting time minimization
using renewable resource
(solar panel)
❑
PAR reduction
TOU
8
.7% PAR and 39.04
%
electricity
cost reduction
Solar
panel maintenance
and
installation
cost are
not
considered
Linear programming
and first come first
serve policy [8]
❑
Electricity cost minimization
❑Using energy storage
❑
Regulate the peak demand
Day
-ahead RTP
Reduced
20%
electricity
cost
and 21.6% peak
load
Overlooked
waiting time
of
appliances
Related Work (2/2)
7
[9]. Azar, A. G., & Jacobsen, R. H. (2016). Appliance Scheduling Optimization for Demand Response.
International Journal on Advances in Intelligent Systems, 9(1 & 2).
[10]. Zhang, Y., Zeng, P., Li, S., Zang, C., &Li, H. (2015). A novel multiobjective optimization algorithm for
home energy management system in smart grid. Mathematical Problems in Engineering, 2015.
Technique(s)
Objective(s)
Price Tariff
Result
Limitation(s)
❑
Single objective
❑Evolutionary
Algorithm (EA)
❑
Multi objective
optimization
❑
Dynamic programing
(mapping problem as
knapsack) [9]
❑
Conflicting objectives are:
❑Minimizing electricity cost and
maximization of total number of
allowed power demand requests
❑Minimization of Carbon dioxide
emission and maximization of total
number of allowed power demand
requests
❑
Rescheduling
Danish
electricity
prices
Peak demand by
40% and 1% cost
Low
cost reduction
Particle Swarm Optimization
(PSO)
Mixed
-Integer Linear
Programming (MILP) [10]
❑
Cost minimization (single objective)
❑
Conflicting objectives are:
❑Minimizing electricity cost
and maximization user comfort
level (based on indoor outdoor
temperature)
❑
Solar panel implementation
RTP
Reduced
32.76
%
cost
Maximize
32.36
%
user
comfort
for
different
scenario
Ignored
PAR
and
solar panel
maintenance
cost
Problem Statement
8
DSM has planned and developed strategies like load shifting using DR.In this respect, an efficient HEMS is required which helps user
to shift the load from On-peak to Off-peak hours in a balanced way. Though many systems like [11]-[14] have been developed with
targets: minimization of cost, PAR and discomfort where trade off occurs between cost and PAR , and cost and comfort. These trade-offs
and are not fully considered by all researchers, as authors in [12] did not consider user comfort, PAR is ignored by [11], and [12]
overlooked the operational waiting time of appliances. Though, [9], [11] and [14] have been developed to overcome conflicting
objectives with trade-off. Article [11], targeted electricity cost minimization and utility stability maximization while ignoring waiting
time of appliances. Where, [14] considered electricity cost and user discomfort reduction, however, totally ignored peak formation
during Off-peak hours, which can burden the utility and generation unit by creating the peak. To handle such situation some other
strategies have been proposed as, [13] has incorporated Renewable Energy Resources (RESs) to reduce cost and PAR. These resources
require the installation and maintenance cost. However, [13] did not consider installation and maintenance cost. Also, fluctuation in the
generation of RES is possible because of varying weather conditions seasonally as cited by [15].
Beside these, user comfort as system flexibility is always a challenging task. Such that scheduler's suppleness can deal with sudden
changes without effecting demanded load and overall cost. In this respect, [11] and [9] work relates with system flexibility. System
flexibility is concern with dealing user run-time demand, where system can be switched OFF and ON according to the user demand, and
perform rescheduling. [13] and [9] performed the rescheduling while ignoring the peak formation during Off-peak hours that can
increase the PAR.
[11] Jovanovic, R., Bousselham, A., & Bayram, I. S. (2016). Residential Demand Response Scheduling with Consideration of Consumer Preferences. Applied Sciences, 6(1), 1-16.
[12] Vardakas, J. S., Zorba, N., & Verikoukis, C. V. (2016). Power demand control scenarios for SG applications with finite number of appliances. Applied Energy, 162, 83-98.
[13] Gao, B., Liu, X., Zhang, W., & Tang, Y. (2015). Autonomous Household Energy Management Based on a Double Cooperative Game Approach in the Smart Grid. Energies, 8(7), 7326-7343.
[14] Cortes-Arcos, T., Bernal-Agustin, J. L., Dufo-Lopez, R., Lujano-Rojas, J. M., & Contreras, J. (2017). Multi-objective demand response to real-time prices (RTP) using a task scheduling
methodology. Energy.
[15] Rabiee, A., Sadeghi, M., Aghaeic, J., & Heidari, A. (2016). Optimal operation of microgrids through simultaneous scheduling of electrical vehicles and responsive loads considering wind and
PV units uncertainties. Renewable and Sustainable Energy Reviews, 57, 721-739.
9
Load shifting
Bring consumption schedule load curve close to objective load
curve [16]
= minimize (
-
) ………… (1)
………… (2)
=
.………… (3)
where
Problem Formulation (1/5)
[16]. Logenthiran, T., Srinivasan, D., & Shun, T. Z. (2012). Demand side management in smart grid using
heuristic optimization. IEEE Transactions on Smart Grid, 3(3), 1244-1252.
=
.………… (4)
10
Constraints:
.………… (5)
=
, .………… (5a)
= sum(
) - std
), ………… (5b)
= std
) + η×min
), .………… (5c)
= mean
), .………… (5d)
− min
)
− min
), .………… (5e) h={1, 2, 3, . . . ., 24}
Problem Formulation (2/5)
Load (kwh)
Cost (¢)
Off- Peak Feasible
Region
search space
search space
On- Peak Feasible
Region
search space
search space
11
Cost minimization
= minimize (
) .………… (6)
=
) ………… (7)
Constraints:
Equation 5.
PAR reduction
= minimize(PAR)
= (
………… (8)
where N = {1,2,3,……….,24}
Problem Formulation (3/5)
➔Objective load curve, minimum cost and PAR reduction
all three objectives are achieved through our defined fitness
function given in equation 5
➔Fitness function defined on the bases of constraint i.e.,
limitation of maximum and minimum load based on price rate
12
Maximize the user comfort
Rescheduling the appliances on user demand through
coordination
maximize (Comfort) ………….… (9)
…….......… (10)
n={1 ,2 3,……..9}
=|
| ………… (11)
where
scheduled time by the scheduler
Available Space
App 1 App 4 App 5
App 1
Given Schedule at Hour 1
App 1 Interrupted
Empty Slot is Allocated to App 2
App 3 App 6 App 7 App 8 App 9
Emergency Call for Appliance
Coordination
App
7
App 4 App 5 App
7
App 3 App 6 App 7 App 8 App 9
App 7 Interrupted
Empty Slot is Allocated to App 2 and App 4
Coordination
Given Schedule at Hour 4
One Hour Interval
App 4 App 5 App
7
App 2
App 3 App 6 App 7 App 8 App 9
App 2 App 4
One Hour Complete Process One Hour Complete Process
App 3
App 3
App 3
App 1
Problem Formulation (4/5)
❑The proposed multi objective optimization model is transformed into
single objective optimization as presented in [10].
❑Adapt weighted sum method
λ(minimize () (1-λ) Comfort ………… (12)
where λ= 1 for scheduling and λ= 0 for rescheduling
❑Adapt Constraint method
= minimize (
-
),
subject to and
13
Model Transformation Multi-Objective
Optimization
Problem
Minimize f1
Minimize f2
.
.
.
Minimize fn
With respect to constraints
High Level
Information
Estimation of
relative importance
Vector
(w1, w2, w3, …… wn)
Single Objective
Optimizer
One Objective Solution
http://wiki.ece.cmu.edu/ddl/index.ph
p/Multi-Objective_Optimization
Home energy management system
❑Day-ahead scheduling
❑Schedule the home appliance before the start of day
❑Schedule the home appliances by load shifting
❑On-peak hours load on Off-peak hour
❑Target the objective load curve
❑Real time scheduling
❑Reschedule appliance(s) on user demand during the day
❑Coordination among appliances
❑Reduced the waiting time of user priority appliances
Proposed Solution
14
System Model (1/2)
15
Day-ahead scheduling
Meta-heuristic algorithm (nature-inspired)
Hybrid Bacterial Foraging Algorithm and Genetic
Algorithm (HBG)
Why nature-inspired [17].
Simple: most of these algorithms are simple to
implement
Flexible: can deal with a broad range of
problem
Ergodicity (positive recurrent aperiodic state
of stochastic systems):
they can easily escape from local minima [17]. Yang, X. S., Cui, Z., Xiao, R., Gandomi, A. H., & Karamanoglu,
M. (Eds.). (2013). Swarm intelligence and bio-inspired computation:
theory and applications. Newnes.
System Model (2/2)
16
Real time scheduling
Mapping the rescheduling problem as
knapsack
Mathematical or conventional techniques
divide the problem into subproblem (helpful
for small problems)
Dynamic Programing
Its ability to divide the problem into smaller
problems
Memorizing
17
GA
Techniques for Home Energy Management System (1/4)
❑GA meta heuristic Evolutionary based
algorithm
❑GA Consists of set of chromosomes
(possible solutions)
❑A single Chromosome is consists of
gene (appliance)
❑After a maximum generation
❑Best solution is selected on the bases of
objective function
Initialize the Population
Evalution
Parent Selection
Mutation
Cross Over
No
Update
Population
Stopping
Criteria
Stop
Yes
18
Bacterial Foraging Algorithm (BFA)
❑Based on foraging behavior of E. coli bacteria
❑Each bacteria (possible solution) has bacterium (appliance)
❑Main steps of BFA are:
Elimination and Dispersal
Reproduction
Chemotaxis
Swarming
Bacteria position will be updated using equation
=
…………. (13)
Population will be evaluated using equation
=
)+ - 1) ) ……. (14)
Optimal solution at the end will be selected based on equation 12.
Techniques for Home Energy Management System (2/4)
Start
Initialize Population
Fitness Evolution (Jlast)
Reproduction Step
Chemotaxis Step
Tumble Swim
Fitness Evaluation (Ji)
Every Bacterium
Replace Jlast with Ji
Evaluate for Selection
k<Nr
l<Ne
Select the Best one
Randomly Eliminate and
Disperse old population
Fitness Evolution
End
No
Yes
Yes
No
Yes
No
Yes
No
Tumble
Elimination & Dispersal Step
No
Fitness Evolution (Ji)
Yes
Yes
j<Nc
i<Np
Ji<Jlast
Swim
Algorithm 1:
HBG for scheduling in SG.
Initialization (PoP) and set parameters: N
p= 30, Ne=24, Nr=Nc=5, Ns
=2, C=0.01
Evaluate the initial PoP using equation 15 J
last→ J
for l = 1 → N
e
for k = 1 → Nr
for j = 1 → Nc
for i= 1 → Np
Tumble bacteria, update position θiand compute the Fitness Ji
for s = 1 → Ns
if Ji < Jlast then
Jlast → Ji
Swim bacteria
else
Tumble bacteria and move in that direction
end if
end for
end for
end for
Calculate Ffand select the best one using equation 4
end for
Elimination-dispersal step using GA Sch(l) ←best
end for
Techniques for Home Energy Management System (3/4)
HBG
❑BFA
❑Exploit the population (random
movement)
❑GA
❑Explore the population (crossover)
❑Premature conversion
❑HBG
❑Solution will lie in local as well as
global optimum
❑Chemotaxis:
Swarming:
Reproduction:
Elimination and Dispersal Steps
Using GA
❑Formulate our problem as knapsack problem
❑Objective of knapsack
❑Fill the available capacity
❑Total weight of selected items should
be less than or equal to available
knapsack capacity
❑Knapsack weight capacity is (available
time interval)
❑Weight of an item (available working time
slot of appliance)
❑Value of an item (cost)
20
Real Time scheduling (Coordination)
Techniques for Home Energy Management System (4/4)
Algorithm 2: Real time coordination among appliances :
Dynamic programing.
Input: (Scheduled per hour (Sch) and Scheduled of M
appliances
for Hour = 1 → 24
if Ĭ == yes (Ĭ is interrupt generated by user during real time)
Ask for appliance user wand to off (Appoff)
Time available = 60-Interptedtime,
Maintain Scheduled hour (, working time interval (
) and value
end for
end if
← , ← and n= length()
Create two tables T ← 0 and S ← 0 of size +1
for ỉ = 2 → n+1
for ĵ = 2 → +1
if
(ỉ -1) < ĵ
T[ỉ, ĵ] T[ỉ − 1, ĵ], Value(ỉ−1) + T[ỉ − 1, ĵ −
(ỉ − 1)]]
else
T[ỉ, ĵ] ← T[ỉ-1, ĵ]
end if
if T[ỉ, ĵ] == Value(ỉ−1) + T[ỉ − 1, ĵ −
(ỉ − 1)]
S[ỉ, ĵ] ← 1
else S[ỉ, ĵ] ← 0
end if
end for
end for
end if
end for
❑Scenario
❑Single home with nine appliances and three groups:
❑Interruptible burst load appliances
❑vacuum cleaner, water heater, water pump and dish washer
❑Base load
❑refrigerator, air conditioner and oven
❑Non-interruptible burst load
❑washing machine and cloth dryer
❑The cloth dryer will always be scheduled after
washing machine.
❑However, user will be allowed to turn OFF any
appliance during its operation
21
Group Appliances
Power Rating
(kWh)
Operational
time (hour)
Interruptible
burst load
Vacuum cleaner 0.7 ≤ 6
Water heater 5 12
Water pump 1 8
Dish washer 1.8 ≤ 8
Interruptible
base load
Refrigerator 0.225 18
AC 1.5 15
Oven 2.15 10
Non-
interruptible
load
Washing machine
0.7 5
Cloth dryer 5 4
Simulation
❑Priority Appliances rescheduled on user demand
❑Dish washer and vacuum cleaner
22
Results and Discussion (1/17)
Results includes
❑Before Coordination (Day-ahead Scheduling)
❑After Coordination (Run-time scheduling)
❑For three price tariff: RTP[15], TOU[17] and CPP[17]
Performance measuring parameters
Check the per hour load and cost
Over all cost and PAR
Waiting time (coordination)
Proposed techniques confidence interval for 95%
0
5
10
15
20
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
TOU
0
5
10
15
20
25
30
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
RTP
0
20
40
60
80
100
120
140
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
CPP
[17] Waterloo North Hydro, https://www.wnhydro.com/en/your-
home/time-of-use-rates.asp.
23
Results and Discussion (2/17)
RTP Tariff
Per hour electricity load before coordination Per hour electricity load after coordination
24
Results and Discussion (3/17)
RTP Tariff
Per hour electricity cost before coordination Per hour electricity cost after coordination
25
Results and Discussion (4/8)
RTP Tariff
Per day electricity cost before coordination Per day electricity cost after coordination
Before
Coordination
After
Coordination
GA
BFA
HBG
GA
BFA
HBG
14
.70 % dec
16
.31% dec
14
.22% dec
5
% dec
8
% dec
8
% dec
26
Results and Discussion (5/17)
Per day PAR before coordination Per day PAR after coordination
RTP Tariff
Before
Coordination
After
Coordination
GA
BFA
HBG
GA
BFA
HBG
51 % dec
47
% dec
58
% dec
2
% inc
4
%
inc
9
% inc
27
Results and Discussion (6/17)
RTP Tariff
Group wise average waiting time before coordination Group wise average waiting time after coordination
28
Results and Discussion (7/17)
TOU Tariff
Per hour electricity load before coordination Per hour electricity load after coordination
29
Results and Discussion (8/17)
Per hour electricity cost before coordination Per hour electricity cost after coordination
TOU Tariff
30
Results and Discussion (9/17)
TOU Tariff
Per day electricity cost before coordination Per day electricity cost after coordination
Before
Coordination
After
Coordination
GA
BFA
HBG
GA
BFA
HBG
16% dec
17
% dec
17
% dec
4
%
dec
5
% dec
5
% dec
31
Results and Discussion (10/17)
TOU
Per day PAR before coordination Per day PAR after coordination
TOU Tariff
Before
Coordination
After
Coordination
GA
BFA
HBG
GA
BFA
HBG
48 % dec
34
% dec
51
%
dec
6
% inc
14
% inc
17
% inc
32
Results and Discussion (11/17)
TOU Tariff
Average waiting time before coordination Average waiting time after coordination
33
Results and Discussion (12/17)
CPPTariff
Per hour electricity load before coordination Per hour electricity load after coordination
34
Results and Discussion (13/17)
CPPTariff
Per hour electricity cost before coordination Per hour electricity cost after coordination
35
Results and Discussion (14/17)
CPPTariff
Per day electricity cost before coordination Per day electricity cost after coordination
Before
Coordination
After
Coordination
GA
BFA
HBG
GA
BFA
HBG
37% dec
42
%
dec
42
% dec
8
% dec
7
%
dec
9
% dec
36
Results and Discussion (15/17)
CPP
Per day PAR before coordination Per day PAR after coordination
Before
Coordination
After
Coordination
GA
BFA
HBG
GA
BFA
HBG
59% dec
40
% dec
44
% dec
16
% inc
13
%
inc
14
%
inc
37
Results and Discussion (16/17)
CPPTariff
Group wise average waiting time before coordination Group wise average waiting time after coordination
38
Results and Discussion (17/17)
❑Over all results shows the outstanding performance
of HBG
❑Also HBG shows low convergence rate as compare
to other techniques
❑Results of with and without coordination shows that
interrupt by the user did not have much effect on
the PAR
❑On the other hand it reduced the cost because of
load curtailment
Tariff
GA BFA HBG
Cost
Lower
Upper
Lower
Upper
Lower
Upper
RTP 1536 1653 1528 1601 1575 1632
TOU 1498 1435 1462 1547 1548 1518
CPP 3600 5369 3643 4656 3789 4525
PAR
RTP 1.75 2.58 1.73 2.99 1.79 1.98
TOU 2.27 3.43 2.88 4.34 2.2 3.14
CPP 1.82 2 1.98 3.65 2.59 2.59
Confidence Interval
39
Conclusion and Future Work
❑PAR and cost are reduced through distributing the load
❑While defining limits using our defined fitness function to achieve objective load curve
❑Price signals do not have much affect on the performance of scheduler
❑Further, it is observed that real-time rescheduling procedure helps to handle interrupts
❑Coordination is established for appliance(s) rescheduling
❑In this scenario we transform multi objective problem into single objective
❑In future we will use dominated solution set to solve the conflict(trade-off)
❑This problem is studied for single home
❑However, in future we will check the performance of proposed solution for multiple
