A Smart Home Energy Management Strategy Based on Demand Side Management

Abstract

In this paper we propose an ECG optimization model for a smart home based on DSM.The proposed model is an efficient SHEM strategy. The model is proposed keeping in view the minimization of energy consumption,energy consumption cost and energy generation cost.The model is based on efficient scheduling of appliances and an ECG optimization algorithm is proposed.We are using and optimizing energy from two energy sources namely lceg and lcd which are also known as macrogrid and microgrid respectively.The problem is solved as cost optimization problem using genetic algorithm and mathematically formulated using binary MNKP. The simulation results show that our ECG model efficiently reduces the cost of energy consumption, energy generation and energy consumption utilization.

Full text

A Smart Home Energy Management Strategy
based on Demand Side Management
Zafar Iqbal1, Nadeem Javaid2,∗, Mobushir Riaz Khan1,
Farman Ali Khan3, Zahoor Ali Khan4,5, Umar Qasim6
1Pir Mehr Ali Shah Arid Agriculture University, Rawalpind 46000, Pakistan
2COMSATS Institute of Information Technology, Islamabad 44000, Pakistan
3COMSATS Institute of Information Technology, Attock 43600, Pakistan
4Faculty of Engineering, Dalhousie University, Halifax, NS, B3J 4R2 Canada
5Higher Colleges of Technology, Fujairah 4114, United Arab Emirates
6University of Alberta Edmonton, Alberta T6G 2J8, Canada
∗Corresponding author: nadeemjavaidqau@gmail.com, www.njavaid.com
Abstract—In this paper we propose an ECG optimization
model for a smart home based on DSM.The proposed model
is an efficient SHEM strategy. The model is proposed keeping
in view the minimization of energy consumption,energy con-
sumption cost and energy generation cost.The model is based
on efficient scheduling of appliances and an ECG optimization
algorithm is proposed.We are using and optimizing energy
from two energy sources namely lceg and lcd which are also
known as macrogrid and microgrid respectively.The problem is
solved as cost optimization problem using genetic algorithm and
mathematically formulated using binary MNKP. The simulation
results show that our ECG model efficiently reduces the cost of
energy consumption, energy generation and energy consumption
utilization.
Keywords:-Demand Side M anagement,
Microgrid,Distributed Energy Resources,
Smartgrid,Appliance scheduling.
I. INTRODUCTION
Demand side management and demand response program
are the area of smart grid research which encourage the utility
consumer to shift their high peak load to to low peak hours
to reduce their electricity bill,avail user comfort and ensure
balanced use of energy.In recent days the smart grid and
microgrid is replacing the traditional electricity grid due to the
increasing demand of the energy for industrial and residential
area,this increasing demand of energy leads to more genera-
tion of energy and exploring new ways to fulfill increasing
energy demand which leads to more air pollution and carbon
emission.Smart grid means traditional electricity grid plus IT
which can also be named as macrogrid while microgrid is
a local energy generation set up for homes,residential area
and buildings on a small scale which is related to distributed
energy resources(DER) e.g. local diesel or gas generator,solar
power,wind power,vehicle-to-grid,waste-to-energy etc.
In this work we proposed an optimization model known as
ECG.The model aim to minimize energy consumption,balance
use of energy,energy consumption cost and energy generation
cost.We have proposed a set of appliances S(sh)in a smart
home which in our model is using energy from two main
sources that is lceg and lcd which is further divided into
load categories as discussed,and each one is mathematically
formulated for one time slot and 24 hour time horizon.The
subcategories of load is not given here in this work due to
space limitation.The energy generation model is using local
generator and solar energy to generate power for running
regular load in a smart home.
A similar work is done by [1].In which the authors proposes
a distributed optimization framework for demand response
based on cost optimization and each user in this model
is receiving information in response to varying electricity
prices.For scheduling of appliances they use approximate
greedy iterative algorithm to minimize the cost.In [2]the
authors proposes a new demand side management technique
and a new energy efficient scheduling algorithm to operate the
appliances in such a way to reduce monetary expense of the
consumers based on varying price model taking into account
the uncertainties in appliances operation time and renewable
energy generation. [3] Proposes a communication based load
scheduling protocol for the smart home appliances.They use
two types of appliances one which have delay flexibility
and second which consume power as they desire.They use
energy management controller to decide variation in electricity
prices and formulate the optimization problem as cost saving
problem.
The rest of the paper is organized as follows.In section two
we discuss the related work,and section three present a com-
prehensive system model consisting of energy consumption
model,energy consumption cost model and energy generation
model,while section four discusses the mathematical model of
the problem including problem discussion,objective function
and proposed algorithm.Section five are conclusion and future
work and finally references in section sex.

II. RELATED WORK
In this section we discuss the similar work done by re-
search community on appliances scheduling and optimization
of cost and energy consumption.In [4] the author consider
the minimum electricity cost scheduling and optimization
problem for smart home appliances and solve the problem
using mixed integer linear programing.They consider several
realistic scenarios based on real time price and optimal power
profile signal to minimize cost.
In [5] the authors discusses consumer demand response
signal so that consumers can control their electricity price
and carbon footprints.The consumers use this information as
an incentive to minimize peak load demand by balancing
consumption of load.In work [6] a new user aware demand
side management technique is proposed for handling resi-
dential load while keeping in view the appliance waiting
time.Knapsack technique is used to keep trade off between
energy consumption and appliance waiting time while for
electricity cost and peak load demand reduction K-Wind
Driven Optimization technique is used.
In [7] the authors propose a user comfort aware effi-
cient residential management strategy and considering the
minimization of user frustration model,peak-to-average ratio
and electricity cost and solve the optimization problem by
proposing EREM algorithm.
In [8]the authors proposes a residential demand response
algorithm for the scheduled plan of home appliances.They
used mixed integer linear programming to minimize electricity
cost and peak load demand based on time-varying electricity
price.On the other hand they using solar system and storage
to make the schedule plan more effective.In [9] the authors
presents an optimal residential load commitment framework
to achieve cost saving and PAR reduction.
III. SYSTEM MODEL
Distributed energy resources-DER are power generation
systems or power storage technologies on a local and small
scale consisting of energy generation and energy storage like
co-generation, solar energy generation, wind power genera-
tion,local power generation via diesel or gas generator, hydro
power generation, waste-to-energy generation, vehicle-to-grid
generation, PV storage, flywheels, etc, which is also called
microgrid. The microgrid is a combination of local sources
of energy generation. The microgrids are modern, localized,
decentralized and small scale power generation units (having
capacity of 10 MW or less) contrary to large scale, centralized
and traditional electric grid called macrogrid. Microgrids are
typically low voltage source and generate AC power, and work
independently from electric grid and are usually installed by
the community or homes which serve them locally.In our
system model we take load from two main sources i.e. from
electric grid and microgrid which are known as lceg and lcd
load and divided these two main types of load into further
categories as mentioned below.Our microgrid include local
gas generator and solar energy,the system model consists of
three sub model 1. Energy consumption model 2. Energy con-
sumption cost model 3. Energy generation model to minimize
electricity cost,energy consumption and minimization of DER
cost.
A. Energy Consumption Model.
In this section the energy consumption is discussed.The
energy consumption from energy sources are categorized into
two main types i.e. lceg and lcd which are further divided
into sub categories, and the appliances load is also cate-
gorized according to these types of loads.The total load of
appliances is running on lceg source except regular or base
load which is using lcd for energy consumption.This type of
load includes home lighting,television,laptop,mobile charging
etc.These types of load usually consume very less wattage and
are must run load and users can not compromise on comfort
here,such load are not scheduled and can not be included
in demand side management otherwise user comfort will be
severely compromise and not enough saving in cost.Energy
consumption of such type of load from DER can be calculated
as in equation(1).
E(totalec,rl,α)=
24
X
t=1
nrl
X
α=1
e(α,rl,t)−P(pvg)(t)=0 (1)
where
E(totalec,rl,α)is total energy consumption by appliances αfor
base load from microgrid,αis appliance,totalec is total energy
consumed from microgrid,rl is base load,nrl is total no of
appliances consuming base load,P(pvg)(t)is power generated
by local generator and PV.We are assuming a smart home
with a set of αappliances denoted by S(sh).The mathematical
formulation of energy consumption of each type of load
(appliances) is given in following equations.In our scenario
t is a time slot which is one hour and belong to T which
is a total time horizon or total schedulable time which is 24
hours i.e. T=24 ,each appliance is denoted by αwe assume
that appliance αconsume an energy Eα(t) in time slot t and
t∈T.The energy consume by each appliance αand total
energy consume by all appliances i.e. of a set of all appliances
Ssh, means the overall energy consumption of smart home
is formulated in equation (2) and equation (3)respectively as
below.
Etotalec(S,α)=
24
X
t=1
n
X
α=1
eα(t)(2)
E(totalec,t,S(sh))=
n
X
α=1
nleg
X
lega=1
mder
X
ldera=1
e(legα,lderα,Ssh,t)(3)
∀t∈T, α ∈Ssh,lega ∈Segα,lder ∈Ssderα ∀Segα ∈
Ssh, lder ∈Sder,α,i.e.Ssh = (Segα )S(Sder,α)
Where
Etotalec(S,α)is total energy consumption by appliance α
belong to set S, eα(t)is energy of appliance αin time slot t,
totalec is total energy consumption, Ssh is set of appliances
in a smart home, nleg is no of appliances consuming lceg

load, mder is no of appliances consuming lcd load, legα
is an appliance consuming lceg load, lderα is an appliance
consuming Microgrid load,As we already assumed that S is a
set of appliances αin a smart home and αi∈S i.e.
S={α1, α2, α3, . . . , αn}
The set S comprises of all the appliances belong to different
categories of appliances which are categorized based on
load types.The energy consumption of each appliance α
for one time slot and for 24 hour time horizon as in
equation(2) and equation(3) respectively. The load types
defined are of two categories,load type which are connected
to electric grid and load type which are connected to DER’s-
Distributed Energy Resources,in our case the DER include
PV(Photovoltaic) and local diesel generator,which is also
called microgrid,thus second category of load is connected to
microgrid source.Hence we have two main sources of load
for our smart home appliances of set Ssh i.e. lceg and lcd.The
first category of load lceg-Load connected to electric grid is
further divided into the following load types.1. Deferrable
load-dl:Pool pump,dryer,water heater,washing machine,Air
conditioner,kettle,hob,washer,dishwasher,heating,EV battery
charging,batch printing,photocopying 2. Non deferrable
load-ndl:lights 3. Non deferrable and non interruptible
load-ndnil:refrigerator,lightening and servers 4. Deferrable
and interruptible load-dil:washer,dryer,dishwasher,batch
printing,photocopying 5. Deferrable and non interruptible
load-dnil: 6. Non deferrable and interruptible load-ndil:The
second category of load lcd-load connected to DER is
further divided into the following two types i.e.7. Regular
load-rl:lightening,heating etc. 8. Elastic load-el:washing
machine,dishwasher,
Load of both categories is scheduled except regular load or
base load. Each type of load energy consumption and energy
consumption cost is calculated, energy consumption of the
load connected to electric grid in one hour or one time slot t
and in total time horizon i.e. 24 hours is given in equation(4)
and equation(5) respectively.
EL(lceg ,t,ec)=
24
X
t=1
nlceg
X
α=1
EG(t,lceg )(4)
EL(lceg ,ec,t)≤EL(lceg,ec,t)(4a)
EL(lceg ,ec,t)< EL(lceg,ec,t)< Q(t)(4b)
where
lceg means load connected to electric grid, nlceg is no of
appliances consuming lceg load,t is time slot such that t ∈
T, EG is electric grid,lceg includes {dl,ndl,ndnil, dil,dnil,ndil}
types of load, ec means energy consumption,Q(t)is the total
quantity of energy in time slot t.The energy consumption of
second category of load lcd for one time slot and for 24 hour
time horizon is given in equation (5) and equation (6) as
follow.
EL(lcd,t,ec)=
n
X
α=1
eα(lcd,α,t)(5)
EL(lcd,t,ec)=
24
X
t=1
nlcd
X
lcdα=1
eα(lcd,α,t)(6)
∀t∈T, α ∈Sα
where
nlcd is no of appliances consuming lcd load, lcdα is an
appliance consuming lcd load,lcd means load connected to
microgrid. In the above section the total energy consumption
load is calculated,energy consumption of grid-connected and
DER-connected and for each appliance for one time slot and
for 24 hour time slot is separately calculated and mathemati-
cally formulated.
B. Energy Consumption Cost Model
In this section we discuss the energy consumption cost cal-
culation and formulation.The energy consumption cost model
of the set Sαof appliances αis as follow.Now we have to
calculate energy consumption cost for lceg load, lcd load
and its sub types deferrable load,non deferrable load,non de-
ferrable and non interruptible load,deferrable and interruptible
load,deferrable and non interruptible load,deferrable and non
interruptible load,non deferrable and interruptible load,regular
load,elastic load.The energy consumption cost of each type
of load is calculated by using the following equations.First
we calculate the total energy consumption cost of overall
load including grid-connected and DER-connected is given in
equation (7) and equation (8) below.
E(totalecc,Ssh,t,α)=
24
X
t=1
mlceg
X
lcegα=1
olcder
X
lcderα=1
e(totalecc,Ssh,lceg α,lcderα,t)EP (t)(7)
∀t∈T, α ∈SA
Total daily energy consumption cost is calculated by consid-
ering the energy generation cost as well.So,daily energy con-
sumption cost is calculated by using the following equations.
E(totalecc,S(sh),t)=
24
X
t=1
E(lceg,ecc,t)
+E(lcd,ecc,t)+P(pvg)(t)(8)
∀t∈T, α ∈SA
where, E(totalecc,S(sh),t)is the total daily energy consumption
cost of the user including energy generation cost and energy
consumption cost,E(lceg,ecc,t)is the energy consumption cost
of lceg power source, E(lcd,ecc,t)is the energy consumption
cost of lcd power source, P(pvg)(t)is the energy consumption
cost of power generated by local generator, totalecc is total
energy consumption cost, lceg is load connected to electric
grid, lcd is load connected to der, pvg is der sources i.e. solar
and generator energy, mlceg is no of appliances consuming
lceg load, olcder is no of appliances connected to microgrid,
lcegα is an appliance consuming lceg load, lcderαis an

appliance consuming lcd load.
The above energy consumption cost model include the total
consumption cost of a set S of appliances αin a smart home
also the set can be denoted by Ssh i.e. a set of all appliances
in a smart home.As the appliances load is divided into two
categories of which cost of each one is separately calculated
and of each load types inside these two categories.
Now the energy consumption cost of load connected to electric
grid for each appliance αfor one time slot and for 24 hour
time slot is given in equation (9) and equation (10) below.
E(lceg,ecc,t,α)=
nlceg
X
α=1
e(α,lceg,t)E P (t)(9)
E(lceg,ecc,t,α)=
24
X
t=1
nlceg
X
α=1
e(α,lceg,ecc,t)E P (t)(10)
∀t∈T, α ∈S
where
lceg is load connected to electric grid, nlceg is no of ap-
pliances consuming lceg load, ecc is energy consumption
cost,where lceg includes {dl,ndl,ndnil,dil,dnil,ndil} types of
loads.The energy consumption cost of lcd-connected load for
one time slot and for 24 hour time slot is given in equation
(11) and equation(12) as below.
E(lcd,ecc,t,α)=
n
X
α=1
e(α,lcd,ecc,t)EP (t)(11)
E(lcd,ecc,t,α)=
24
X
t=1
nlcd
X
lcdα=1
e(α,lcd,ecc,t)EP (t)(12)
C. Energy Generation Model.
As our second category of appliances are running on der
or microgrid which include in our case, photovoltaic(PV)
generation and local diesel generator generation and the der-
connected load is using it.Therefore the energy generation
model is formulated as total expense minimization cost of der
as given in equation (13) below, the total expense include gen-
eration fuel cost GFEp(g),cost due to losses C(lo)negligible in
this case ,fixed cost F(ct)i.e. startup cost ST(ct),depreciation
cost DP(ct),maintenance cost M(ct).
Minimize
n
X
g=1
λ(g)P2g
+GFEp(g)Pg+F(ct)+M(ct)(13)
subject to
P(g,mn)≤Pg≤P(g,mx)(13a)
24
X
t=1
n
X
g=1
Pg(t) = E(t,der)(t)(13b)
n
X
g=1
Pg(t) = E(t,der)(t)(13c)
where t is number of time slots, g number of generators, λ(g)is
cost due to losses, negligible in this case, GFEp(g)is total fuel
expense of g units, F(ct)is fixed cost Fct =ST(ct)+DP(ct),
ST(ct)is start up cost, DP(ct)is depreciation cost, M(ct)
is maintenance cost, P(g)is power produced by generator
g, P(g,mx)is maximum power generation limit, P(g,mn)is
minimum power generation limit, E(t,der)is total energy
consumed by der-connected load.
IV. MATHEMATICAL MODEL OF THE PROB LEM
A. Problem Discussion.
In this section we discuss and formulate the mathemat-
ical model of our problem.Our problem is to reduce total
energy consumption,minimize total energy cost and also to
reduce the expenses incur by der energy source.To solve
this optimization problem we propose an optimization model
discussed in system model.Our optimization model includes
energy consumption minimization,energy cost minimization
and energy generation.We have a set of appliances in a smart
home which will consume energy from two main sources i.e.
lceg and lcd which are discussed and formulated in previous
section.These two types of energy source is divided into
further load types which are discussed in start.The regular load
is running on lcd load which is also known as microgrid or
der.After modeling the parameters of our optimization problem
and and its constraints,the model is formulated using binary
MNK P (multiple knapsack problem).The problem is solved
using Genetic Algorithm.Fitness function of genetic algorithm
can be written as below in equation(14).
F F =
24
X
t=1
n
X
α=1
BV(αt)EL(lceg,t,ec)EP (t)
+
24
X
t=1
n
X
α=1
BV(αt)EL(lcd,t,ec)EP (t)(14)
where FF is fitness function.
B. Objective Function.
The objective function can be written as follow.
Minimize
24
X
t=1
n
X
α=1
BV(αt)EL(lceg,t,ec)
+
24
X
t=1
n
X
α=1
BV(αt)EL(lcd,t,ec)(15)
subject to the following constraints
24
X
t=1
n
X
α=1
BV(αt)=LOTα(15a)
24
X
t=1
EαBV(αt)≤ψ(t)(15b)
24
X
t=1
EαBV(αt)≤ψ(t)(15c)

δα≤T−LOTα∀lceg (15d)
tst
α≤δα≤tst
α+LOTα∀lcd (15e)
δα=tst
α∀lcd, tst
α∈Tst
lcd (15f)
BV(αt)∈ {0,1}(15g)
where
Tst
lcd time in which the appliance αis ON,ψ(t)is energy
consumption limit or total quantity of energy, δαis scheduled
time of appliance α,tst
αis starting time of appliance
α,BV(αt)is a binary variable which shows that appliance
is on or off,LOTαis appliance αtotal time in which it operate.
ECH Algorithm
1: Required pattern of appliances before scheduling,tbs
α,ps,
mxgeneration,αn.
2: Start random population representing the pattern of
appliances.
3: for t=1:24 do
4: d=0
5: for α=1:ps do
6: Execute fitness function equation (14)
7: FF=fitness
8: if (F(α)< F (α−1)&&(ET(t)) < CP (t)) then
9: F(α) = F(α)
10: if lcegα(t−1) == 1 then
11: tst
α=t
12: dα=tbs
α-tst
α
13: if (dα≥3)&&(δα≤LOTα)
14: lceg(t)=1 end if end if
15: if lcdα(t)==1 then
16: tst
(α)=t
17: dα=tbs
α-tst
α
18: if (dα≥3)&&(δα≤LOTα)
19: lcd(t)=1 end if end if else
20: F(α) = F(α−1) end if end for
21: ap(1,:)=pn(1,:)
22: if apα==1 then
23: LOTα=LOTα−1end if
24: d=lcd(t)
25: solve optimization problem equation(13).
Generation of new population.Selection of crossover pair x,y
26: if Pc> rand then
27: crossover (x,y) end if
28: if Pm> rand then
29: crossover (x,y) end if
30: pn(ps,αn) end for
where αnis no of appliances,ap is appliances pattern,CP(t)
is total energy capacity or limit in time slot t,ps is popsize and
pn is popnew.
V. CONCLUSION
In this research work we proposed an efficient ECG op-
timization model including energy consumption minimiza-
tion,energy consumption cost minimization and energy gen-
eration cost optimization for smart home energy management
based on DSM. An ECG algorithm is proposed for balancing
total energy consumption load, cost of energy consumption
and energy generation cost for efficient scheduling of appli-
ances.We are using two main types of energy sources i.e. lceg
and lcd can also be named as macrogrid and microgrid for our
smart home appliances.The lcd energy source act as microgrid
or DER are used to reduce burden of load consumption and
run regular load on it.The simulation results show that our
proposed model and ECG algorithm balencing the energy
consumption and reduces both types of cost as compared to
the results without ECG model.In future we intend to further
elaborate this work by including PAR model,user discomfort
model and implementation of appliances specific constraints.
REFERENCES
[1] Phani Chavali, Student Member, IEEE, Peng Yang, Student Mem-
ber, IEEE, and Arye Nehorai, Fellow, IEEE, “A Distributed Al-
gorithm of Appliance Scheduling for Home Energy Management
System”, IEEE TRANSACTIONS ON SMART GRID, VOL. 5,
NO. 1, JANUARY 2014.
[2] Xiaodao Chen, Tongquan Wei, Member, IEEE, and Shiyan Hu,
Senior Member, IEEE, “Uncertainty-Aware Household Appliance
Scheduling Considering Dynamic Electricity Pricing in Smart
Home”,IEEE TRANSACTIONS ON SMART GRID, VOL. 4,
NO. 2, JUNE 2013.
[3] Kin Cheong Sou, James Weimer, Henrik Sandberg, and Karl
Henrik Johansson, “Scheduling Smart Home Appliances Using
Mixed Integer Linear Programming”,2011 50th IEEE Conference
on Decision and Control and European Control Conference
(CDC-ECC) Orlando, FL, USA, December 12-15, 2011.
[4] Kin Cheong Sou, Mikael K´
lordel, Jonas Wu, Henrik Sand-
berg and Karl Henrik Johansson,“Energy and CO2 Efficient
Scheduling of Smart Home Appliances”,2013 European Control
Conference (ECC)July 17-19, 2013, Zurich, Switzerland.
[5] M. B. Rasheed, M. Awais, N. Javaid,Z. Iqbal, A. Khur-
shid, F. A. Chaudhry, F. Ilahi,“An Energy Efficient Residen-
tial Load Management System for Multi-Class Appliances in
Smart Homes”,Conference:18th IEEE International Conference
on Network-Based Information Systems (NBiS-2015), Taipei,
Taiwan.
[6] M. Awais, N. Javaid,N. Shaheen,Z. Iqbal,G. Rehman,K. Muham-
mad,I. Ahmad,“An Efficient Genetic Algorithm Based Demand
Side Management Scheme for Smart Grid”,Conference:18th IEEE
International Conference on Network-Based Information Systems
(NBiS-2015), Taipei, Taiwan.
[7] Chen Chen, K.G. Nagananda, Gang Xiong, Shalinee Kishore,
Member, IEEE and Lawrence V.Snyder, Member, IEEE,“A
Communication-Based Appliance Scheduling Scheme for
Consumer-Premise Energy Management Systems”,International
Journal of Smart Home Vol. 7, No. 4, July, 2013.
[8] Taeyoon Yu,Dong Sik Kim and Sung-Yong Son,“Optimization of
Scheduling for Home Appliances in Conjunction with Renewable
and Energy Storage Resources”,IEEE TRANSACTIONS ON
SMART GRID, VOL. 4, NO. 1, MARCH 2013.
[9] Mohammad Rastegar,Mahmud Fotuhi-Firuzabad,Farrokh Amini-
far,“Load commitment in a smart home”,Applied Energy 96
(2012) 45 54.