10th IFAC Symposium on Control of Power and Energy Systems

10th
10th IFAC
Tokyo, Symposium
Japan,
IFAC on
September
Symposium Control
on 4-6, 2018of
Control of Power
Power and
and Energy
Energy Systems
Systems
10th IFAC
Tokyo,
10th Symposium
Japan,
IFAC on
September
Symposium Control
4-6,
on 4-6, 2018of
Control Available
of Power
Power andonline
and EnergyatSystems
Energy www.sciencedirect.com
Systems
Tokyo, Japan,
10th IFAC September
Symposium on 4-6, 2018
Control of Power and Energy Systems
Tokyo, Japan, September 2018
Tokyo, Japan, September 4-6, 2018
Tokyo, Japan, September 4-6, 2018
ScienceDirect
IFAC PapersOnLine 51-28 (2018) 486–491
Load Shedding Optimization Considering Consumer Appliance Prioritization
Load Shedding Optimization
Using Considering
Genetic Algorithm ConsumerApplication
for Real-time Appliance Prioritization
Load
Load Shedding
Shedding Optimization
Optimization Considering
Considering Consumer
Consumer Appliance
Appliance Prioritization
Prioritization
Using Genetic Algorithm for Real-time Application
Using GeneticMarven
Algorithm for Real-time Application
Using Genetic Algorithm for
E. Jabian, Real-time
Ryohei Application
Funaki, and
Marven E. E. Jabian,
Jabian, Ryohei Funaki, and and
Marven Junichi Ryohei
MurataFunaki,
Marven
Marven E. E. Jabian,
Junichi Ryohei
Jabian, MurataFunaki,
Ryohei Funaki, and and
Marven E. Jabian,Junichi Murata
Ryohei Funaki, and
Junichi Murata
Junichi Murata
Kyushu University, 744 Motooka, JunichiNishi-ku,
Murata Fukuoka, Fukuoka, Japan
Kyushu
Kyushu University,
University,
(Tel : 744
744 Motooka,
Motooka, Nishi-ku,
+81-92-802-3675; Nishi-ku, Fukuoka,
Fukuoka, Fukuoka,
Fax: +81-92-802-3692) Fukuoka, Japan
Japan
Kyushu
Kyushu University,
University,
(Tel : 744
744 Motooka,
Motooka,
+81-92-802-3675; Nishi-ku,
Nishi-ku,
Fax: Fukuoka,
Fukuoka,
+81-92-802-3692) Fukuoka,
Fukuoka, Japan
Japan
(Tel : +81-92-802-3675;
Kyushu University, Fax: +81-92-802-3692)
744 Motooka, Nishi-ku,
(marven.jabian@g.msuiit.edu.ph) Fukuoka, Fukuoka, Japan
(Tel
(Tel :: +81-92-802-3675;
+81-92-802-3675; Fax:
Fax: +81-92-802-3692)
+81-92-802-3692)
(marven.jabian@g.msuiit.edu.ph)
(marven.jabian@g.msuiit.edu.ph)
(Tel : +81-92-802-3675; Fax: +81-92-802-3692)
(marven.jabian@g.msuiit.edu.ph)
(marven.jabian@g.msuiit.edu.ph)
(marven.jabian@g.msuiit.edu.ph)
Abstract: In the event of energy supply-demand imbalance caused by deficiency in energy generation,
Abstract:
Abstract:
Distribution In the event
InUtilities
the event of energy
of
(DUs) energy
implementsupply-demand
supply-demand
load sheddingimbalance
imbalance
methods caused
caused by deficiency
by
to avoid deficiency in energy
in
system damages. energy generation,
In generation,
the current
Abstract:
Abstract:
Distribution InUtilities
In the event
the event of energy
of
(DUs) energy
implementsupply-demand
supply-demand
load imbalance
imbalance
shedding methods caused
caused
to by deficiency
by
avoid deficiency
system in energy
in energy
damages. In generation,
generation,
the
Distribution
Abstract:
distribution InUtilities
the
set-up, event(DUs)
of
consumers implement
energy load
supply-demand
experience shedding methods
imbalance
unscheduled or to
caused
scheduled avoid
by system
deficiency
total blackoutsdamages.
in energy
with no In the current
current
generation,
control over
Distribution
Distribution
distribution Utilitiesconsumers
Utilities
set-up, (DUs) implement
(DUs) implement
experience load
load shedding methods
shedding
unscheduled methods
or to avoid
to
scheduled avoid
totalsystem
system
blackoutsdamages.
damages.
with no Incontrol
In the current
the current
over
distribution
Distribution
which set-up,
Utilities
appliances to consumers
(DUs)
spare. This experience
implement
paper unscheduled
load shedding
introduces a novel or scheduled
methods
method toto total
avoid blackouts
system
implement with
damages.
automated no Incontrol
load the over
current
shedding,
distribution
distribution
which set-up,toconsumers
set-up,
appliances consumers
spare. This experience
experience
paper unscheduled
unscheduled
introduces a novel ormethod
or scheduled
scheduledto total blackouts
total blackouts
implement with no
with
automated no control
control
load shedding,over
over
which appliances
distribution
considers to spare.
set-up,appliance
consumersThis paper
experience
activities introduces a
and unscheduled
priority novel method
or scheduled
levels to
as predefinedimplement
total by automated
blackouts load
with no control
theautomated
consumers, shedding,
inshedding,over
a smart
which
which
which appliances
appliances
considers to
to spare.
spare.
appliance This
This paper
paper
activities introduces
introduces
and a
a novel
novel method
method to
to implement
implement automated load
load shedding,
which considers
appliances
distribution system. appliance
to spare.
The activities
This
proposed and priority
papermethod priority
introduces levels
levels
a novel
utilizes the as
as predefined
predefined
method by
by the
to implement
information from the consumers,
consumers,
automated
the loadin
distributed aa smart
inappliance
smart
shedding,
which
which considers
considers appliance
appliance activities and
and priority
activities method priority levels
levels as predefined by
by the consumers, in aa smart
distribution
distribution
which system.
system.
considers
controllers which areThe
The
appliance proposed
proposed
assumedactivities
to havemethod
and power utilizes
utilizes
priority the as
the
levels
monitoring as
predefinedfrom
information
information
and directfrom
predefined by
the
loadthethe
theconsumers,
distributed
distributed
consumers,
control
inappliance
smart
inappliance
capabilities a smart
with
distribution
distribution
controllers system.
system.
which areThe
The proposed
proposed
assumed to method
method
have utilizes
utilizes
power the
the
monitoring information
information
and directfrom
from
load the
the distributed
distributed
control appliance
appliance
capabilities with
controllers
distribution
bidirectional which
system. areThe
communication. assumed
proposed to consumer
Since have
method power monitoring
utilizes
appliance the and is
information
switching direct load
from
binary in control
the capabilities
distributed
nature, Genetic appliance
Algorithm with
controllers
controllers
bidirectional which
which are
are assumed
communication. assumed to have
have power
to consumer power monitoring
monitoring and
and is direct
direct load
load control capabilities
controlGenetic
capabilities with
with
bidirectional
controllers
(GA) is used
bidirectional
communication.
which
to perform
communication. the Since
are assumed Since
to consumer
have power
optimization
Since consumer
appliance
appliance
which switching
switching
monitoring
is to allocate
appliance
andtheis
switching is
binary
binary
direct
binary
in
in nature,
load
available
in
nature, Genetic
controlsupply
power
nature, toAlgorithm
Algorithm
capabilities
Genetic as many
Algorithm
with
bidirectional
(GA)
(GA) is used
is used communication.
to perform
to perform the Since
the Since consumer
optimization
optimization appliance
which
which is to switching
allocate
is to allocate theis
theis binary in
available
available nature,
power
power Genetic
supply
supply toAlgorithm
as
toAlgorithmmany
as power
many
bidirectional
appliances as communication.
possible considering theconsumer appliance
consumer-defined switching
appliance binarylevels.
priority in nature,
With Genetic
the limited
(GA)
(GA) is
is used
appliancesused to
to perform
perform
as possible
possible the
the optimization
consideringoptimization which
which is
theisconsumer-defined
consumer-definedis to
to allocate
allocate
appliance the available
thepriority
available
priority power
Withsupply
power
levels. supply
theeach to
to as
limited many
as power
many
appliances
(GA)
supply, as
is consumer
used to perform
powerconsidering
the the
optimization
allocation appliance
which isbyto executing
determined allocate the
two GA levels.
available With
power
processes, inthe
supply limited as power
toappliancemany
appliances
appliances
supply, as
as possible
possible
consumer powerconsidering
considering
allocation the
the isconsumer-defined
consumer-defined
determined by appliance
appliance
executing priority
priority
two GA levels.
levels. With
With
processes, inthe
the limited
limited
each power
power
appliance
supply,
appliances
controllerconsumer
as possible power
and in thepower allocation
considering
centralallocation the is determined
consumer-defined
station, respectively. Thebyby executing
appliance
GAexecuting two
process intwo GA
priority
eachGA processes,
levels.
appliance With inthe
controller each appliance
limited
allocates power
the
supply,
supply,
controllerconsumer
consumer
and in thepower
in capacity
centralallocation
station, is determined
is determined
respectively. The by
GA executing intwo
processtheir eachGA processes,
processes,
appliance in each
in
controller each appliance
appliance
allocates the
controller
supply,
available and
consumer
supply the central
power to station,
allocation
the respectively.
enrolledis determined
appliances The GA
by
to process
executing
determine in each
two GAappliance controller
processes,
switching on or off allocates
inconsidering
each appliance the
their
controller
controller
available and in
and
supply in capacity
the central
the central
to station,
station,
the respectively.
respectively.
enrolled appliances The
The toGA
GA processtheir
process
determine in each
in each appliance
appliance
switching on controller
controller
or off allocatestheir
allocates
considering the
the
available
controller
priority supply
and
levels. in capacity
the
In order toto
central the
tothe enrolled
station,
avoid appliances
respectively.
repeated The to
switching determine
GA process their
in
of particular switching
each appliance
appliances, on or off considering
controller
'fairness' allocates
of switching their
the
available
available
priority supply
supply
levels. capacity
capacity
In order to
to the enrolled
enrolled
avoid appliances
appliances
repeated to
to
switching determine
determine
of their
their
particular switching
switching
appliances, on
on or
or off
off considering
considering
'fairness' of switchingtheir
their
priority
available levels. In order
supply capacity
implementations to
toby avoid repeated
thea proposed
enrolled appliancesswitching of particular
to determine their appliances,
switchingsupply 'fairness'
on or off of switching
considering their
priority
priority levels. In
levels. Inis judged
order to
order to avoid
avoid repeated
repeated
criterion. The
switching
switching
remaining
of particular
of
unallocated
particular appliances,
appliances,
capacity
'fairness'
'fairness' ofis
of
collected
switching
implementations
implementations
priority
and Inis
is judged
order toby
judged
levels.redistributed
optimally by
by aaGA
proposed
proposed
avoid repeated
in the criterion.
criterion. The
The remaining
switching
central station. remaining
of
The unallocated
unallocated
particular
case studyappliances,
results supply
supply capacity
capacity
'fairness'
showed that ofisswitching
is
the collected
collected
switching
proposed
implementations
implementations is
is judged
judged by by a proposed criterion. The
The remaining unallocated supply capacity is collected
and
and optimally
optimally
implementations
method ensures redistributed
redistributed byaaGA
by
is judged power
optimum by
proposed
GA in
proposed the criterion.
in the
utilizationcentral
central station.
station.
criterion.
to avoid
remaining
ThetotalThe
The case unallocated
case
remaining study
study results
results
unallocated
blackouts with
supply
showed
showed
fastsupply
capacity
that
that the
capacity
convergence
is proposed
the collected
is proposed
collecteda
signifying
and
and optimally
optimally
method ensures redistributed
redistributed
optimum by
by
power GA
GA in
in the
the
utilizationcentral
central
to station.
station.
avoid The
The
total case
case
blackoutsstudy
study results
results
with fast showed
showed
convergencethat
that the
the proposed
proposed
signifying
method
and
promising
method
ensures
optimally optimum
redistributed
capability
ensures optimum for power
by GAutilization
real-time
power utilization
to avoid
inapplications.
the central
to
station.total
The
Furthermore,
avoid total
blackouts
case
the
blackouts
studywith
proposed
with
fastmethod
results
fast
convergence
showedisthat
convergence ablethe involveaaa
signifying
proposed
to
signifying
method
promising
promising ensures optimum
capability
capability for
for power
real-time
real-time utilization to
applications.
applications. avoid total
Furthermore,
Furthermore, blackouts
the
the with
proposed
proposedfast convergence
method
method is able signifying
to involveaa
to involve
is ablesignifying
method ensures
consumers optimum
in deciding whichpower utilization
appliances to avoid
to deload total their
through blackouts
priority with fastinputs.
level convergence
promising
promising
consumers capability
capability
in deciding for
for real-time
real-time
which appliancesapplications.
applications.
to deload Furthermore,
Furthermore,
through their the
the proposed
proposed
priority level method
method
inputs. is
is able
able to
to involve
involve
consumers in decidingfor
promising capability which appliances
real-time to deloadFurthermore,
applications. through their the priority level inputs.
proposed method is able to involve
consumers
Keywords:
© 2018, IFAC
consumers Power
in deciding Distribution,
which
(International
in deciding Frequency
appliances
whichFederation
appliances to
of to Stabilization,
deload
Automatic through Distribution
their
Control)their
deload through priority
Hosting Automation,
level
by Elsevier
priority inputs. Distributed Control,
Ltd. All rights reserved.
level inputs.
consumers in deciding
Power which appliances
Distribution, Frequency to deload through Distribution
Stabilization, their priorityAutomation,
Distribution level inputs. Distributed
Keywords:
Keywords:
Genetic
Keywords:
Power
Algorithm,
Power
Distribution,
Load Frequency
Distribution,
Frequency
Control,Stabilization,
Frequency Load Regulation,
Stabilization, OptimalAutomation,
Distribution Control, Priority
Automation,
Distributed Control,
Control,
Distributed Control,
Control,
Keywords:
Genetic
Genetic Power
Algorithm,
Algorithm, Distribution,
Load Frequency
Load Frequency Frequency Stabilization,
Control,Stabilization,
Control, Load Regulation,
Load Distribution
Regulation, Automation,
OptimalAutomation,
Optimal Control, Priority
Control, Distributed
Priority
Keywords: Power Distribution, Frequency Distribution Distributed Control,
Genetic Algorithm,
Genetic Algorithm, Load Load Frequency
Frequency Control,Control, Load
Load Regulation,
Regulation, Optimal
Optimal Control,
Control, Priority
Priority
Genetic Algorithm, Load Frequency Control, Load Regulation, Optimal Control, Priority
1. INTRODUCTION interruption and increased electricity rates, among others. The
1.
1. INTRODUCTION
INTRODUCTION interruption
interruption
repeated and
and increased
blackouts caused electricity
increased electricity
inconveniencerates,
rates,toamong
both others.
among others.
householdsThe
The
A Distribution Utility 1. INTRODUCTION
1. (DU)
INTRODUCTION
in the Philippines, as a franchise interruption
interruption
repeated and
and
blackouts increased
increased
caused electricity
electricity
inconveniencerates,
rates, among
among
to both others.
others.
householdsThe
The
A Distribution 1.
Utility INTRODUCTION
(DU) in repeated
and blackouts
interruption
businesses, and since caused
increased inconvenience
electricity
consumers in rates,
the to
affected both
among households
others.
areas have The
no
A Distribution
holder of a Utility
certain area (DU)
is in the
obligedthetoPhilippines,
Philippines,
source for itsas aa franchise
as consumers’
franchise repeated
repeated
and blackouts
blackouts
businesses, since caused
caused inconvenience
inconvenience
consumers in the to
to
affected both
both households
households
areas have no
A
A Distribution
Distribution Utility
Utility (DU)
(DU) in the
in thetoPhilippines,
Philippines, as consumers’
as aa franchise
franchise and businesses,
repeated
control blackouts
over which since consumers
caused in the
inconvenience affected
to areas
both have
households no
holder
holder
A of a
Distribution
power certain
ofconsumption
a certain area
areaand
Utility is
(DU) obliged
is obliged
maintain source
source
the for
in thetoPhilippines, its
for its
as consumers’
required a franchise
technical and
and businesses,
businesses,
control over sinceappliances
since
which consumers to
consumers
appliances in
to
spare,
in the
the not even
affected
affected
spare, not areas aahave
areas
even
single
have no
no
single
holder
holder of a certain area is obliged to source for its consumers’ control
and over
businesses, which
since appliances
consumers to
in spare,
the not
affected even
areas a single
have no
power of
power
holder
standards
a certain areaand
consumption
ofconsumption
aascertain areain
specified
is obliged
and
isthe maintain
maintain
obliged
to source
to source
Distribution
for its consumers’
the Code
the required
required
for of
its the technical
technical
consumers’
country.
lighting
control
control
lighting
fixture.
over
over
fixture.
On the
which
which
On the
other hand,tothe
appliances
appliances
other hand,to
country
spare,
spare,
the countrynot
not
has a traditional
even
even
has a a
a single
single
traditional
power
power consumption
consumption and
and maintain
maintain the
the required
required technical
technical lighting
control
grid fixture.
over
power On
which
system the other
applianceshand,tothe country
spare, not has a
even traditional
a single
standards
standards
power
In as specified
as specified
consumption
accordance with in
in
the the Distribution
andtheexisting
Distribution
maintain Code
the Code of the
of therules
required
implementing country.
country.
technical
and lighting
lighting
grid fixture.
fixture.
power systemOnwherein
On the otheraa hand,
the other
wherein
unidirectional
hand, the country
the country
unidirectional
infrastructure
has
infrastructure
flow
has aa traditional
traditional
flow
standards
standards
In accordance as
as specified
specified
with in
in
the the
the Distribution
Distribution
existing Code
Code
implementing of
of the
the country.
country.
rules and grid
of power
lighting
electricity system
fixture. isOninwhereinothera hand,
theeffect. unidirectional
Power thesystem
countryinfrastructure
adjustments flow
has a traditionalare
In accordance
standards
regulations thewith
asofspecified the
in the
Electric existing
Power implementing
Distribution
Industry Code
Reform of the
Act rules
of and
country.
2001 grid
grid
of power system
power
electricity system
is inwherein
wherein
effect. a unidirectional
a unidirectional
Power system infrastructure
infrastructure
adjustments flow
flow
are
In
In accordance
accordance
regulations of thewith
with the
the
Electric existing
existing
Power implementing
implementing
Industry Reform Act rules
rules
of and
and
2001 of
gridelectricity
power system
communicated is
from inwherein
effect.
grid Powertosystem
a unidirectional
operators each of adjustments
infrastructure
the DUs are
flow
system
regulations
In
law,accordance
the of the
DUs in Electric
with the Power
Mindanao, existing
a Industry
southern Reform
implementing Act
part/island ofof2001
rules and
the of
of electricity is
electricity
communicated is
from in grid
in effect.
effect. Powertosystem
Power
operators system
each of adjustments
adjustments
the DUs are
are
system
regulations
regulations
law, the of
of the
DUs the
in Electric
Electric
Mindanao, Power
Power a Industry
Industry
southern Reform
Reform Act
Act of
part/island ofof
of2001
2001
the communicated
of electricity
operators to doisfrom in grid
the operators
effect.
manual Power tosystem
switching.eachIn of the DUs
adjustments
times system
of massive are
law, the
regulations
Philippines, DUs
of in
the Mindanao,
Electric Power a southern
Industry part/island
Reform Act of the
2001 communicated
communicated from
dofrom grid
the grid operators
operators to each
to each of the
oftimes DUs
the DUs system
system
law,
law, the DUs
the
Philippines, DUshave in
have
base load apower
in Mindanao,
Mindanao,
base load apower
sources
southern
southern
consisting
part/island
part/island
sources consisting of the
of
of
the
of
operators
operators
communicated
power to do
to
supply-demand the
from manual
manual
grid switching.
switching.
operators
imbalance, to each
the In operators
In
grid oftimes of have
of
the DUs massive
massive
system
the
Philippines,
law, the
indigenous DUs have
and in base
Mindanao,
renewable load apower sources
southern
energies: consisting
part/island
hydro, of
geothermal, of
the operators
operators
power to
to do
do the
supply-demand the manual
manual switching.
switching.
imbalance, the In times
times of
In operators
grid massive
of have
massive
Philippines,
Philippines,
indigenous have renewable
have
and base load
base load energies:
power sources
power sources
hydro, consisting
consisting
geothermal, of
of power
capacitysupply-demand
operators toto do
do an imbalance,
the automatic
manual or the
switching.
manual In operators
grid of have
timesdropping
load massivethe
the
to
indigenous
Philippines,
biomass and
and andhave renewable
base
solar power load energies:
power
plantsenergies: hydro,
sources
(EPIRA, 2001). geothermal,
consisting of power
power supply-demand
supply-demand imbalance,
imbalance, the
the grid
grid operators
operators have
have the
the
indigenous
indigenous
biomass and ofand
solar renewable
renewable
power plantsenergies:
(EPIRA,that hydro,The
hydro,
2001). The
country’s
geothermal,
geothermal,
country’s
capacity
capacity
power to
to do do
supply-demand
stabilize the grid an
an(ERC,automatic
automatic
imbalance,
2001). or
or manual
manual
the load
grid operators
Typically, load dropping
load dropping
dropping
have can to
to
the
biomass
indigenous
Department and solar
and power
renewable
Energy plants
(DOE) (EPIRA,
energies:
reported 2001).
hydro,
onlyThealmostcountry’s
geothermal,40% capacity
capacity
stabilize to
to
the do
do
grid an
an(ERC,automatic
automatic
2001). or
or manual
manual
Typically, load
load
load dropping
dropping
dropping to
to
can
biomass
biomass
Department and of
and solar
solar power(DOE)
power
Energy plantsreported
plants (EPIRA,that
(EPIRA, 2001).
2001).
onlyThe
Thealmostcountry’s
country’s
40% stabilize
capacity
be resolvedthe
to by grid
do an(ERC,
installing 2001).
automatic
more orTypically,
power supplyload
manual load dropping
dropping
capacities. can
to
In fact,
Department
biomass
of the total of Energy
andcapacity
solar power
came (DOE)
plants
from reported
(EPIRA,
renewablethat onlyThe
2001).
energy almost
and 31%40%
country’sof stabilize
stabilize
be resolvedthe
the bygrid
grid (ERC,
(ERC,
installing 2001).
2001).
more Typically,
Typically,
power supply load
load dropping
dropping
capacities. In can
can
fact,
Department
Department
of the total of Energy
of Energy
capacity came (DOE)
(DOE)
from reported
reported
renewablethatenergy
that only almost
only almost
and 31%40%
40%of be resolved
stabilize the
privately-owned by installing
grid (ERC,
coal-firedmore
2001).
powerpower supply
Typically,
plants capacities.
load
started dropping In
its testing fact,
can
and
of thecame
that total from
Department capacity
of Energy came
hydro fromreported
(DOE)
power renewable
plants thatenergy
but and 31%
onlyunfortunately
was almost 40%of be
be resolved by
resolved
privately-owned by installing
installing
coal-firedmore
morepowerpower
power supply
supply
plants capacities.
capacities.
started In fact,
In
its testing
testing fact,
and
of
of the
the
that total
total
came capacity
capacity
from came
came
hydro from
from
power renewable
renewable
plants but energy
energy
was and
and 31%
31%
unfortunately of
of privately-owned
be
commissioning coal-fired
resolved by installing
stages morepower
early plants
power
2017 started
supply
(DOE its
capacities.
Report, In and
fact,
2017).
that
of thecame
having total from
capacity
reduced hydro
came
generation powerdue plants
from but energy
renewable
to adverse was
effectunfortunately
and 31%
made of
by El privately-owned
privately-owned
commissioning coal-fired
coal-fired
stages power
power
early plants
plants
2017 startedReport,
started
(DOE its testing
its testing and
and
2017).
that
that
havingcame
came from
from
reduced hydro
hydro
generation power
powerdue plants
plants
to but
but
adverse was
was
effectunfortunately
unfortunately
made by El commissioning
privately-owned
However, stages
coal-fired
this isstages early
a costlyearlypower 2017
and 2017 plants(DOE
not easily started Report,
its
accessible testing2017).
and
solution
having
that
Niño came
thatreduced
from
lead togeneration
hydro power
curtailment due
of to adverse
plants
power but
supplyeffect
was made
unfortunately
during thebyfirst
El commissioning
commissioning
However, this is stages
a costlyearly
and 2017
not (DOE
(DOE
easily Report,
Report,
accessible 2017).
2017).
solution
having
having reduced
reduced generation
generation due
due to adverse effect made by El However,
commissioning this is a costly
stagesdemand.and not easily accessible
early 2017 (DOE Report, 2017). solution
Niño
Niño
having
half ofthat
that
the lead
yearto
lead
reduced curtailment
to2016
curtailment
generation
(DOEdue of to
of adverse
power
power
to
Report, supply
supply
adverse
2016).
effect
DUs
made
during
during
effect inmade the
thethe
byfirst
El
byfirst
El
region
especially
However,
However,
especially
during
this
during is peak
this is aa costly
peakcostly and not
and
demand. not easily
easily accessible
accessible solution
solution
Niño
Niño
half that
that
ofthat lead
lead
thelead to
to curtailment
curtailment
yearcomplaints
2016 (DOE of
of power
power
Report, supply
supply
2016). DUs during
during
in the
the
thethe first
first
region especially
However, duringthis is peak demand.
a costly and not easily accessible solution
half
Niño
are of the
receiving year to2016 (DOE
curtailment Report,
fromof their 2016).
power supplyDUs
consumers in
during
due thetoregion
first
poor especially during peak
The utilization
especially during peak demand.Load Control (DLC) with
of Direct
demand.
half
half
are of
of the
the year
receiving 2016
2016 (DOE
yearcomplaints (DOE Report,
their2016).
Report, 2016). DUs
DUs in in thetoregion
thetoregion especially during peak demand.Load Control (DLC) with
are
half receiving
of the
service yearcomplaints
quality, 2016 (DOE
unscheduled from
from their
Report, consumers
consumers
2016).
blackouts, longDUs due
due
in
hours the
of poor
poor
region
power The
The utilization
utilization
bidirectional of Direct
of
communicationDirect Load
is a Control
favourable(DLC) with
alternative
are
are receiving
receiving
service complaints
complaints
quality,complaints
unscheduled from
from their
their
blackouts,consumers
consumers
long hours
hoursdue
due to
to poor
poor
oftopower
power The
The utilization
utilization
bidirectional of
of
communicationDirect
Direct Load
Load Control
is aa Control (DLC)
favourable(DLC) with
with
alternative
service quality,
are receiving unscheduled fromblackouts, long
their consumers dueof poor bidirectional
The utilization communication
of Direct Load is favourable
Control alternative
(DLC) with
service quality,
service quality, unscheduled
unscheduled blackouts,
blackouts, long
long hours
hours of of power
power bidirectional communication
bidirectional communication is is aa favourable
favourable alternative
alternative
service quality, unscheduled blackouts, long hours of power bidirectional communication is a favourable alternative
2405-8963 ©
Copyright © 2018, IFAC (International Federation of Automatic Control)
2018 IFAC 486Hosting by Elsevier Ltd. All rights reserved.
Peer review©under
Copyright 2018 responsibility
IFAC of International Federation of Automatic
486Control.
Copyright © 2018 IFAC 486
10.1016/j.ifacol.2018.11.750
Copyright
Copyright © 2018
© 2018 IFAC
IFAC 486
486
Copyright © 2018 IFAC 486
 IFAC CPES 2018
Tokyo, Japan, September 4-6, 2018
Marven E. Jabian et al. / IFAC PapersOnLine 51-28 (2018) 486–491 487
solution to the power supply-demand imbalance and sudden
changes in frequency and voltage levels in the distribution
system. The study of Mortaji et al. (Mortaji et al., 2016)
showed that smart DLC was able to reduce high peak load
demand to average demand and minimized the blackouts. Liu,
Hill and Zhang (Liu, Hill and Zhang, 2016) designed a
switching controller that enables DLC to quickly respond to
frequency disturbance and then gradually transfer the
frequency restoration to the power generators to achieve a
smooth transition from load control to automatic generation
control in real time. Moreover in Dandan Zhang et al. (Dandan
Zhang et al., 2016) study, the strategy was optimized in a load
shedding control that considered economic cost of load
shedding and multi node frequency balancing. These studies
have shown that investments on DLC equipment and
infrastructure in a smart distribution system is efficient in
optimizing the consumption of the available power supply.
However, these studies did not include consumer inputs as part
of the load shedding control variable.
A smart distribution system is an improved power distribution
network using technologies that enhance communication
between DUs and their consumers to effectively respond to
changes or disturbances in the network. With this set-up, the
information on frequency and voltage levels are easily
monitored which is crucial in the management of power
supply-demand balance.
In this paper, we propose a method to automate load shedding
in a smart distribution system that considers consumers'
appliance priority levels and switching activities to avoid
repeated deloading. The research assumes that DU’s
distribution and designated consumer nodes are installed with
smart appliance controllers that have power monitoring, data
logging, and DLC capabilities with bidirectional
communication. The proposed method uses LFC droop
relation as in the study of (Elrayyah, Cingoz and Sozer, 2017)
to verify that the adjusted load or the new supply allocation
keeps the system frequency within the allowable range prior to
optimization. Then the method selects suitable appliances to
be deloaded considering the consumer-defined priority levels,
amount of loads to be shed and the appliances’ switching
activities. The method further utilizes a two-level hierarchical
structure to facilitate the optimization and switching
operations that involve a large number of appliances for real
time application. In binary combinatorial problems, the use of
metaheuristic techniques is highly favorable in dealing with a
large number of variables. Among them, Genetic Algorithm
(GA) is promising due to its execution speed and considerable
accuracy (Uddin, Abido and Rahman, 2003), which is the case
in the current problem as will be validated in the case studies.
GA is executed in each level, at each appliance controller (K)
and at the central station (CS), respectively. In K, GA finds the
set of appliances to be switched on. The appliances in the set
should have higher priority levels and the sum of their power
consumption should be as close as possible to, but not higher
487
than the power supply capacity allocated to K by CS. The
resulting appliance combinations, from GA execution, are
subjected to the developed fairness function. This is done to
eliminate or avoid biased/unfair switching actions. The next
GA is in CS. Before its execution, the remaining unallocated
power and appliance nominations with the least power
consumption and low switching activity from each K are
collected. The accumulated power becomes the available
power supply for reallocation. In CS, given the appliance
nominations from each K to CS, GA finds a set of suitable
appliances to be switched on. The sum of the power
consumption of the appliances in the set should be as close as
possible to, but not higher than the remaining available power
supply. The combined results of the GA executions determine
the final solution for each K, for the final switching
implementation. Case study simulations show that the
proposed method enables the consumers to have control over
which appliances to switch off and avoid total blackouts.
Furthermore, the GA optimization processes in this study
converge quickly, which signifies that the proposed method is
capable of real time applications.
2. PROPOSED SYSTEM DESIGN
2.1 System Network Topology
Figure 1 illustrates the overall system network topology. The
Central Station (CS) is the DU’s main server for power
management, monitoring, data logging and switching controls,
K1 , K 2 , … and K 𝑛𝑛 are the distributed appliance controllers,
where n corresponds to the number of installed K, and X𝑗𝑗1 , X𝑗𝑗2 ,
… and X𝑗𝑗𝑗𝑗 are individual consumer appliance control and
monitoring nodes under K𝑗𝑗 , where m corresponds to the
number of enrolled consumer appliances. Each of the X𝑗𝑗𝑗𝑗
consists of 5 appliances, with corresponding consumer priority
levels p, set as 1 to 5. The appliance rated power is denoted as
𝐿𝐿1𝑗𝑗𝑗𝑗 ,…, 𝐿𝐿𝑗𝑗𝑗𝑗
5 5
, with 𝐿𝐿𝑗𝑗𝑗𝑗 as the appliance rating with the lowest
priority level, indicating that the consumer readily accepts that
X2m X32
X31
X22
K2 K3
X3m
X21
CS
X11
Xn1
K1 Kn
X12
Xn2
X1m Xnm
Fig. 1. Proposed Overall System Topology
IFAC CPES 2018
Tokyo, Japan, September 4-6, 2018
488 Marven E. Jabian et al. / IFAC PapersOnLine 51-28 (2018) 486–491

such appliance is being deloaded at a specified time. The total 2.3 Optimization in Appliance Controller K
number of K is DU determined, taking into consideration the
After each K𝑗𝑗 has received its individual power allocation, K𝑗𝑗
strategic locations of the participating consumers while the
number of X depends on the capacity of each K. The distributes the available power to each of its corresponding X𝑗𝑗𝑗𝑗
communication between K𝑗𝑗 and CS can either be wired or based on the consumer-defined appliance priority level p.
wireless or even both for a more reliable network. Starting with the top priority level, level 1, it is checked if the
power supply capacity is large enough to energize the
2.2 System Monitoring appliances at that priority level. If it is, all the appliances in
that level are allocated with power and the remaining supply
The CS keeps monitoring the DU’s system technical data such capacity is handed over to the next priority level as the
as system frequency, voltage level, total enrolled/connected following equation indicates:
loads of each K𝑗𝑗 , and the real-time total power delivered by the
𝑆𝑆𝑗𝑗1 = 𝑆𝑆𝑗𝑗𝑛𝑛𝑛𝑛𝑛𝑛 , (5)
power generating plants. In the event of system low frequency
and voltage level, CS performs a short term demand 𝑝𝑝 𝑝𝑝−1 𝑝𝑝−1 (6)
𝑆𝑆𝑗𝑗 = 𝑆𝑆𝑗𝑗 − ∑𝑚𝑚
𝑘𝑘=1 𝐿𝐿𝑗𝑗𝑗𝑗 ,
forecasting for the next 15 minutes of the connected load based
on historical data and the current load using the equation 𝑝𝑝
where 𝑆𝑆𝑗𝑗 is the power supply capacity available for the
below:
appliances at priority level p or less important under the
𝑝𝑝
𝐿𝐿𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = 𝐿𝐿𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 + (𝐿𝐿𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 × 𝐺𝐺𝑡𝑡𝑎𝑎𝑎𝑎𝑎𝑎 ), (1) controller Kj and 𝐿𝐿𝑗𝑗𝑗𝑗 indicates the total load of appliances of
priority level p owned by user k under control of Kj. If the
𝑑𝑑−𝑙𝑙 𝑑𝑑−𝑙𝑙
∑𝐷𝐷
𝐿𝐿𝑡𝑡 −𝐿𝐿𝑡𝑡−1 power supply becomes, at a certain priority level 𝑝𝑝𝑐𝑐 ,
𝑙𝑙=1
𝐿𝐿𝑑𝑑−𝑙𝑙 (2) 𝑝𝑝 𝑝𝑝𝑐𝑐
𝐺𝐺𝑡𝑡𝑎𝑎𝑎𝑎𝑎𝑎 = 𝑡𝑡−1
, insufficient, i.e., 𝑆𝑆𝑗𝑗 𝑐𝑐 < ∑𝑚𝑚
𝑘𝑘=1 𝐿𝐿𝑗𝑗𝑗𝑗 , then the first optimization
𝐷𝐷
using GA is executed to determine which set of enrolled
where, 𝐿𝐿𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 is the forecasted demand after 15 minutes in appliances of priority level pc is a candidate to be turned-off or
MegaWatts (MW), 𝐿𝐿𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 is the actual total connected load deloaded. The GA minimizes the objective function (7) below
power in MW, 𝐺𝐺𝑡𝑡𝑎𝑎𝑎𝑎𝑎𝑎 is the average load power increase, t whose decision variables are defined by (8),
denotes the data logging time interval, 𝐿𝐿𝑑𝑑−𝑙𝑙
𝑡𝑡 is the demand 𝑝𝑝
𝑆𝑆𝑗𝑗 𝑐𝑐 − ∑𝑚𝑚
𝑝𝑝
𝑘𝑘=1 𝑥𝑥𝑗𝑗𝑗𝑗 𝐿𝐿𝑗𝑗𝑗𝑗 ,
𝑐𝑐 𝑐𝑐 𝑝𝑝
(7)
power at time t on day 𝑑𝑑 − 𝑙𝑙 in MW, d is the current day, and
D is the number of available historical data.
𝑝𝑝 1, 𝑜𝑜𝑜𝑜
Given the forecasted power demand, the system checks if the 𝑥𝑥𝑗𝑗𝑗𝑗𝑐𝑐 = {
0, 𝑜𝑜𝑜𝑜𝑜𝑜
. (8)
frequency will still be within the allowable range using Subject to,
equation (3) (Elrayyah, Cingoz and Sozer, 2017) as follows: 𝑝𝑝 𝑝𝑝 𝑝𝑝
𝑆𝑆𝑗𝑗 𝑐𝑐 ≥ ∑𝑚𝑚
𝑘𝑘=1 𝑥𝑥𝑗𝑗𝑗𝑗 𝐿𝐿𝑗𝑗𝑗𝑗 ,
𝑐𝑐 𝑐𝑐 (9)
𝜔𝜔𝑠𝑠 = 𝜔𝜔𝑟𝑟𝑟𝑟𝑟𝑟 − 𝑍𝑍𝑃𝑃𝐿𝐿 , (3)
until the following stopping condition holds,
where, 𝜔𝜔𝑠𝑠 is system angular frequency in rad/sec, 𝜔𝜔𝑟𝑟𝑟𝑟𝑟𝑟 the
𝑝𝑝 𝑝𝑝 𝑝𝑝 𝑝𝑝
reference angular frequency in rad/sec, 𝑃𝑃𝐿𝐿 is the forecasted 𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 = 𝑆𝑆𝑗𝑗 𝑐𝑐 − ∑𝑚𝑚
𝑘𝑘=1 𝑥𝑥𝑗𝑗𝑗𝑗 𝐿𝐿𝑗𝑗𝑗𝑗 < 𝐿𝐿𝑗𝑗𝑘𝑘𝑚𝑚𝑚𝑚𝑚𝑚 ,
𝑐𝑐 𝑐𝑐 𝑐𝑐 (10)
demand 𝐿𝐿𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 , and Z is a frequency droop coefficient and 𝑝𝑝
where 𝑥𝑥𝑗𝑗𝑗𝑗𝑐𝑐 represents the switching operation of the appliance
is dependent on the actual connected load which varies for
at priority level pc owned by user k under control by Kj,
every DU. When the frequency is forecasted to be within the
range, CS will do nothing. However, when it is not the case, 𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 is the remaining power supply capacity after
𝑝𝑝 𝑝𝑝 𝑝𝑝𝑐𝑐
the following process will be activated. The new power optimization, and 𝐿𝐿𝑗𝑗𝑘𝑘𝑐𝑐𝑚𝑚𝑚𝑚𝑚𝑚 is the smallest among 𝐿𝐿𝑗𝑗1𝑐𝑐 , … , 𝐿𝐿𝑗𝑗𝑗𝑗 .
allocation for each Kj is determined by:
In actual scenarios, consumers can have identical devices with
𝑆𝑆 𝑛𝑛𝑛𝑛𝑛𝑛 = 𝑃𝑃 × 𝐿𝐿 , (4) the same power rating enrolled in the DLC program. Hence,
𝑗𝑗 𝑅𝑅 𝑗𝑗
there is a possibility that multiple optimal solutions are found.
where, 𝐿𝐿𝑗𝑗 is the total connected load for K𝑗𝑗 in all priority
Also, during the switching implementation there is a
levels, 𝑆𝑆𝑗𝑗𝑛𝑛𝑛𝑛𝑛𝑛 is the new total power supply to be allocated to possibility that a certain consumer experiences a biased
K𝑗𝑗 , and 𝑃𝑃𝑅𝑅 denotes the ratio of the total available power supply implementation or repeated deloading of the same appliance
S𝐺𝐺 provided by the transmission grid operator to the forecasted over a certain time. To avoid the biased and unfair switching
total load 𝐿𝐿𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 . operations, a fairness function below is imposed on the
solutions to evaluate fairness:
𝑚𝑚
𝑝𝑝 𝑝𝑝 𝑝𝑝 𝑂𝑂𝑂𝑂
𝐹𝐹𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 (𝑥𝑥𝑗𝑗1𝑐𝑐 , … , 𝑥𝑥𝑗𝑗𝑗𝑗𝑐𝑐 ) = 𝑎𝑎 ∑ 𝑥𝑥𝑗𝑗𝑗𝑗𝑐𝑐 𝑅𝑅𝑗𝑗𝑗𝑗,𝑝𝑝𝑐𝑐
𝑘𝑘=1

488
 IFAC CPES 2018
Tokyo, Japan, September 4-6, 2018
Marven E. Jabian et al. / IFAC PapersOnLine 51-28 (2018) 486–491 489

+ 𝑏𝑏 ∙ 𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 , (11) a DU-specific system power rated values, we use 2.36x10 -8

Hz/W as the frequency droop coefficient Z based on a Matlab
𝑂𝑂𝑂𝑂
𝑂𝑂𝑂𝑂 𝑛𝑛𝑗𝑗𝑗𝑗,𝑝𝑝
(12) simulation. The appliances used are the common loads of
𝑅𝑅𝑗𝑗𝑗𝑗,𝑝𝑝 = 𝑐𝑐
,
𝑐𝑐 𝑂𝑂𝑂𝑂
𝑛𝑛𝑗𝑗𝑗𝑗,𝑝𝑝 𝑐𝑐
𝑂𝑂𝑂𝑂𝑂𝑂
+𝑛𝑛𝑗𝑗𝑗𝑗,𝑝𝑝 𝑐𝑐
residential and commercial establishments, and their rated
value are listed in a local DU website. Based on this list, the
𝑂𝑂𝑂𝑂
where, 𝑅𝑅𝑗𝑗𝑗𝑗,𝑝𝑝𝑐𝑐
is the ratio of the number of turning-on appliance’s power rating and its corresponding priority levels
operations to the total number of switching operations of the p are randomly assigned to each Xjk. Figure 2 shows a sample
consumer k’s appliance of priority level pc over a certain past of randomly selected appliance ratings with priority level 4. It
𝑂𝑂𝑂𝑂
period, 𝑛𝑛𝑗𝑗𝑗𝑗 ,𝑝𝑝 is the number of turning-on switching operations is assumed that the power generation is less than the DUs
𝑐𝑐
of the appliance over the period, 𝑛𝑛𝑗𝑗𝑗𝑗,𝑝𝑝 𝑂𝑂𝑂𝑂𝑂𝑂
is the number of actual connected demand and load shedding needs to be
𝑐𝑐
initiated. In the figure, a few appliances have similar power
turning-off switching operations, a and b are weight constants
ratings indicating that some consumers use such appliances at
to be determined by the DU. The value of the first term in (11)
the same time, which is a possibility in practical scenarios.
is high when appliances that have been on are again decided to
be on leaving those which have been off still kept off. The
weights balance between the ‘unfairness’ and the quality of
supply capacity application 𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 . The smaller 𝐹𝐹𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓
is, the better the solution is. In case of multiple solutions
having the least fairness value, a roulette wheel selection is
used to select a single solution. Once the unique solution is
obtained, the appliance with the minimum rated power
consumption among those which are to be turned off is found
and nominated as a candidate for possible turning-on in the
next step. Let its rated power consumption be denoted by
Fig. 2. Appliance power rating at priority level 4
𝐿𝐿𝑗𝑗𝑘𝑘𝑛𝑛𝑛𝑛𝑛𝑛 . The sum of unallocated supply capacity 𝑆𝑆𝑇𝑇𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 =
∑𝑛𝑛𝑗𝑗=1 𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 is reallocated to the controllers K. 3.2 GA Population Initialization

2.4 Optimization in Central Station CS Following the usual GA population initialization which is
random in nature, we encountered convergence issues after
The second GA maximizes the power utilization by using the several test runs. From these runs, it was observed that if the
following equations, where equation (13) is the objective 𝑝𝑝
available power supply 𝑆𝑆𝑗𝑗 𝑐𝑐 is below 10% or above 90 % of the
function: 𝑝𝑝
total load power rating (𝐿𝐿𝑗𝑗 𝑐𝑐 = ∑𝑚𝑚
𝑝𝑝
𝑘𝑘=1 𝐿𝐿𝑗𝑗𝑗𝑗 ), GA was unable to
𝑐𝑐

𝑆𝑆𝑇𝑇𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 − ∑𝑛𝑛𝑗𝑗=1 𝑥𝑥𝑗𝑗𝑐𝑐𝑐𝑐 𝐿𝐿𝑗𝑗𝑘𝑘𝑛𝑛𝑛𝑛𝑛𝑛 , (13) come up with a solution in 100 generations. To speed up the
convergence and facilitate the real-time applications, we use
1, 𝑜𝑜𝑜𝑜 different population initialization strategies depending on the
𝑥𝑥𝑗𝑗𝑐𝑐𝑐𝑐 = {
0, 𝑜𝑜𝑜𝑜𝑜𝑜
, (14) 𝑝𝑝
relationships between the available power supply 𝑆𝑆𝑗𝑗 𝑐𝑐 and the
𝑝𝑝
and the constraint is, total target load𝐿𝐿𝑗𝑗 𝑐𝑐 as followsǣ
𝑝𝑝 𝑝𝑝 𝑝𝑝
(0.9 × 𝐿𝐿𝑗𝑗 𝑐𝑐 ) ≤ 𝑆𝑆𝑗𝑗 𝑐𝑐 < 𝐿𝐿𝑗𝑗 𝑐𝑐 , (16)
𝑛𝑛
𝑆𝑆𝑇𝑇𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 ≥∑ 𝑥𝑥𝑗𝑗𝑐𝑐𝑐𝑐 𝐿𝐿𝑗𝑗𝑘𝑘𝑛𝑛𝑛𝑛𝑛𝑛 . (15) 𝑝𝑝 𝑝𝑝 𝑝𝑝
𝑗𝑗=1 (0.1 × 𝐿𝐿𝑗𝑗 𝑐𝑐 ) < 𝑆𝑆𝑗𝑗 𝑐𝑐 < (0.9 × 𝐿𝐿𝑗𝑗 𝑐𝑐 ), (17)
𝑝𝑝 𝑝𝑝 𝑝𝑝
𝐿𝐿𝑗𝑗𝑘𝑘𝑐𝑐𝑚𝑚𝑚𝑚𝑚𝑚 < 𝑆𝑆𝑗𝑗 𝑐𝑐 ≤ (0.1 × 𝐿𝐿𝑗𝑗 𝑐𝑐 )Ǥ (18)
After the execution of this final GA, CS communicates the
optimized power reallocation to each Kj. This is done to make When the condition expressed by equation (16) holds, it is
sure that the total unallocated power is close to zero. expected that the solution contains the majority of switching-
on actions, which corresponds the majority of gene values
3. CASE STUDY equal to 1 since there is nearly sufficient power to distribute
among the connected appliances. When this condition holds,
3.1 DU Model
each individual in the initial population is generated so that its
To evaluate the effectiveness of the proposed method, we 90% of genes have the value 1. When equation (17) is satisfied,
apply the method to the model of a local DU in Mindanao, the GA population is randomly initialized with equal
Philippines. In this case study, the DU has approximately probabilities for 0s and 1s. Furthermore, in the case where
20MW total power demand and 130 installed Ks, each with equation (18) holds, the initial solutions should have
approximately 150kW total connected load. Also, considering approximately 10% chance to have a value of 1 or mostly have
values equal to 0.

489
IFAC CPES 2018
Tokyo, Japan, September 4-6, 2018
490 Marven E. Jabian et al. / IFAC PapersOnLine 51-28 (2018) 486–491

3.3 Appliance Controller K
After several simulations, it is found that the following GA
parameters produced considerable results: the number of
generations = 100, population size = 50, the number of genes
or the number of consumer appliances connected = 50, the
probability of mutation = 35%, the probability of crossover =
40%, and the weight constants a and b for ‘fairness’ evaluation
are set to 1. Bit inversion and order changing are the two
mutation methods used, and for crossover methods, single
point crossover, double point crossover and uniform point
crossover are utilized. A mutation and a crossover methods are
randomly selected and applied in every generation.
Figure 3 shows a sample optimization result in one K. The total
enrolled/connected load for this specific K is 145.844kW
based on the randomly selected appliance ratings mentioned in
the previous section. From the forecasted total power demand,
we assumed that only 90kW was allocated for this particular
K. All the load with the top priority level, level 1, was allocated
with necessary power, and similarly load of levels 2 and 3 was
determined to be energized, which left the supply capacity of
2666W for the priority level 4 load that required 34717W in
total. Here the GA optimization was executed and the solution
was found at the 61st generation or after 51.75ms
(milliseconds) of CPU time, which left only 2W of unallocated
power supply. The priority level 5 load was totally left off.
Figure 5 shows the resulting fairness values for each of the
generated solution (combination of switching statuses of 50
appliances). It has been observed that the 8 th combination has
the least/best fairness value hence it is selected as the final
appliance combination for this specific K.
Fig. 5. Fairness values of the switching combinations
3.4 Central Station (CS) Controller
After determining the switching statuses of appliances in each
Kj, there is corresponding 𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 or the remaining power
allocated to Kj but are not allocated to the appliances under its
control since it is not an optimal choice to further allocate. The
second stage of GA optimization is based on the sum of
𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 reported by each Kj along with the appliance with
the minimum power rating 𝐿𝐿𝑗𝑗𝑘𝑘𝑛𝑛𝑛𝑛𝑛𝑛 , and its corresponding
𝑂𝑂𝑂𝑂
recorded switching activities 𝑅𝑅𝑗𝑗𝑘𝑘 𝑛𝑛𝑛𝑛𝑛𝑛
. In this case study, GA
parameters are set as follows: the number of genes or the total
number of installed K = 130, population size = 50, maximum
generation = 1000, crossover method = single point crossover,
mutation method = bit inversion, and the weight constants 𝑎𝑎
and 𝑏𝑏 for ‘fairness’ evaluation are set to 1. Figure 6 shows the
assumed unallocated power 𝑆𝑆𝑗𝑗𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 ”‡’‘”–‡†„›‡ƒ…ŠŒǤ

Fig. 3. Power allocation in K

Figure 4 shows the optimization process over 100 generations.
In each iteration, a minimum value is saved and is updated
until the 100th generation. The solutions that qualifies equation
(10) are the candidates for fairness function evaluation.

Fig. 6. Reported unallocated power from each Kj

After the GA process, there are three switching combinations
obtained and evaluated by the fairness function as shown in
Fig. 7. The result shows that the 2nd switching combination is
the optimum combination of appliances for additional
switching implementation to be communicated to each
corresponding Kj to update its initially found solution during
each Kj’s optimization process.
Fig. 4. Optimization process for 100th generation

490
 IFAC CPES 2018
Tokyo, Japan, September 4-6, 2018
Marven E. Jabian et al. / IFAC PapersOnLine 51-28 (2018) 486–491 491
Fig. 7. Fairness values of the switching combinations
Figure 8 shows the result of the optimization process using the
2nd switching appliance combination. For the combined
2618W unallocated power 𝑆𝑆𝑇𝑇𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 , a total of 2617W load
power is allocated out of 3590W total load nominated from
each Kj. So the final remaining unallocated power is 1W after
30.1ms of CPU execution time.
Fig. 8. Appliance power reallocation
4. CONCLUSION
In this paper, we proposed a method for real time applications
to automate load shedding that gives consumers control over
which appliances to deload by letting them set the appliance
priority levels and avoids total blackouts while maximizing the
utilization of the available power supply.
As shown in the case study, the proposed method presents
promising outcomes which can be considered suitable for real-
time application with substantial accuracy. The method was
able to find optimum combinations of switching operations for
each appliance controller quickly and accurately through two
separate genetic algorithm executions, considering the varying
power requirements and involving large number of consumer
appliances, each with corresponding priority level and a power
rating. As observed, power allocations were distributed
primarily based on consumer defined priority settings, giving
the consumers some degree of liberty to decide which
appliances they can readily deload and not caught by surprise
deloading. Also, by the implementation of the proposed
method, the utilization of available supply from the grid is
maximized and shutting down of the whole substation feeder
can be avoided. In addition, the developed fairness function
ensures that no same appliance is repeatedly deloaded at a
certain time, to avoid biased load shedding operations. The
final unallocated power is insignificant in actual scenario
491
considering the total DU demand and the available power
supply from the main grid. Thus the proposed method was able
to maximize utilization of the available power supply
signifying that more consumer appliance have power
allocation. As a result, DUs will have an automated and
efficient load shedding process which improves system
stability, reliability and power quality.
5. ACKNOWLEDGEMENT
The researchers would like to thank and acknowledge the
contribution of the Republic of the Philippines, Department of
Science and Technology, Engineering Research for
Development and Technology (DOST-ERDT) for the
scholarship support.
REFERENCES
Dandan Zhang, Wenbo Li, Zhe Jiang, Naihu Wu and Yanwen
Hou (2016) ‘Load shedding strategy coordinate optimization
based on sensitivity analysis’, in 2016 IEEE PES Asia-Pacific
Power and Energy Engineering Conference (APPEEC). IEEE,
pp. 1893–1897.
DOE Report (2016) 2016 Philippine Power Situation Report |
DOE | Department of Energy Portal. Available at:
https://www.doe.gov.ph/electric-power/2016-philippine-
power-situation-report-0 (Accessed: 17 February 2018).
DOE Report (2017) ‘Power Supply and Demand Highlights
Total Non-coincidental Peak Demand, Philippines in MW’.
Available at: http://www.doe.gov.ph (Accessed: 21 March
2018).
Elrayyah, A., Cingoz, F. and Sozer, Y. (2017) ‘Smart Loads
Management Using Droop-Based Control in Integrated
Microgrid Systems’, IEEE Journal of Emerging and Selected
Topics in Power Electronics, 5(3), pp. 1142–1153.
EPIRA (2001) ‘Rules and Regulations To Implement’, (9136).
Available at: http://www.transco.ph/downloads/EPIRA-
IRR.pdf.
ERC (2001) ‘Philippine Grid Code’, (December). Available
at: https://www.doe.gov.ph/sites/default/files/pdf/downloads/
final_grid_code.pdf.
Liu, T., Hill, D. J. and Zhang, C. (2016) ‘Non-Disruptive
Load-Side Control for Frequency Regulation in Power
Systems’, IEEE Transactions on Smart Grid, 7(4), pp. 2142–
2153.
Mortaji, H., Ow Siew Hock, Moghavvemi, M. and Almurib,
H. A. F. (2016) ‘Smart grid demand response management
using internet of things for load shedding and smart-direct load
control’, in 2016 IEEE Industry Applications Society Annual
Meeting. IEEE, pp. 1–7.
Uddin, M. N., Abido, M. A. and Rahman, M. A. (2003) ‘Real-
time performance evaluation of a genetic algorithm based
fuzzy logic controller for IPM motor drives’, in 38th IAS
Annual Meeting on Conference Record of the Industry
Applications Conference, 2003. IEEE, pp. 731–737.