2020 Fifteenth International Conference on Ecological Vehicles and Renewable Energies (EVER)

Photovoltaic System Sizing for Reliability

Improvement in an unreliable Power Distribution
System
Fitsum Salehu Kebede, Salvy Bouyguet, Jean-Christophe Olivier
IREENA-Universite de Nantes
37 Bd de l'université, 44600 Saint-Nazaire, France
Email: {fitsum.kebede, salvy.bourguet, jean-christophe.olivier}@univ-nantes.fr

Abstract—This paper proposes a photovoltaic-battery energy system with sufficient security and quality of
storage system (PV-BSS) to improve the reliability of a supply.
power distribution network with frequent outages, Generally, about 90% of the electrical power
encountered in many developing countries. The system distribution supply service interruption experienced by
sizing methodology takes into account the loss of power
customers is attributed to distribution systems. The causes
supply probability (LPSP) and the cost of energy. A
genetic algorithm is used for achieving improved reliability
of power outages can either be unplanned or planned.
with an optimal cost of energy by determining the Pareto Component failures, operational faults, human and other
solutions set. A detailed outage survey characterized by the natural errors contribute to the unplanned outages.
number and duration of outages experienced by a typical However, the planned ones are performed for
customer on a selected weak distribution system is maintenance and operation [2].
examined. For this study, real load demand and outage In an effort to seeking options for enhanced power
data have been recorded for a university building in supply like in [3], [4], grid outage prediction and
Ethiopia. Moreover, long-term outage data was analyzed forecasting are presented believed to support enhancing
and studied to understand the overall outage situation, the grid management system and reducing outage events.
which clearly showed its randomness. The proposed PV-
The authors reported that the prediction can help in
BSS helps to reduce the loss of power and achieve a
reliable power supply with zero LPSP for the case
informed energy resources scheduling and outage
considered. Currently, without any standby and backup handling in severe weather conditions. Also in [5], a
system, the building has a poor power supply with an probabilistic estimation of a failing distribution
erratic outage of about 6% LPSP. component is used to predict outage, consequently
helping for prioritized power restoration for critical
Keywords—Outage; reliability; power distribution customers.
system; loss of power supply probability (LPSP); PV- PV systems integrated with battery storage, for
battery. various applications, have been explored by many
researchers. Their various objectives include mainly the
I. INTRODUCTION reduction of utility energy bills and CO2 emission, peak
Recurrent power interruption from distribution demand shaving, improve energy mix, grid voltage
networks is overwhelmingly becoming an everyday stability and supply electric power for off-grid areas.
problem in developing countries [1]. It is even more For instance, in [6], a PV-Battery system is presented
serious due to the fact that most interruption occurs out of for reliability improvement in weak grid supply. Hybrid
the control of utility operators. Likewise, these PV-Wind-Battery system can also help to support a
distribution systems are weak due to the common use of deficiency of power due to seasonal effects [7].
radial arrangement and aged components [2]. For Similarly, hybrid systems are useful in supplying power
instance, in Ethiopia, we observed since the last decades demands of off-grid standalone applications [8].
an increase in the number of interruption. The interruption On the other hand, planned grid outage alleviation
while lasting more often from few hours to several days in through power management system with distributed
worst cases, greatly affects a wide range of consumers generation consisting of Photovoltaic system, diesel and
from households to large businesses leading to the use of battery is proposed in [9].
diesel generators (DG). The associated emission, cost of In this paper, a hybrid PV-battery system is considered
diesel, O&M, etc. in turn calls for a renewable source of to improve the reliability of a weak power grid. Our main

978-1-7281-5641-5/20/$31.00©2020 European Union

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.
 objective is the optimal sizing of this power system, outgoing 15kV feeders. It has two 25/31.5MVA and one
considering a techno-economic approach. 12MVA transformers [12].
The rest of the paper is structured as follows: Section Fig. 1 shows the outage record and its severity from
II illustrates and discusses the distribution system 2012 to 2016 [13]. For instance, the duration and number
topology and outage profile survey in Ethiopia. Section III of outages grew by more than 150% and 23% from 2012
describes the building’s case study and the proposed to 2013 respectively. In contrast, the outage trend has
system mathematical models followed by the system’s shown a decrease from 2015 to 2016 by nearly 32% and
economic model in section IV. In section V, the system 17%. Likewise, the five-year data show that an average of
simulation results are presented with a conclusion in the 52 minutes outage duration per event was registered.
last section. The causes of outage are categorized as listed below:
• Earth fault (EF): which includes both permanent
II. POWER SECTOR IN ETHIOPIA
and temporary faults,
Ethiopia’s energy sector, currently with an installed • Overload (OL): either on the distribution line,
electricity generating capacity of more than 4,200MW is transformer or system overload,
experiencing a shortage of power [10]. The major causes
include a highly unreliable electrical network due to • Operational (OPE): an intended interruption of
interruption or outage and huge electricity annual demand power for maintenance, load transfer, and expansion
growth of about 30% [11]. Its economic development, the or upgrading works,
transition from agricultural to an industrial economy, • Short circuit (SC), and
population growth and urbanization are some of the • Under frequency (UF).
factors in raising this modern energy demand. EF 51 min./event
2,000 61 min./event
Nowadays in Ethiopia, even in urban areas like its 1,800
OL 67 min./event
OPE
capital Addis Ababa, the electric power interruptions are
Outage Duration (hr.)

1,600 SC
increasingly common. Such incidents are leaving lots of 1,400 UF 50 min./event
1,200
customers without power for a few minutes up to several 1,000
hours and days in a single interruption event. In this 800 33 min./event
context, many customers installed backup DGs to 600
400
maintain the supply during the outages. These systems 200
imply high maintenance and operating cost and the use of 0
2012 2013 2014 2015 2016
non-environmental friendly diesel groups. The objective Year

of this work is to propose an alternative and more Fig. 1: Causes contributions to the outage durations (top) and number
ecological way to improve the grid reliability. (bottom) for the 2012 – 2016 records
300 250
A. Overview of Addis Ababa Electrical Distribution Duration
Number
Network 250
200

Mainly the city is interconnected with 400kV, 230kV 200

150
and 132kV transmission network with a sub-transmission 150
system of 66kV, 45kV, and 33kV lines. Moreover, the 100
100
15kV distribution system has 28 Bulk Supply Points
50
(BSP) or substations [12]. Customer level voltages are 50

380V for three-phase and 220V for a single phase. The 0 0

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
distribution system has an overhead radial topology, Month
lacking sufficient redundancy in case of outage of major Fig. 2: Monthly numbers of outages and durations in the year 2016
circuit or transformer. In this paper, the outage number
and duration is surveyed for a particular highly Fig. 1 also shows the contributions of the causes for
unreliable 132/15kV electrical distribution substation. the outage in the same period. On top of the bar, the
average outage duration sustaining in a single event (i.e.
B. Outage Study of Addis Center Distribution Substation minutes per one outage event) is shown. In all the years,
Addis Center substation (ADC) is one of the BSPs the operational outage has the highest durations and also
with wide area coverage and large number of customers. the highest number of events.
It is also known to experience frequent outage. It is Besides, the monthly outage durations and number of
supplied by a single 132kV overhead line with 11 outage for the year of 2016 is shown in Fig. 2. Both the
outage duration and number experienced a drastic

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.
decrease between January and February. Apparently, the operating in islanded mode from the grid. The core energy
relatively high values from June to September could be management system (EMS) is responsible for energy
due to the rainy season. conversions and dispatches.
Further outage records on one of the feeders (ADC-9) S
emanating from ADC showed relatively higher outages. Grid

Fig. 3 shows the total monthly outage duration and the Surplus
corresponding average load for the year 2016. To give an Interruption
economic perspective during the longer outage duration in Battery
June, the total energy not supplied (ENS) is about EMS
264MWh. Considering an average electricity sales price PV array
of 0.0321€/kWh [14], it leads to an approximate monthly
sales revenue loss of about €8,500 for this single feeder. It
Building
thus seems obvious that these outages have a significant Load
impact on the economy and solutions must be considered.
Fig. 4: Energy flow simplified schematic for the PV-BSS

A. Load Measurement and Outage Recording

Fluke 435 Series II three-phase power quality and
energy analyzer is used to measure the load demand and
record the outage events. The device is installed at one of
AAiT’s buildings with regular functions like offices,
classrooms, laboratories and conference rooms. Even with
the standby generator, the power supply experiences
interruption dips due to total outage when the diesel is not
yet supplying power (due to the time to switch between
Fig. 3: Feeder ADC-9 monthly total outage duration and monthly the grid and the generator). This can also occur when
average load either the generator is out of fuel or faulty ATS. Based on
C. Customer-based Outage Survey: A case for Addis
these sequences of loss of power, it becomes possible to
Ababa Institute of Technology automatically detect the start and end of an outage. The
data measurement was taken for about 27 days from
Further outage and load profile case study has been October to November 2018. To capture the power supply
made on a university campus located in Addis Ababa. The switching between the grid and generator, the time step
institute’s electrical grid power is supplied from the sole was set to one second. After that, for the sizing analysis, a
government-owned power utility company. It has also minutely averaged load and outage data are synthesized to
installed a 650kVA standby diesel generator to sustain perform the energy systems modeling.
outages. An Automatic Transfer Switch (ATS) will turn Fig. 5 illustrates the power profile for one day, on the
on the generator when the outage is detected and turn off grid side. During grid outages, the power is null and the
if the grid is back. diesel group ensures the load demand. It must be noted
III. METHODS AND MATERIALS that the outages 1, 3 and 4 are long enough to be seen on
the figure. For outage 2, its duration is only 3 minutes and
We consider a photovoltaic power system in parallel it is not obvious to see it on the figure.
with the grid. Battery storage is also included for added 50
Durations:
degree of freedom to the system in the energy Outage 1 = 414 minutes
Grid Power Supply (kW)

40 Outage 2 = 3 minutes
management, which also increase the system flexibility Outage 3 = 110 minutes
for outage reductions. 30
Outage 4 = 347 minutes
with Outages

The general energy system schematic for the

considered PV-BSS is as depicted below in Fig. 4. During 20

when there is an active grid, isolator switch S is closed, 10

Outage 1 Outage 3

whereby the energy flow is represented by the black Outage 2 Outage 4

arrows. Apart from this when the battery is fully charged, 0

0 2 4 6 8 10 12 14 16 18 20 22 24
surplus PV power can be fed to the grid (green arrow). Time (hr.)
Conversely, if there is grid interruption, S will be opened. Fig. 5: A typical-day grid supply scenario with outages (for 31-Oct-
Hence the system follows the blue arrows, thereby 2018)

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.
 Week Days Weekend Days data for the location [16] is used with a long term annual
average irradiance from 4.5 to 6.5kWh/m2/day and yearly
50
48.4 kW
mean ambient temperature of 16°C. In fact, the seasonal
45 variation is insignificant but still contributes to the lower
40 radiation during the rainy season from June to August and
Load Demand (kW)

35 higher contributions for the rest of the sunny months. The

30 solar radiation and temperature are both used in (2) to
25 24.4 kW estimate the efficiency of the photovoltaic panels. A time
20 step of 1 minute is used for the solar radiation profile.
15
13.3 kW C. Battery Model
10
Nov 12 Nov 13 Nov 14 Nov 15 Nov 16 Nov 17 Nov 18 A key device for renewable power systems is the
Time (day)
storage, which is here connected to the grid through a
Fig. 6: One-week total average load/power demand of the building in power electronic converter (see Fig. 4).
kW (from Nov 12 to Nov 18, 2018) Because this paper only addresses the problem of
Fig. 6 shows the load profile for one week and with a system sizing and global energy management, self-
time step of one minute. Because the building is a discharging and temperature effects of the battery storage
university campus, it can be seen that the consumption is are not considered. Moreover, the charging and
higher during the first five days (from Monday to Friday) discharging battery efficiencies are taken to be a
than during the weekend. The profile is on the load side, manufacturer’s roundtrip efficiency.
for which the power comes from the grid or diesel. The battery modeling in this work is based on its two
operational modes either as charging or discharging. The
B. Photovoltaic Sizing Mathematical Models
sizing rule is governed by comparing the sum of the grid
The solar power that can be generated by the PV is and PV power with that of the load demand. This relation
mathematically modeled for maximum power point is defined by the battery State of Charge (SOC) at any
tracking by (1) as used in [7], [8]: time instant. For our case, a 1 minute time step is used for
( , )= ( , ) (1) the simulation of the system.
The first one is the charging mode: For this case, if the
where Ppv is the PV power output, Tamb is the ambient total production is higher than the load and the battery is
temperature, pv the PV panel efficiency, Apv is the PV not yet fully charged, the battery will be charged as given
array area and Girr is the global solar irradiance hitting the below by (5) with the power available from the PV and
PV surface. The efficiency of the PV is dependent on the grid, a maximum subscribed grid power of 50kW is
PV cell temperature and the irradiance, and hence is given assumed whenever there is no outage [7]. Hence if,
by (2) [8]. ( )≥ ( )⁄ where Ptot is the total power as
in (4) coming from the grid Pg and the PV array Ppv with n
( , )= 1− − (2) discrete times of simulation:
( )= ( )+ ( ) (4)
= − , (3)
Then, charging mode is used, and the SOC is given by:
where Tcell, TNOCT, and Tamb, NOCT are the PV cell ( )
temperature, the Nominal Cell Operating Temperature ( )= ( −1 ) + ( )− _
∆ (5)
(NOCT) and the ambient temperature for the NOCT
conditions respectively. where inv is inverter efficiency, bat_cha is the battery
The PV module from LG electronics [15] with the charging efficiency and Cbat is the battery bank capacity in
following parameters is considered in this paper: Wh.
reference temperature Tref = 25°C, TNOCT = 44±3°C, Tamb, The second mode is the storage discharge: This mode
NOCT = 20°C, manufacturer’s module efficiency pv = appears once the total need of the load is higher than the
0.203%, the power temperature coefficient, pv = - produced PV electricity and grid outage. In this case, the
0.3%/°C and the NOCT irradiance, GNOCT = 800W/m2. batteries are used to ensure all the load power. That is
To have the most realistic model results, experimental when ( )≤ ( )⁄ , the SOC variation is
data with the 27 days of load profile incorporating the given by:
outages presented in section A are used. Global solar ( )
( )= ( )+ − ( ) ∆ (6)
irradiance and ambient temperature actual measurement _

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.
which is an Euler approximation and where bat_discha is the = ,
(9)
battery discharge efficiency. The SOC of the battery is
subjected to the following operational constraints at which
where, the annualized total cost, Ctot,a in [€] is given as,
the battery will be operating between the minimum and
maximum allowable charge states. , = + + × + , (10)
≥ ( )≥
= + = × + × (11)
where the SOCmax and SOCmin are the maximum and
minimum SOC. In addition, the maximum charging and where Cpv and Cbat are respectively the PV and battery
discharging power also govern the battery charge- capital costs [€] per unit PV area Apv and battery capacity
discharge. In this modeling, a Lithium-Ion battery from Bcap, CRF is the capital recovery factor and Ccha,g is
[15] is considered. The technical data of these batteries charging electricity cost from the grid. Also, CCtot is the
are Cbat = 7kWh, a maximum charging and discharging total capital investment cost [€] and OMCpv is the total
power of 3.5kW and 7kW, respectively, a maximum SOC annual operation and maintenance cost [€] with an O&M
= 0.9, a DOD of 0.8 and a roundtrip efficiency bat = 0.97 cost fraction of the capital cost αpv [%/m2/year]. The
are taken. OMCpv can then be found by [18]:
Now, the first step to define the reliability of the ×
system is to calculate the Loss of Power Supply (LPS), =∑ ( )
(12)
the power that cannot be met by the PV-battery bank
system. If SOC(tn) < SOCmin, therefore, SOC(tn) is set to Here, noting that the operation and maintenance cost
SOCmin and, the Loss of Power (LPS) is calculated as [6], for the battery storage system is neglected, the battery
[7]: replacement cost RCbat [€] with a lifetime of 6 years that
will have 3 replacement occurrences through the life of
= − +( − ) (7) the project is included. Hence, it is given by [18]:
_ ∆

From the LPS, a satisfactory index of the sizing can be =∑ ( ) ×

(13)
built. It is a well-known reliability index indicator, named
Loss of Power Supply Probability (LPSP) [7], [8], which where the replacement cost is equal to the cost of the
is the fraction of the LPS with respect to the load and is battery, and the CRF can be calculated from [18]:
given by:
( )
=( )
(14)
∑ ( )
= (8)
∑ ( )
where dr is the annual real discount rate [%], tbat and nrep
On the other hand, when the SOC(tn) is greater than are the battery lifetime and number of battery
SOCmax, the surplus energy produced considered as replacements. TABLE 1 summarizes all the economic
dumped power after meeting the load demand and the data used for this study.
battery is fully charged can be supplied to the grid. TABLE 1: PARAMETERS USED IN THIS PAPER

IV. SYSTEM ECONOMIC MODEL PV capital cost Cpv (€/m2) 200

PV annual O&M cost αpv (€/m2/year) 1% of Cpv
Battery capital cost Cbat (€/kWh) 550
For renewable energy technologies, the Cost of Energy Discount rate dr (%) 2
(CoE) is a well-adapted parameter to estimate the Lifetime of PV (years) 25
practicability to invest in energy projects. It is calculated Lifetime of battery (years) 6
Diesel generator capital cost (€/kW) 370
based on the total annualized system costs as well as the Generator O&M cost (€/h) 0.05
electricity price to charge the storage and with respect to Generator characteristics F0 (l/h/kW) 0.0271
the total demanded energy through the lifetime (here T =
25 years is considered) of the project. The annualized cost Faced to all the possible PV array area – battery
considered in this study consists of the capital, operation capacity pairs, we have used Genetic Algorithm (GA)
and maintenance, and replacement costs. optimization and identify the Pareto Front for lower LPSP
The economic parameters of the PV-BSS are given in with the optimal cost of electricity, see Fig. 7.
TABLE 1 below, which are average values from the Furthermore, the CoE for the PV-BSS is compared
literature. with the cost of energy for a diesel-only system by
Hence according to literature such as [17], [18], the considering a generator for the studied building. With a
CoE [€/kWh] for the proposed PV-BSS is given by:

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.
similar analogy as above, the cost of energy can be outage hours, the lifetime of the generator can be assumed
computed by: to be equivalent to the project life without major error.
The hours of operation are the outage hours, which is a
, , ,
CoE = (15) total of 32.93 hours during the measurement period. The
annualized cost is calculated by converting to annual
Based on the daily maximum load of 48.4kW recorded outage hours assuming the same occurrences. The current
on 12th November (see Fig. 5), a 48kW rated power diesel price of 0.559€/l in Ethiopia is used. When
generator is considered in this study. The fuel required, a currency conversion of one Euro equals to
consumption characteristic of the diesel generator is given 32.278 Ethiopian birr is used [21].
by its fuel curve and the efficiency performance [17], [18] V. RESULTS AND DISCUSSION
which are given by (16) and (17) as a function of the
generator’s output power. Apart from the electrical utility generic outage
situation, an exploratory study of real outage exposure
= + (16) alongside the load profile for a building located in
Ethiopia has been examined and presented. It is evident
×
= (17) from this investigation and data acquisition that the
× ×
customer outage is severe in which the outage duration
and occurrence experienced were about 32 hours and 29
Start events, respectively. Currently, without any backup and
standby power system, the building is found to have a
READ: Load Profile, Grid status loss of grid power supply probability of 5.91%.
(active grid or outage), PV and
battery component parameters, As stated earlier, the modeling objective targeted at a
associated costs and constraints reduced or zero loss of power is achieved by optimizing
the LPSP and CoE for all the possible combinations of
Generate initial population (Ai, Nbat )
the system’s PV array area and battery bank capacity.
Fig. 8 shows the PV-Battery pairs for desirable LPSP
values and the corresponding CoE. For exhaustive
PV-BSS model, searching, the possible combinations in a wider range of
Eqs. (1-6) PV and battery with upper bounds of 600m2 and
350kWh, respectively, are explored.
LPSP and CoE objective functions, Similarly, as can be seen from the Pareto frontier, an
Eqs. (8-9) Genetic N
operator absolute zero LPSP can be achieved with relatively high
battery capacity and PV array, but still one can see that
Fitness function evaluation beyond the 0.10% LPSP contour region, it shall be
acceptable reliability. The equivalent electricity price for
this reliability amounted to 16 c€ per kWh.
Stopping criteria?
One can note that the impact of extended outage
duration in a single event could be significant on the
Y
LPSP value, hence demanding larger system size.
Possible optimal LPSP – CoE combinations, Accordingly, the higher PV area and large capacity
Pareto front
battery banks are attributed to the higher outage duration
per event, which was recorded at 8.78 hours/outage
Stop
event. Here it should be highlighted that at higher battery
Fig. 7: Simplified chart for GA optimization sizes, the LPSP starts to hardly respond to changes in PV
size. This is due to the fact that the above-mentioned
where F is the fuel consumption at any time interval outage happened during the night time on 30th – 31st
[l/h], F0 is the fuel curve intercept coefficient [l/h/kW], F1 October where there is no solar power.
is the fuel curve slope [l/h/kW], Prated is the rated capacity,
Pgen is generated power [kW], gen is its efficiency, fuel is End users can decide to have an approach to sustain
the density [kg/m3] and LHVfuel is the lower heating value longer duration events with high investment. This can be
[kJ/kg] of diesel fuel. arbitrage between deciding to have an emergency power
A generator such as given in [19] is selected and supply or a reliable standby system either with lower or
economic parameters as given in [20] are considered. higher investment costs, respectively.
Furthermore, with a standby operation only during grid

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.
 In fact, a very small outage could have adverse VI. CONCLUSION
aftereffects for consumer groups like hospitals, research
In this paper, we carried out a PV-Battery system
laboratories, etc. which shall be better equipped with
sizing mechanism for a poor electrical distribution
reliable systems such as this. Besides, the emergency
system with frequent outages. We recorded the load
system can give an opportunity to monitor daily
demand and outage characteristics for a university
activities to save the loss of important data, for example
building located in Ethiopia as a case study. Following
in teaching and research laboratories. This is because
the outage phenomena of the distribution system in the
professors and lecturers in the institute always claim loss
country in general and particularly in the building, the
of data due to interruption of experiments and loss of
optimal sizing methodology has taken an exploration
productive lecture hours, which are among other
approach of system components sizes through a genetic
consequences of the outage. This and other crucial
algorithm based on the real load profile and grid outage
insights can be a compromise for the higher costs.
scenario. Despite a relatively high CoE compared to the
Here, with these comparable CoE for the PV-BSS diesel generator, the presented system provides greater
pairs, it is suggested that a system with large battery reliability improvement and can also serve as a basis for
capacity and small PV shall have a fairly better further comprehensive research in such a weak grid
preference as the battery can also be charged from the supply. It can also be noted that lower costs with an
grid. This is because of the cheap grid power and acceptable choice of LPSP than zero shall give the
reducing dumped surplus power. In the latter case, freedom to choose other PV-BSS combinations, for
surplus generation could be fed to the grid though users to make an informed decision in implementing
currently there is no feed-in structure in Ethiopia yet. such systems.
ACKNOWLEDGMENT
The authors gratefully thank the Ethiopian national
electric power and utility companies, and meteorological
agency for the data access.
REFERENCES
[1] World Bank Group, “Enterprise Surveys - Ethiopia 2015
Country Profile,” 2015.
[2] R. E. Brown, Electric Power Distribution Reliability. 2009.
[3] R. Eskandarpour and A. Khodaei, “Machine Learning Based
Power Grid Outage Prediction in Response to Extreme Events,”
IEEE Trans. Power Syst., vol. 32, no. 4, pp. 3315–3316, 2017,
doi: 10.1109/TPWRS.2016.2631895.
[4] R. Eskandarpour, A. Khodaei, and A. Arab, “Improving power
grid resilience through predictive outage estimation,” 2017
North Am. Power Symp. NAPS 2017, 2017, doi:
10.1109/NAPS.2017.8107262.
[5] A. A. Nanadikar, V. N. Biradar, and D. V. S. S. Siva Sarma,
Fig. 8: Loss of Power Supply Probability (LPSP) and Cost of Energy
“Improved outage prediction using asset management data and
(CoE) optimization
intelligent multiple interruption event handling with fuzzy
The CoE for ideally operated generator without delay control during extreme climatic conditions,” 2014 Int. Conf.
to achieve a zero LPSP based on (15) is 0.1054€/kWh, Smart Electr. Grid, ISEG 2014, pp. 1–7, 2015, doi:
10.1109/ISEG.2014.7005617.
this can correspond with the PV-BSS’s LPSP of 0.64%. [6] J. Khoury, R. Mbayed, G. Salloum, and E. Monmasson,
Despite this slightly low cost, in the range of 50 to 60% “Optimal sizing of a residential PV-battery backup for an
less than its PV counterparts, issues such as the emission intermittent primary energy source under realistic constraints,”
concern, instability, and rise of imported diesel price, Energy Build., vol. 105, pp. 206–216, 2015, doi:
spiking maintenance hours and cost, etc. are yet to be 10.1016/j.enbuild.2015.07.045.
considered for such investment. [7] Mohammad Masih Sediqi, Atsushi Yona, Tomonobu Senjyu,
Mohammed E. Lotfy, and Masahiro Furukakoi, “Optimal
The developed system, notably for unpredictable and Economical Sizing of Grid-Connected Hybrid Renewable
unplanned grid distribution system outages, provides Energy System,” J. Energy Power Eng., vol. 11, no. 4, pp. 244–
greater reliability improvement and also a basis for such 253, 2017, doi: 10.17265/1934-8975/2017.04.005.
a weak grid supply scenario for further comprehensive [8] S. Diaf, M. Belhamel, M. Haddadi, and A. Louche, “Technical
research. and economic assessment of hybrid photovoltaic/wind system
with battery storage in Corsica island,” Energy Policy, vol. 36,
no. 2, pp. 743–754, 2008, doi: 10.1016/j.enpol.2007.10.028.
[9] M. Hijjo, F. Felgner, and G. Frey, “Energy Management

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.
 Scheme for Buildings Subject to Planned Grid Outages,” vol. 3,
no. 3, pp. 58–65, 2016.
[10] “Ethiopia Country Commercial Guide (Ethiopia - Energy),”
2016. [Online]. Available:
https://www.export.gov/article?id=Ethiopia-Energy. [Accessed:
26-Jun-2018].
[11] Ministry of Water and Energy, “Scaling-Up Renewable Energy
Program (Ethiopia Investment Plan-Draft Final),” 2012.
[12] P. Brinckerhoff, “Addis Ababa Distribution Master Plan Study
- Modelling and Analysis of Existing Network,” 2014.
[13] Ethiopian Electric Power, “Substation Outage datalog.” p.
Unpublished raw dataset, 2018.
[14] Ministry of Water Irrigation and Electricity of Ethiopia,
“Electricity Tariff Amendment (የኤሌክትሪክ ታሪፍ ማስተካከያ, in
Amharic),” Addis Ababa, 2018.
[15] LG Electronics, “LG Neon,” 2013. [Online]. Available:
www.lg.com. [Accessed: 07-Jun-2018].
[16] National Meteorology Agency-Ethiopia, “Weather Data.” p.
Unpublished raw dataset, 2018.
[17] B. K. Das and F. Zaman, “Performance analysis of a PV/Diesel
hybrid system for a remote area in Bangladesh: Effects of
dispatch strategies, batteries, and generator selection,” Energy,
vol. 169, pp. 263–276, 2019, doi:
10.1016/j.energy.2018.12.014.
[18] HOMER Energy, “HOMER Pro microgrid software.” [Online].
Available:
https://www.homerenergy.com/products/pro/docs/3.11/index.ht
ml. [Accessed: 10-Apr-2019].
[19] Kohler Power Asia Pacific, “Diesel Generators.” [Online].
Available:
http://www.kohlerpower.sg/industrial/detail.htm?sectionNumbe
r=13261&categoryNumber=11961&filter_1=50
Hz&prodnum=35854602. [Accessed: 10-Apr-2019].
[20] O. N. Kuznetsov and H. M. Sultan, “Economic Feasibility
Analysis of PV / Wind / Diesel / Battery Isolated Microgrid for
Rural Electrification in South Egypt,” 2019 IEEE Conf. Russ.
Young Res. Electr. Electron. Eng., pp. 1001–1006, 2019.
[21] Commercial Bank of Ethiopia, “Currency Rate.” [Online].
Available:
https://www.combanketh.et/More/CurrencyRate.aspx.
[Accessed: 17-Apr-2019].

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:02:06 UTC from IEEE Xplore. Restrictions apply.