Performance and reliability analysis of an off-grid PV mini-grid system in

rural tropical Africa: A case study in southern Ethiopia

Downloaded from: https://research.chalmers.se, 2023-11-08 15:30 UTC

Citation for the original published paper (version of record):

Wassie, Y., Ahlgren, E. (2023). Performance and reliability analysis of an off-grid PV mini-grid
system in rural tropical
Africa: A case study in southern Ethiopia. Development Engineering, 8.
http://dx.doi.org/10.1016/j.deveng.2022.100106

N.B. When citing this work, cite the original published paper.

research.chalmers.se offers the possibility of retrieving research publications produced at Chalmers University of Technology.
It covers all kind of research output: articles, dissertations, conference papers, reports etc. since 2004.
research.chalmers.se is administrated and maintained by Chalmers Library

(article starts on next page)

 Development Engineering 8 (2023) 100106

Contents lists available at ScienceDirect

Development Engineering
journal homepage: www.elsevier.com/locate/deveng

Performance and reliability analysis of an off-grid PV mini-grid system in

rural tropical Africa: A case study in southern Ethiopia
Yibeltal T. Wassie, Erik O. Ahlgren *
Chalmers University of Technology; Department of Space, Earth and Environment; Division of Energy Technology, SE- 412 96, Gothenburg, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: Although some progress has been made in recent years, ensuring universal access to electricity remains a major
Off-grid PV mini-grids challenge in many countries in sub-Saharan Africa, particularly in rural areas. In light of this challenge, solar
Power generation photovoltaic (PV) mini-grid systems have emerged as a promising solution for off-grid electrification. However,
Performance ratio
little is known about their actual performance and reliability when used in real-world applications. Using real-
Load-shedding
Reliability
time monitored data and IEC’s evaluation standard, this paper examines the performance and reliability of a 375
Ethiopia kWp off-grid PV mini-grid system installed in a remote small town in Ethiopia. The findings showed that the
mini-grid produced 1182 kWh/day of electricity compared to the estimated generation of 2214 kWh/day, a
difference of 1032 kWh/day (46.6% less). In contrast, 87% of the average daily electricity generated was
delivered to the load. The discrepancies can be attributed to average PV capture losses of 2.75 kWh/kWp/day
and system losses of 0.40 kWh/kWp/day. The performance evaluation results revealed that the mini-grid system
is performing poorly, with average on-site module efficiency (ηpc), temperature corrected performance ratio
(PRcorr), capacity factor (CF) and overall system efficiency (ηsys) of 9.85%, 42%, 13%, and 8.76%, respectively. It
was found that the daily PV energy output could not meet the daily demand. As a result, the load is shed off from
the power supply for 13 h a day; between 17:00 and 19:00 and again between 21:00 and 08:00. The study
demonstrated that accurate demand assessment and robust system sizing, taking into account the impact of local
weather conditions and prospective electricity demand growth is critical to ensure high performance and reli­
ability of off-grid PV mini-grid systems.

growing body of evidence suggests that stand-alone PV mini-grid sys­

1. Introduction tems with batteries are cost-effective options to deliver affordable and
reliable electricity to remote off-grid communities in many developing
“Ensuring access to affordable, reliable, sustainable and modern countries (Moner-Girona et al., 2018), (Come-Zebra et al., 2021). These
energy for all” is at the heart of the United Nations Sustainable Devel­ systems work independent of the main grid, and deliver electricity near
opment Goals (SDG 7) (IEA, 2020). However, providing access to elec­ the point of use. In this regard, successful experiences of rural electri­
tricity remains a major challenge in many sub-Saharan African (SSA) fication using off-grid PV mini-grid systems have been documented, for
countries, particularly in rural areas. While some progress has been example, in India, Kenya, Tanzania, Nepal, and Namibia (Come-Zebra
made in recent years, more than 500 million people in SSA still live et al., 2021), (Pedersen et al., 2021). Other studies, however, have found
without basic access to electricity, most of them in rural areas (IEA, that many PV mini-grid systems installed in developing countries are
2020). The difficulty arises, in part, from the fact that extending national either underperforming; have reliability issues; have failed or have been
grids to remote and isolated villages requires huge investment owing to abandoned prematurely (Azimoh et al., 2016), (Hartvigsson et al.,
their geographic location and the high cost of transmission lines and 2018), (Katre et al., 2019), (Numminen and Lund, 2019).
sub-stations (Wassie and Adaramola, 2021). Against this backdrop, In light of this, substantial research has been conducted to evaluate
decentralized renewable energy systems such as solar photovoltaic (PV) the performance of PV power systems. For example, the German
mini-grids and micro hydropower plants have emerged as viable alter­ Development Agency (GIZ) assessed the performance of two off-grid PV
native solutions to provide electricity in off-grid areas (Vezzoli et al., mini-grid systems in rural Indonesia. The study found that the perfor­
2018). With the rapid fall in the price of solar panels in recent years, a mance ratio (the ratio of the annual measured energy output to the

* Corresponding author.
E-mail addresses: tebikew@chalmers.se (Y.T. Wassie), erik.ahlgren@chalmers.se (E.O. Ahlgren).

https://doi.org/10.1016/j.deveng.2022.100106
Received 24 June 2022; Received in revised form 14 December 2022; Accepted 26 December 2022
Available online 29 December 2022
2352-7285/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

Abbreviations MMEMS Mini-grid Monitoring and Energy Management System

MPPT Maximum power point tracker
A PV array effective area (m2) Pnom Rated power/nominal capacity of the PV array (kWp)
AC Alternating current PR Performance Ratio (%)
CF Capacity factor (%) PRcorr Temperature-corrected Performance Ratio (%)
DC Direct current PV Photovoltaic
DOD Depth of discharge SAIDI System’s Average Power Interruption Duration Index
E DC DC energy output (kWh) SOC State of charge
EAC AC energy output (kWh) SSA Sub-Saharan Africa
EAC Load Energy consumed by the load (kWh) STC Standard Test Conditions (25 ◦ C, 1 kW/m2; A.M. = 1.5)
EEU Ethiopian Electric Utility (electricity provider) UEAP Universal Electricity Access Program
ESMAP Energy Sector Management Assistance Program of the Ui Total power interruption time per day (hours)
World Bank YA Array yield (kWh/kWp)
GSTC Irradiance at standard test conditions (1, 000 W/m2) YF Final yield (kWh/kWp)
Gtotal Total solar irradiance incident on the tilted PV array (kWh/ YR Reference yield (kWh/kWp)
m2) ηpc Power conversion efficiency of PV modules under real
IEC International Electrotechnical Commission working conditions (%)
LC Capture losses ηSTC Module efficiency under STC (%)
LS System losses ηsys Overall system efficiency (%)
MG Mini-grid

estimated output based on the system name-plate rating) was below PV mini-grid systems, but nearly all of these studies focused on grid-
30% due to the high energy losses in the energy capture and conversion connected systems. As a result, not much is known about the perfor­
process (GIZ, 2017). Sharma and Chandel (2013) investigated the per­ mance, reliability and efficiency of off-grid PV mini-grids when used in
formance of a 190 kWp grid-tied PV plant in northern India; and found real-world conditions, particularly in hot tropical climates in developing
that the plant produced 98.8% of the estimated annual energy yield. The countries. This gap in knowledge and on-site practical experience can
authors reported the annual mean performance ratio and overall system hinder future development and improvement of off-grid PV mini-grid
efficiency of the plant to be 74% and 8.3%, respectively. Costa and systems as a viable alternative solution for rural electrification.
Villalva (2020) studied the performance of a PV-diesel hybrid mini-grid The main aim of this study is to investigate the actual performance,
system for the Tapajós-Arapiuns extractive reserve in Brazil, and showed efficiency and power supply reliability of a 375 kWp off-grid PV mini-
that with a ‘load-following’ dispatch strategy, 85.6% of the energy de­ grid system with energy storage batteries installed in a remote small
mand can be met by the PV system and 14.4% by the diesel generator town in Ethiopia using real-time measured weather data, and power
(DG). Using an estimated annual electricity demand of 63,875 kWh, generation and load data. The paper essentially seeks to answer the
Akinyele and Rayudu (2016) assessed the performance of different-sized following research questions.
PV micro-grids for a small village in Nigeria. They found that PV
micro-grids with a capacity of 55–82.5 kWp can deliver sufficient power ⁃ What is the efficiency of PV modules in converting available solar
to the communities. Similarly, Banguero et al. (2019) evaluated the energy to useful electrical energy under real conditions compared to
performance of a 20 kWp off-grid PV micro-grid in Chocó, Colombia, under standard test conditions?
and discovered that the micro-grid’s overall efficiency ranged from ⁃ How does the measured energy yield and performance of the mini­
10.3% to 11.09%. grid system compare with its nominal capacity, nominal efficiency
These studies, along with many others, provide insight into the and estimated/calculated energy yield?
performance of PV mini-grid systems. The majority of these studies, ⁃ How has the power supply reliability of the PV mini-grid system
however, were simulations based on synthetic (artificial) load profiles changed over time?
generated from assumed energy demand and usage. But, artificial load ⁃ What factors are affecting the energy performance and reliability of
profiles are prone to error and may not accurately reflect actual energy the PV mini-grid system?
demand. As a result, performance assessment of PV mini-grid systems
based on artificial load data may fall short of accurately measuring the 2. Methodology
performance and reliability of the systems when operated in real-world
situations (Louie and Dauenhauer, 2016), (Mandelli et al., 2016). In this 2.1. Research approach
regard, Hartvigsson and Ahlgren (2018) found that the night load per
household estimated from appliance data collected through interviews Research on rural electrification through PV mini-grids involves
in a rural village in Tanzania was 37 W, compared to the measured night analyzing the interactions and relationships between a wide range of
load of 280 W per household. The authors showed that the variables, from the technical and operational elements of the mini-grid
interview-based load profile underestimated the load factor and the to the economic and institutional characteristics of customers (Hart­
capacity factor of the plant by 34–117%. This shows that assessing vigsson et al., 2021). Therefore, the present study was conducted using
mini-grid performance based on estimated load data could lead to poor an interdisciplinary case study approach that integrates both quantita­
conclusions about the system’s technical and economic viability. tive and qualitative research methods. The application of the interdis­
Furthermore, previous mini-grid performance analyses have been ciplinary research methodology allows for a more comprehensive
focused on technical issues, largely overlooking the impact of local understanding of the influences and interactions of various factors on
economic activities and productive use customers on the magnitude and the performance of the mini-grid system, as well as the mapping of these
timing of the peak load; and the performance responses of the PV plants influences to underlying problems or causes. The case study method, on
once operational (Hartvigsson et al., 2021). the other hand, allows the researchers to conduct an in-depth investi­
There have indeed been a few studies that used real data to evaluate gation and contextual analysis of the phenomenon/problem within a

2
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

defined real-world setting, using real-time data drawn from multiple The main system components of the MG include: PV modules, con­
sources, and methods. As shown in Fig. 1, the current case study was verters (solar direct current (DC) to alternating current (AC) inverters,
conducted following six linear but iterative processes based on Yin and battery DC/AC inverters), battery energy storage system (BESS), MG
(2014). These processes include 1) planning the case study research and monitoring and energy management system (MMEMS), a diesel gener­
identifying the research questions, 2) developing the case study research ator (DG), a distribution panel (with three AC power feeders) and loads.
design, defining the unit of analyses and criteria (standards) for inter­ However, the DG has not been used thus far due to the high cost of diesel
preting the results, 3) preparing the case study protocols and data fuel (US$1/liter) in the area. A detailed technical description of the MG
collection instruments, 4) collecting data, 5) analyzing and interpreting system is presented in Appendix 1.
the data, and 6) sharing the findings and knowledge gained.
2.3. Data sources and collection methods
2.2. Description of the case study
2.3.1. Measured meteorological data
2.2.1. Location of the case study site Measured data on solar irradiation on the tilted panels and ambient
The case study was conducted on a stand-alone PV mini-grid system air temperature were obtained through direct access to the Ethiopian
(hereafter MG) installed in a small remote town named Omorate in Electric Utility’s (EEU) Universal Electricity Access Program (UEAP)
southern Ethiopia. The MG was selected for the study owing to its database; and from the contractor that built the MG system. Although
location in a hot tropical climate, the availability of operational data and the MG began producing power in late April 2021, on-site measurement
the fact that the MG is among the first PV power plants installed in and recording of incident solar irradiance and ambient air temperature
Ethiopia. The town lies between 4◦ 80′ 16′′ N Latitude and 36◦ 3′ 29′′ E had been underway since January 2020. The measurement and
Longitude with an average elevation of 368 m. a.s.l. The mean annual recording of the solar irradiance (W/m2) and ambient temperature (◦ C)
temperature in Omorate is 28.2 ◦ C. In 2021, the town had a population were performed in accordance with the IEC standard 61 724–1: 2017 for
of 3,852, which included 770 households. The MG system began measuring equipment and methods (IEC, 2017). Two standard pyran­
generating electricity in late April 2021. In May 2021, the number of ometers had been used; one mounted at an angle on the array plane for
customers served by the MG was 97. However, by the time this case measuring the in-plane irradiance, and the other one hidden to measure
study was completed (January 2022), the total number of customers had the ambient temperature. The irradiance was measured between 06:00
climbed to 443; of which 301 (68%) were ordinary households, 112 and 19:00 and the ambient air temperature recorded for 24 hours. Both
(25.3%) were small enterprises, mostly household-based businesses, and measurements were taken on a daily basis at 15 min intervals, with
30 (6.7%) were public institutions. As the number of customers grew uncertainty of <2%. Using these recorded values from two years, the
sharply, the power generated by the MG was no longer sufficient to meet average total solar insolation (kWh/m2/day) and the average ambient
the daily electricity demand. As a result, daily load shedding has been in air temperature are calculated for each day from January to December.
effect since mid-July 2021. These values are used to estimate the reference yield of the array and
calculate some of the performance metrics.
2.2.2. Technical specification of the MG system
The MG in Omorate has a total installed capacity/rated power (Pnom) 2.3.2. Real-time PV electricity generation and load data
of 375 kWp. The PV array consists of 1210 series-connected mono­ This study is based on measured on-site power generation and con­
crystalline PV modules from Jinko (Model: JKM310M-60). Each PV sumption data. The data on hourly DC output power from the PV array,
module has a rated power of 310 Wp and a rated efficiency of 18.94%. hourly AC electricity generated by the PV inverters, and the hourly
The modules are assembled into 9 strings in two parallel rows (Fig. 2). energy consumed by the load were all retrieved directly from the mini-
Each string is connected to one inverter from Growatt (Model: MAX grid’s Monitoring and Energy Management System (MMEMS) portal as
50KTL3 LV) that has a maximum output power of 50 kWp. Each inverter shown in Table 1.
has 6 maximum power point trackers (MPPT). All the modules in each As the MG began operating on April 20, 2021, complete data on
string are fixed on ground-mounted racks and positioned in a direction hourly power generation and consumption were available only for the
facing towards south at a tilt angle of 15◦ . The MG system is alternating first 245 days or 8 months (1 May – December 31, 2021) of the MG’s
current (AC)-coupled and is equipped with five Lithium Iron Phosphate operation by the time this case study was completed. The MMEMS
(LiFePO4) battery packs with a total rated storage capacity of 600 kWh. system captures data from all inverters in the PV array, the battery
A schematic diagram of the MG system in Omorate is presented in Fig. 3. system (BESS), the DG, and the power feeders to the loads. The data is

Fig. 1. Processes in conducting a case study research (Yin, 2014).

3
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

Fig. 2. A view of the PV MG infrastructure in Omorate, Dasanech district, southern Ethiopia.

Fig. 3. Schematic diagram of the AC-coupled PV mini-grid system in Omorate, southern Ethiopia.

Table 1
Partial display of the daily AC power report retrieved from the MG’s monitoring system.
Site OMORATE Date 2021-12-31

Time PVG BESS

Ia(A) Ib(A) Ic(A) P (kW) Q (kVar) COS Ia(A) Ib(A) Ic(A) P (kW) Q (kVar) COS

00:00 0 0 0 0 0 0.00 14 1 4 4 0 0.98

01:00 0 0 0 0 0 0.00 1 1 4 1 0 0.93
02:00 0 0 0 0 0 0.00 1 1 4 1 0 0.93
03:00 0 0 0 0 0 0.00 6 1 4 2 0 0.99
04:00 0 0 0 0 0 0.00 6 1 4 2 0 0.99
05:00 0 0 0 0 0 0.00 1 1 4 1 0 0.93
06:00 0 0 0 0 0 0.00 1 1 4 1 0 0.94
07:00 15 15 14 9 − 1 0.90 14 15 13 − 10 0 − 0.99
08:00 89 89 89 62 2 1.00 83 89 86 − 60 − 2 − 1.00
09:00 211 211 211 140 41 0.96 72 74 63 − 47 12 − 0.96
10:00 303 303 303 208 38 0.99 164 156 150 − 97 17 − 0.97
11:00 363 363 363 248 53 0.98 206 209 208 − 145 3 1.00
12:00 223 223 224 152 29 0.99 70 79 88 − 48 24 0.84

4
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

made available from the MMEMS portal on a daily basis in three power power of the PV array (kWp).
report formats: 1) hourly DC output power from the PV array, 2) AC
power output from the PV Generator (PVG), the BESS and DG, and 3) c) The Final Yield (YF):
power supplied to loads through the feeders. All the power reports are in
kW. As shown in Table 1, the daily AC power report, for example, pro­ In stand-alone PV MGs, the Final Yield (YF) represents the energy
vides detailed information on the hourly PV generation, peak generation actually delivered to the load by the plant in a given time period
amount and the hour, battery charging (-ve values) and discharging normalized by the rated power of the plant (Ma et al., 2013). The daily
(+ve values), and depth of charging, among other things. Likewise, the YF is expressed by equation (3):
daily load report provides detailed information on hourly electrical
EAC,Load
loads, the peak load, load-shedding hours, etc. Using these daily power YF = [kWh / kWp / day] (3)
P nom
and energy reports, a time-series dataset was created for the 8 months.
This dataset is utilized for many of the analyses in this paper. where EAC, Load is the daily total AC energy delivered to the load (kWh/
day) and Pnom is the rated power/nominal installed capacity the MG
2.3.3. Field visits, customer surveys, and in-depth interviews system (kWP).
In addition to the real-time energy data extracted from the MMEMS
of the power plant, data on the monthly electric consumption of each d) Capture Loss (LC):
customer was obtained from the local EEU billing office. Furthermore,
surveys were conducted with customers from different sectors (house­ The Capture Losses (LC) embodies energy losses from the PV array
holds, productive/commercial users, and public/state institutions). The during the energy collection and conversion process. These losses can
surveys were conducted by using semi-structured questionnaires that stem from the effect of weather conditions such as high temperature,
were designed, pre-tested and revised following a pilot study and module quality, ageing, wiring, shading and soiling, dirt and dust
guidelines suggested by Yin (2014). The surveys were administered deposition or any other inefficiency in the PV array. It is calculated as
through face-to-face interviews from 20 November to December 15, shown in equation (4):
2021 by the researchers. Throughout the field study period, repeated
visits to the MG, as well as in-depth interviews and discussions with the LC = YR − YA [kWh / kWp / day] (4)
operators, EEU staff, local authorities, community leaders, and business
owners were conducted. e) System Loss (LS):

2.4. Data analysis and evaluation parameters The System Loss (LS) denotes energy losses and inefficiencies in the
various system components including the battery bank, wiring and
The system performance analyses were carried out according to the distribution losses, energy used by the plant for operation and unused
IEC Standard guidelines and methods IEC-TS 61724–3: 2016 (IEC, losses. It is expressed by equation (5):
2016). The IEC standard specifies methods and parameters for
LS = YA − YF [kWh / kWp / day] (5)
measuring and evaluating the power generation and energy yield of a
power system under real-world conditions relative to estimated energy
output from the same plant under Standard Test conditions (STC). The f) Solar power conversion efficiency/on-site module efficiency (ηpc):
IEC performance evaluation methods also allow for comparative anal­
ysis of the PV module and system efficiencies of a PV MG in real-time The solar power conversion efficiency (also known as electrical ef­
situations against nominal efficiencies at STC. The normalized param­ ficiency) of a PV MG system (ηpc) can be defined as the ratio of the PV
eters and equations used to evaluate the performance, efficiency and array’s actual energy output (EAC) to the total solar radiation incident on
reliability of this MG system are presented as follows. the array surface (kWh/m2/day). It measures how much of the solar
energy incident on the PV plane is converted to useable electrical energy
a) The Reference Yield (YR): when operating under real conditions. Following Joshi et al. (2009), ηpc
is given by equation (6)
The Reference Yield (YR) represents the theoretically possible energy EAC,generated
output of the PV array if it operated at its nominal efficiency normalized ηpc = (6)
A ∗ Gtotal
by the rated power of the plant under STC without any losses. Following
Ma et al. (2013) the daily YR of a PV plant is given by equation (1):
∫ g) Performance Ratio (PR):
Gi dt A ∗ ηSTC ∗ Gtotal
YR = = [kWh / kWp / day] (1)
GSTC 1 ∗ Pnom The performance ratio (PR) of an isolated PV plant is the ratio of the
2 energy delivered to the load (EAC, Load) to the energy (DC) that would be
where A is PV array total effective surface area (m ), ηSTC is the PV
produced if the system was operating at its nominal efficiency and rated
module efficiency at STC, Gtotal is the total daily in-plane solar irradiance
power at STC. PR reflects the influence of losses incurred during energy
(kWh/m2), Pnom the MG system’s rated power (kWp), and GSTC is the
capture, conversion, storage and distribution on the rated output of the
global tilted irradiance at STC (1 kW/m2).
MG system (Ma et al., 2013). PR values are typically expressed in per­
centages and are given by equation (7):
b) The Array Yield (YA)
YF
PR = [%] (7)
The Array Yield (YA) denotes the total daily DC energy output from YR
the PV array (kWh) normalized by the PV’s rated power. The daily YA is
given by equation (2): h) Temperature Corrected Performance Ratio (PRcorr):
EDC
YA = [kWh / kWp / day] (2)
Pnom The performance of PV modules and therefore of the MG is strongly
affected by temperature, especially in tropical climates (Kurnik et al.,
where: EDC is the DC energy output (kWh/day), and Pnom is the rated

5
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

2011). Thus, it is imperative to calculate the Temperature Corrected MG customers in this study were calculated based on the daily power
Performance Ratio (PRcorr) in order to measure the effect of the hot interruption hours using equation (12). Data on the number of cus­
tropical climatic conditions at the site on the performance of the MG. tomers in each day (NT), the frequency of power interruptions per day,
The temperature corrected performance ratio (PRcorr) can be calculated and the total duration of power interruptions per day (Ui) were collected
using equation (8) from the IEC (IEC, 2016) from the daily reports of the MG’s energy monitoring system, daily
[ ] operational reports of the operators, and from the customer surveys we
PR
PRcorr = (8) conducted.
1 + α (Tm − Tn )
∑
N

where α is the temperature coefficient of power (α = − 0.39%/◦ C, ac­ Ui Ni

cording to the manufacturer’s data), Tm is the module temperature, and SAIDIm = d=1
(12)
NT
Tn is the nominal module temperature (25 ◦ C).
Reliable measured data on the PV modules working temperatures where Ni is the number of customers in the group, NT the total number of
were not available; the module temperature Tm was thus calculated by customers served by the MG, which is the same as Ni in this case, Ui is the
using equation (9) following Duffie and Beckman (1991). total duration of power outages (hours) in the month, and N is the total
[ ] number of days in that particular month.
Tc,NOCT − Tα,NOCT [ η ]
T m = T α + GT 1− c (9)
GT,NOCT τα
3. Results and discussion
where Tm is the module temperature, Tα is the ambient temperature
(◦ C), GT is the solar radiation incident on the PV array (kW/m2), Tc,NOCT 3.1. Solar irradiation and ambient temperature analysis
is the nominal operating module temperature (45 ± 2 ◦ C, manufac­
turer’s data), Tα,NOCT is the ambient temperature at which the NOCT is According to the measured solar irradiation data, the total annual
solar energy resource incident on the tilted PV array is 2247 kWh/m2/
defined (20 ◦ C), GT,NOCT is the solar radiation at which the NOCT is
year. The average daily solar irradiation is 6.1 kWh/m2, however, it
defined (0.8 kW/m2), ηc is the conversion efficiency of the PV array
varies between 4.6 and 7.5 kWh/m2/day. The monthly average daily
(18.94%) and τα the product of the solar transmittance and solar
solar irradiation and ambient temperature are shown in Fig. 4. The
absorbance. A τα value = 0.9 is used as suggested by Duffie and Beckman
figure displays that the lowest average daily solar irradiation (5.59
(1991).
kWh/m2/day) is recorded in July. The peak irradiation (7.05 kWh/m2/
day) is recorded in January. The average daily ambient air temperature
i) Capacity Factor (CF):
at the MG site is 30.1 ◦ C, with a minimum of 28.2 ◦ C in July and a
maximum of 33.0 ◦ C in February. In general, Fig. 4 shows that the dis­
The capacity factor (CF), also referred to as the capacity utilization
tribution of solar irradiation at the MG site has little seasonal variation.
factor, is defined as the ratio of the AC energy output (EAC) of the MG
system to the amount of energy that the system would generate if it
operates at its rated power (Pnom) continuously for 24 h a day 3.2. Module efficiency under real conditions (power conversion
throughout the year (8760 h). The annual CF of a PV plant is given by efficiency)
equation (10):
The PVAC power conversion efficiency (ηpc), in lieu of the nominal
EAC generated [kWh/year]
CF Annual = [ ] [%] (10) module efficiency at STC, is often used to determine the energy con­
Pnom kWp x 8760h
version and power generation efficiency of PV modules under real
working conditions (Joshi et al., 2009). According to our results ob­
j) Overall System Efficiency (ηsys): tained from equation (6), the average daily power conversion efficiency
of the PV modules over the 8-month period is 9.85%. The result indicates
The overall system efficiency (ηsys) of a MG system represents the that only 9.85% of the solar energy incident on the PV plane is converted
ratio of the total amount of energy delivered to the load (EAC) in a given to useable electrical energy. It also reveals that the module efficiency
time to the total solar energy available at the PV array during that under real outdoor working conditions is almost half of its nominal ef­
period. The monthly system efficiency (ηsys.m) is given by equation (11): ficiency under STC (18.94%). Fig. 5 presents the monthly average daily
on-site module efficiencies over the 8-month period.
EAC,m
ηsys,m = (11) Fig. 5 shows that the monthly average PV module efficiency ranges
A ∗ Gtotal,m
from 9.39% in September to 10.44% in June 2021. However, it is found
that the average daily PV module efficiency varies considerably from 5%
where EAC, m is the energy actually delivered to the load in the month, A
to 11.35%. The lowest daily module efficiencies (5–7.95%) were found
is PV array’s total surface area (m2), Gtotal, m is the total monthly in-plane
in May, July and October, all of which coincided with daily radiation
solar irradiance on the PV array (kWh/m2/month).
values below 5.3 kWh/m2. The highest module efficiencies
(10.94–11.35%) were observed in June and August, and were paired
k) Power supply reliability analysis:
with radiation levels close to the total average daily radiation (6.1 kWh/
m2). Despite the strong correlation between PV module efficiency and
Reliability is an important indicator when assessing the performance
daily in-plane solar irradiance level, the variation in module efficiency
of stand-alone MG systems. The reliability of the MG in this study is
between the eight months was not substantial, as shown in Fig. 5.
evaluated from the angle of energy reliability. The energy reliability of a
The large difference between the on-site module efficiency (ηpc) and
power system can be defined as the system’s ability to supply customers
the nominal efficiency at STC (ηSTC) can be due in part to the effects of
with energy sufficient to satisfy demand with minimum power in­
the high module operating temperatures. The outdoor performance of
terruptions (Song et al., 2019). The energy reliability of a power plant is
PV modules primarily depends on the in-plane solar irradiance and the
typically assessed by calculating the System’s Average Interruption
PV module temperature (Kurnik et al., 2011; Ma et al., 2017). The
Duration Index (SAIDI) (Song et al., 2019). The SAIDI measures the
climate in Omorate is warm and hot all year round. Our calculations
monthly or annual power outage duration for a customer (or a group of
from equation (9), shown in Fig. 5, indicated that the average daily
customers) served by a power plant. The monthly SAIDI values of the
operating temperature of the PV modules in this MG is 55 ± 2 ◦ C. This

6
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

Fig. 4. The daily mean irradiation (kWh/m2/day) and average monthly ambient temperature (oC).

Fig. 5. Average daily on-site PV module efficiency and average daily ambient and module temperatures.

figure is 30 ◦ C above the STC (25 ◦ C), and 25 ◦ C above the average from 460.89 kWh/day to 1602.80 kWh/day. The fitted line in Fig. 6
ambient temperature (30 ◦ C). According to Dash and Gupta (2015), the illustrates that the daily electricity generation, given by the scatter
efficiency of monocrystalline PV modules decreases by 0.45% for every points, is directly proportional to the solar irradiance level that hits the
1 ◦ C increase in module temperature above the STC. Hence, the high panel (adj. R2 = 0.718). In terms of monthly electricity generation, we
operating temperatures may have negatively and significantly affected found that the lowest monthly electricity generation 32,816 kWh was
the efficiency of the PV modules by lowering the output voltage, and as a recorded in July, when the average daily solar irradiation was the lowest
result, reducing the power output. As will be discussed later in this (5.59 kWh/m2/day). Conversely, the highest monthly electricity gen­
article, the low electrical efficiency of the PV modules is also related to eration 40,294 kWh, was recorded in December, when the average daily
the limited battery capacity. solar irradiation was the highest (6.57 kWh/m2/day). Fig. 7 compares
the daily power generation curves of the MG in July and December
2021.
3.3. PV energy estimated, electricity generated, vs. electricity consumed/ As can be seen in Fig. 7, the daily electricity production curves for
load analysis July and December have markedly different characteristics. While the
power generation curves in July largely follow a typical PV plant, the
Using the hourly real-time AC power generation data retrieved from curves in December behave oddly with the power generation consis­
the MG monitoring system (see Table 1), we calculated the daily, tently plunging between 12:00 and 15:00. One reason for this is active
monthly and total electricity generated over the 245 days. The results power curtailment or power clipping. Active power clipping occurs
showed that the total electricity generated by the MG plant during the 8- when the MPPT charge controllers deliberately limit the inverters’ AC
month period was 289 607 kWh. The average daily electricity produc­ output power, as the battery reaches full charge, to match the inverters’
tion during the same period is calculated at 1182 kWh/day; but varies

7
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

Fig. 6. Daily PV electricity generation (kWh/day) plotted versus daily solar irradiation (kWh/m2/day) (Based on data retrieved from the mini-grid’s energy
monitoring system, EEU, 2021).

Fig. 7. Daily PV power generation curves of the MG in July (left) and in December 2021 (right).

power to the load. As a result, with the high solar irradiation in yields:1estimated electricity output,2electricity actually generated,
December, the inverters’ AC power out exceeds the load and battery and3electricity delivered to the load on a daily basis are shown in Fig. 8.
storage capacity early in the day, around 11:00, before the mid-day peak The average daily estimated energy output of the MG (Eqn. (1)) was
load, and active power clipping occurs. This phenomenon was also calculated to be 2214 kWh. When this value is compared to the average
confirmed during our field visits to the MG plant in December 2021. A daily electricity generated by system (1182 kWh/day), a substantial
similar problem of severe power curtailment was found by a GIZ study of difference of 1032 kWh/day is observed. This shows that only 53.4% of
two off-grid PV MGs in rural Indonesia (GIZ, 2017). The power clipping the rated power of the MG system is actually produced whereas the
is also directly related to the limited battery capacity, as will be dis­ remaining 46.6% is not. In contrast, the average daily energy delivered
cussed in the next section. to the load is 1030 kWh/day. This means that 87% of the daily electricity
Another reason could be the effect of high temperatures on module generated is consumed by the load whereas distribution and system
and inverters efficiency. The calculated average module temperature in losses account for the remaining 13% of the generated power. The
December was around 56.3 ◦ C compared to 53.7 ◦ C in July. It is there­ significantly lower AC power output of the PV inverters compared to the
fore possible that the high afternoon temperature and heat wave in estimated (rated) output can be explained by energy losses and in­
December may have reduced the PV modules and inverters efficiency efficiencies in the energy capture and conversion process including
more than it did in July (rainy season). losses due to inverter inefficiencies. Conversely, the high percentage of
Given the data on the daily electricity generated, and the daily daily energy output delivered to the load shows that the load absorbs
electricity consumed by the load, the daily estimated DC energy output nearly all of the AC power generated by the plant each day.
from the PV array was calculated using equation (1). The three energy During the field study and surveys, it was observed that the main

1
Estimated electricity output denotes the maximum daily energy output of
the plant calculated from the rated PV capacity, module efficiency at STC, ar­
ray’s effective surface area (m2), and daily in-plane solar irradiance (kWh/m2).
2
Electricity generated represents the total daily AC energy output from all
the PV inverters (kWh/day).
3
Electricity delivered denotes the total daily AC energy output that is actu­
ally consumed by the load (kWh/day).

8
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

Fig. 8. Daily estimated, generated and consumed electricity (kWh/day) from the MG system.

power feeder to the loads is cut off between 17:00 and 19:00, and again customers. These findings are comparable to those of a study (Hart­
between 21:00 and 08:00 almost every day in December. This is done by vigsson et al., 2021) in a rural village in Tanzania.
the operators as a load management strategy to save energy during low
demand hours and supply it during the evening peak hours. It was also
observed that the load profiles of productive use customers include a 3.4. Hourly energy flow and energy balance analysis
variety of high-power appliances such as welding machines, deep
freezers, air conditioners, compressors, cooking stoves and hair dryers. To further examine the energy performance of the MG, the hourly
A separate analysis of the load profiles of customers based on monthly energy flows and hourly net energy balances were analyzed using the
electricity utility bill data and using equation (12) showed that pro­ hourly AC power generation, energy consumption, and battery power
ductive use customers consumed 53% of the energy delivered by the PV data for December, as an example. The results, shown in Fig. 9, illustrate
installation in December despite representing only 25.3% of the total that the PV power output is in excess of the load during the morning
hours, with the peak output power (230 kW) occurring at 11:00. While

Fig. 9. Hourly energy flows and hourly net energy balances of the MG in December 2021.

9
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

the PV array is producing excess power during these hours, the battery is to prolonged deep-discharging can reduce the battery’s efficiency and
charging (negative values). The maximum charging power (− 128.8 kW) can shorten its life time. Considering the high likelihood of further in­
is reached at 11:00, coinciding with the peak generation. The load kicks crease in the electricity demand, expanding the battery’s capacity may
in around 08:00 and shortly climbs to 92.55 kW at 10:00 and remains help improve the power supply. Other studies (Benavente et al., 2019),
stable and above 90 kW until it is shed off at 17:00. however, argue that increasing the MG system’s capacity as opposed to
The power feeder is reconnected to the load at 19:00, and the load the battery capacity alone is more cost-effective in overcoming power
reaches its peak (126 kW) at 20:00 before being shed off again at 21:00. outages due to capacity shortages.
Throughout the 24 h, the net energy balance remains close to zero. In At the root of the PV generation and battery capacity shortages lie
general, the figure demonstrates two significant phenomena. The first is poor initial load assessment and subsequent undersizing of the MG
that, given the high daily load, the rated battery capacity is insufficient. infrastructure. From our interviews and discussions with UEAP officers
This is shown by the rapid depletion of the battery’s state of charge at EEU headquarters in Addis, it was apparent that the EEU had not
(SOC) within 2–3 h of discharging between 19:00 and 22:00. The second imposed any financial or other constraints when determining the ca­
is that, even when the solar radiation is good and the battery is fully pacities of the PV system and battery bank. However, the load assess­
charged, the PV electricity production is insufficient to fully satisfy the ment and demand forecast that laid the basis for the design and sizing of
demand at certain hours of the day (13:00–15:00). In light of the the MG including the battery bank was done remotely in Addis Ababa
observed low PV module efficiency, this could indicate a mismatch be­ (850 km away from the MG site) by a Canadian consulting firm jointly
tween the PV array’s generation capacity and the demand, meaning that with EEU staff. Apparently, the load assessment and demand forecast
the installed capacity of the MG, including the battery, is undersized were performed based on data that was mostly inaccurate, outdated
compared to the actual load. and/or insufficient such as the number of households in the town, and
satellite imagery; as opposed to extensive and detailed on-site demand
3.5. Battery bank performance analysis assessment. The underlying assumption of the consultant in the initial
load assessment was that the demand for electricity in the town is low
To evaluate the battery bank’s performance, the hourly SOC were since most of the communities (Dasanech tribe) are pastoralists.
recorded and analyzed for the 31 days in December 2021. The battery Contrary to the consultant’s assessment, we found that the town is a
SOC is defined as the ratio of the total amount of energy presently stored melting point of people from all over the country with vibrant small-
in the battery (available capacity Qt) to the maximum possible charge businesses. We also found that productive use customers are
that can be stored in the battery (nominal capacity Qn) (Ma et al., 2017). consuming a substantial amount of the daily power output, and were
The maximum depth of discharge (DOD) of the batteries in this MG is responsible for keeping the electrical load stable and high during the day
limited to 70% and the allowable maximum SOC is limited to 90%. As as well as for the evening peak load. Our results imply that the dimen­
shown in Fig. 10, the battery SOC rises sharply from 30% at 07:00–72% sioning of the MG system, from the very beginning, has failed to account
at 13:00 following the surplus power generated by the PV array. Yet, for the impact of the hot weather condition (hence the use of electricity
from 13:00 to 15:00 the SOC is unsteady due to power discharging to for air conditioning and refrigeration), the large number of productive
offset the energy deficit created by the drop in PV power generation. The users, and the growing urbanization of the town on the electricity de­
SOC peaks (89%) at around 17:00, about 6 h after the peak PV mand. These findings strengthen the assertions of Numminen and Lund
generation. (2019) that inaccurate initial load assessment and poor technical design
However, within 2–3 h of discharging, the SOC deteriorates to 40% are among the main factors behind the poor energy performance and
at around 21:00. Fig. 10 shows that the battery undergoes unsteady and reliability of PV MGs in developing countries.
rapid charging/discharging rates. The battery is also discharging deep
from 19:00 to 21:00. This confirms that the battery capacity is insuffi­ 3.6. Reliability of power supply from the MG system
cient to store enough energy and discharge power to the loads smoothly
over a longer period of time. Rapid charging/discharging and exposure The power supply reliability of this MG system was analyzed by

Fig. 10. Hourly mean variations in the battery state of charge (SOC) in December 2021.

10
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

Fig. 11. Monthly total SAIDI value (hours) and average monthly electricity consumption per customer (based on monthly electricity consumption data per customer
from EEU, 2021 and using equation (12)).

calculating the System’s Average Power Interruption Duration Index barriers (Benavente et al., 2019). The findings clearly demonstrate the
(SAIDI) values, and the average monthly electricity consumption per impact of the dynamics of the electricity demand and usage behaviors of
customer using equation (12). All calculations were made using the rural communities on the reliability of MG systems and the need to ac­
actual data on monthly electricity consumption of each customer ob­ count for such changes in initial system sizing of off-grid MGs.
tained from the EEU billing records and our surveys. The results, pre­ According to the financial data collected from the MG operator, the
sented in Fig. 11, show how the SAIDI increases and average power average monthly operation and maintenance cost of the MG including
consumption decreases as the total number of customers increased from staff salaries was US$ 582 and the average monthly replacement cost
97 in May to 197 in August and to 443 in December 2021. was US$ 1550. This puts the total average monthly expenditure neces­
Following the rapid growth in the number of customers, the monthly sary to run the MG at US$ 2132. In contrast, the average monthly rev­
power interruption hours (SAIDI) per customer increased from 15.5 h (= enue collected from all customers was US$ 1059 (i.e. US$ 173 from
0.5 h/day) in May to 286.75 h in August (= 9.25 h/day) and 402.38 h (= ordinary households, US$ 782 from productive use customers, and US$
12.98 h/day) in December 2021. However, the monthly electricity 104 from institutions). Comparing the average monthly expenditures
supplied by the PV plant has not changed significantly over the 8-month and revenues of the MG operator shows that the MG is currently oper­
period. As a result, the average monthly electricity consumption per ating at a loss (of about 50% of its monthly expenditure). While the total
customer fell from 318 kWh in May to 165.7 kWh in August and 71 kWh economic benefits of the MG may outweigh its operation costs in the
in December (a decline by 77%). The data showed that the power sup­ long run, as it stands now, the MG does not appear to be financially
plied by the MG in May was sufficient to meet the demand; with total viable and sustainable. The data above also reveals that productive users
power outage durations of only 15.5 h, mainly for operational reasons. (small businesses) account for nearly 75% of the revenues collected from
In the months that followed, however, the number of customers grew customers while representing only 25% of the total customers.
fast while the supply remained essentially unchanged. In fact, the MG
operator (EEU’s district office) was aware that a sharp increase in the 3.7. Overall system performance analysis
number of customers could compromise the reliability of the MG’s en­
ergy supply but prioritized providing basic access to electricity to all Findings from the evaluation of the overall system performance and
initially registered and paying customers rather than limiting the num­ efficiency based on the IEC standard normalized parameters and equa­
ber of customers. However, the operator no longer provides connections tions 1–11 indicated that the MG system is performing poorly. The re­
to new customers. sults of the energy yield and efficiency analyses are summarized in
The surveys and in-depth interviews with key informants revealed Table 2.
that most early adopters of the MG service were small business owners; According to our results, the average daily reference yield (YR), array
and they are the ones who are now losing much of their former access to yield (YA) and final yield (YF) of the MG over the 8-month period were
reliable power supply. This, in turn, affects both the small businesses and 5.90 kWh/kWp/day, 3.15 kWh/kWp/day, and 2.75 kWh/kWp/day,
the financial returns of the MG. The results illustrate that the MG is respectively. These results confirm that only 53.4% of the potential solar
currently unreliable in terms of electricity supply, and that the average energy available at the PV array’s surface (YR) is harvested by the
monthly electricity consumption per customer observed in December inverter (YA). As discussed earlier, the large disparity between the
(71 kWh) is rather a ‘suppressed demand’ due to extended hours of load- reference yield (YR) and the array yield (YA) relates to energy losses and
shedding and power interruptions. A suppressed energy demand occurs inefficiencies incurred in the capture and conversion of the in-plane
when the energy supplied by a power system is insufficient to meet the solar energy into useful electricity. Table 2 shows that the total
basic needs of the customers due to limited installed capacity of the average daily energy loss of this MG system is 3.15 kWh/kWp/day;
power plant, inadequacy of power supply, low user income, or other which comes from average daily capture losses (LC) of 2.75 kWh/kWp/

11
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

Table 2
Summary of the results of the energy performance and efficiency analyses of the MG.
Parameters YR YA YF LC LS PR PR corr CF ηsys
Month (kWh/kWp (kWh/kWp/d) (kWh/kWp/d) (kWh/kWp/d) (kWh/kWp/d) (%) (%) (%) %
/d)
May 5.60 2.99 2.67 2.62 0.31 47.8 45.6 12 8.93
Jun 6.08 3.43 2.94 2.64 0.49 48.7 45.1 14 9.09
Jul 5.46 2.82 2.65 2.63 0.17 49.1 46.4 12 9.17
Aug 5.81 3.21 2.81 2.60 0.40 48.5 46.5 13 9.08
Sept 5.88 3.05 2.59 2.83 0.47 44.5 39.8 13 8.42
Oct 5.81 2.97 2.81 2.84 0.17 48.7 40.6 12 9.01
Nov 6.14 3.27 2.74 2.87 0.53 44.8 36.5 14 8.36
Dec 6.43 3.47 2.77 2.96 0.70 43.1 34.4 14 8.01
Average 5.90 3.15 2.75 2.75 0.40 47.9 42.0 13 8.76

day and average daily system losses (LS) of 0.40 kWh/kWp/day. These month. However, it is also important to note that a low PR value for an
figures reveal that energy losses during energy capture and conversion off-grid PV plant does not necessarily mean that the system is experi­
account for most (87%) of the total energy loss; while only 13% is a encing technical difficulties. Instead, it can be caused by a poor match
result of energy losses and inefficiencies in the distribution and system between the MG system’s installed capacity and electricity demand. As
components. noted by Hartvigsson et al. (2020) the viability and operational behavior
The large capture losses can be linked to the low efficiency of PV of distributed MGs heavily depends on the interactions and feedback
modules, power clipping and the effects of high temperatures. Given between endogenous factors (e.g., between electricity demand and
that this is an off-grid PV MG with a storage battery and MPPT tracker, supply) rather than exogenous factors. As such, the low PR value of this
the PV array yield is also influenced by the quantity of energy needed to PV system might be due more to the limited battery capacity limiting the
feed the instantaneous load and to charge the battery bank. In other amount of solar energy the system can capture and deliver. As can be
words, how much of the available solar energy is harnessed by the seen in Fig. 12, the PRcorr and the load in this MG are strongly and lin­
inverter also depends on how much energy the battery bank can store, in early correlated, suggesting that the PRcorr of the system could have been
addition to feeding the active load. The large difference between the higher if the battery capacity had been larger and more energy was
reference yield and the array yield could thus be partly explained by the captured by the inverter and delivered to the load.
low storage capacity of the battery bank. Comparing the final yield (YF) The average capacity factor (CF) of the system is 13%, indicating that
with the reference (YR) observes that 46.56% (2.75 kWh/kWp) of the the energy output of the plant is 13% of the theoretical energy output
total available solar energy on the PV array surface (YR) is delivered to that the MG would provide if it had operated 24 h a day, including
the load each day. nights, at its rated capacity. The overall system efficiency (ηsys) of this
The results of the efficiency analyses showed that this PV power PV power plant is 8.76%. Previous research indicates that the overall
plant performs poorly with an average performance ratio (PR) of system efficiency of PV power systems in tropical climates ranges be­
47.92% and an average temperature-corrected performance ratio tween 5% and 15% (Joshi et al., 2009; Kumar and Sudhakar, 2015;
(PRcorr) of 42%. A PV power system is typically considered to have high Minai, 2022). A study on a 20 kWp PV micro-grid in a hot tropical
performance when its PR vale is above 70% (Kumar and Sudhakar, climate in Colombia (Banguero et al., 2019), also found the overall
2015). Compared to this, the performance of this MG system falls within system efficiency of the micro-grid to be 10.3–11%. Compared to these
the low range. Comparing the normal PR and the temperature-corrected reports, the overall system efficiency of the current MG system is rela­
PRcorr shows that the system experiences drops in performance of nearly tively low.
6% due to the effect of the high temperature in the area. The impact of
temperature is most visible in December, where the MG experienced a
PR drop of nearly 9% despite the high level of solar irradiation in the

Fig. 12. Relationship between the PRcorr and energy delivered to the load by the MG system.

12
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

4. Conclusions in hot tropical climates can be significantly reduced by large drops in PV

module efficiency and energy capture losses that come with operating at
The real-time performance and power supply reliability of a 375 kWp high temperatures. Hence, future deployment and system sizing of PV
off-grid PV mini-grid system installed in a small remote town in Ethiopia power plants in hot tropical areas of Ethiopia and the SSA at large should
is analyzed using measured meteorological data and real-time power critically assess and take into account the effect of local weather con­
generation and consumption data retrieved from the energy monitoring ditions, economic activities and potential growth in electricity demand
system of the mini-grid over an eight-month period (May 01 to on the performance and reliability of MG systems. In this sense, the
December 31, 2021). Findings show that the average PV module effi­ paper provides valuable lessons and practical experiences to inform
ciency (ηpc) under real conditions is 9.85% compared to its nominal policy makers, mini-grid developers, governments, researchers and or­
efficiency of 18.94% under STC. The disparity in module efficiency is ganizations working on MG-based rural electrification in developing
related to the large PV array capture losses (LC = 2.75 kWh/kWp/day). countries.
The mean daily electricity produced by the MG 1182 kWh/day was
found to be considerably lower than the mean estimated energy output Declaration of competing interest
of 2214 kWh/day; with a difference of 1032 kWh/day (46.6% less than
the estimated). In contrast, 87% of the daily electricity generated was The authors declare that they have no known competing financial
delivered to the load. System and distribution losses (LS = 0.40 kWh/ interests or personal relationships that could have appeared to influence
kWp/day) accounted for the remaining 13%. It is found that the mini- the work reported in this paper.
grid system is performing poorly with an average daily array yield
(YA) of 3.15 kWh/kWp/day, final yield (YF) of 2.75 kWh/kWp/day, Data availability
performance ratio (PR) of 47.92%, temperature-corrected PRcorr of 42%,
capacity factor (CF) of 13% and an overall system efficiency (ηsys) of Data will be made available on request.
8.76%.
Analysis of the daily energy flow and power supply reliability Acknowledgements
showed that the energy delivered by the PV plant does not meet the daily
demand to the extent that the load is shed off the power supply for 13 h The authors wish to acknowledge the Swedish Research Council and
each day. The technical poor performance and managerial poor assess­ the Chalmers University of Technology Energy Area of Advance for
ment and design of the MG can be attributed to three main factors: a) the providing financial assistance to conduct this research. The authors
low PV module conversion efficiency, mainly due to the effect of the would also like to express their gratitude to the Ethiopian Electric Utility
high temperatures, large capture losses and active power curtailment; b) (EEU), the coordinators of the Universal Electricity Access Program
mismatch between the PV’s installed capacity (375 kWp) and the de­ (UEAP) at Hawassa and Arba-Minch, and the mini-grid operators at
mand, and c) insufficient rated battery capacity (600 kWh) compared to Omorate for their technical support and facilitation of the field data
the load. The results suggest that the performance of PV power systems collection.

Appendix 1. Technical specification of the PV mini-grid system installed

Component Parameters Specification

Geographic location Omorate, Dasanech district, Southern Ethiopia

Latitude 4◦ 80′ 16′′ N
Longitude 36◦ 3′ 29′′ E
Elevation 368 m. a.s.l.
Mean annual ambient temp 30.1 ◦ C
Installation Fixed ground-mounted racks
Configuration AC-coupled with 9 strings in two parallel rows
PV array PV module type Mono-PERC (mono-crystalline) JKM310M-60-MX
Number of PV cells per module 60
Total number of PV modules 1210
PV module dimension (Length x Width) 1.65 cm × 0.992 cm
Surface area of each PV module 1.6368 m2
Total effective area of the PV array 1980.5 m2
Tilt angel 15◦ facing south
Maximum power per unit of area at STC 189.4 W/m2
Rated output power per module 310 Wp
Global Horizontal Irradiation 5.904 kWh/m2/day
Measured irradiation at the tilted plane 6.07 kWh/m2/day
Module conversion efficiency at STC 18.94%
Total installed/rated PV capacity 375 kWp
PV lifetime 25 years
Temperature coefficient − 0.39%/oC
Converter Inverter model MAX 50KTL3 LV
Total Number of inverters 9
Max input power (DC) of each inverter 75 kWp
Max output power (AC) of each inverter 50 kWp
Nominal voltage 585 V
Max. AC apparent power (kVA) 55.5kVA
Inverter output AC voltage 220/380 VAC three-phase
Max output current 80.5 A
Lifetime 10 years
Maximum inverter efficiency 96%
(continued on next page)

13
Y.T. Wassie and E.O. Ahlgren Development Engineering 8 (2023) 100106

(continued )
Component Parameters Specification

Energy storage (Battery) Max energy storage capacity 600 kWh

Number of battery blocks/racks 5 × 60 kW
Battery cell chemistry LiFePO4
Minimum charge/discharge life cycles 3000 cycles
Max lifetime per battery 10 years
Round trip efficiency (%) 80
Battery cell energy density 125 Wh/kg
DC Power supply voltage 24 V
Diesel generator (DG) Rated output power (kVA/kW) 125/100
DG output AC voltage 220/380 VAC three-phase
Load minimum ratio 30%
Minimum fuel efficiency 32%
Estimated life time in hours 10,000-15,0000
Power factor 0.8

References Joshi, A.S., Dincer, I., Reddy, B.V., 2009. Performance analysis of photovoltaic systems: a
review. Renew. Sustain. Energy Rev. 13 (8), 1884–1897.
Katre, A., Tozzi, A., Bhattacharyya, S., 2019. Sustainability of community-owned mini-
Akinyele, D.O., Rayudu, R.K., 2016. Techno-economic and life cycle environmental
grids: evidence from India. Energy, Sustainability and Society 9 (2).
performance analyses of a solar photovoltaic microgrid system for developing
Kumar, B.S., Sudhakar, K., 2015. Performance evaluation of 10 MW grid connected solar
countries. Energy 109, 160–179.
photovoltaic power plant in India. Energy Rep. 1, 184–192.
Azimoh, C.L., et al., 2016. Electricity for development: mini-grid solution for rural
Kurnik, J., et al., 2011. Outdoor testing of PV module temperature and performance
electrification in South Africa. Energy Convers. Manag. (110), 268–277.
under different mounting and operational conditions. Sol. Energy Mater. Sol. Cell. 95
Banguero, E., et al., 2019. Renewable microgrid operational results and economic
(1), 373–376.
evaluation using RETScreenTM. Int. J. Electr. Comput. Eng. 9 (2), 723–731.
Louie, H., Dauenhauer, P., 2016. Effects of load estimation error on small-scale off-grid
Benavente, F., et al., 2019. Photovoltaic/battery system sizing for rural electrification in
photovoltaic system design, cost and reliability. Energy for Sustain. Dev. 34, 30–43,
Bolivia: considering the suppressed demand effect. Appl. Energy 235, 519–528.
2016.
Come-Zebra, E., et al., 2021. A review of hybrid renewable energy systems in mini-grids
Ma, T., Yang, H., Lu, L., 2013. Performance evaluation of a stand-alone photovoltaic
for off-grid electrification in developing countries. Renew. Sustain. Energy Rev. 144,
system on an isolated island in Hong Kong. Appl. Energy 112, 663–672.
111036.
Ma, T., Yang, H.X., Lu, L., 2017. Long term performance analysis of a standalone
Costa, T., Villalva, M.G., 2020. Technical evaluation of a PV-diesel hybrid system with
photovoltaic system under real conditions. Appl. Energy 201, 320–331.
energy storage: case study in the tapajós-arapiuns extractive reserve, amazon, Brazil.
Mandelli, S., et al., 2016. Effect of load profile uncertainty on the optimum sizing of off-
Energies 13 (11), 2969.
grid PV systems for rural electrification. Sustain. Energy Technol. Assessments 18,
Dash, P.K., Gupta, N.C., 2015. Effect of temperature on power output from different
34–47.
commercially available photovoltaic modules. Int Journal of Engineering Research
Minai, A.F., 2022. Performance analysis and comparative study of a 467.2 kWp grid-
and Applications 5, 148–151.
interactive SPV system: a case study. Energies 15, 1107.
Duffie, J.A., Beckman, W.A., 1991. Solar Engineering of Thermal Processes, second ed.
Moner-Girona, M., et al., 2018. Electrification of sub-saharan Africa through PV/hybrid
Wiley, New York.
mini-grids: reducing the gap between current business models and on-site
GIZ, 2017. Technical Report Measuring System Performance of Isolated Photovoltaic
experience. Renew. Sustain. Energy Rev. 91, 1148–1161.
Mini-grid in Rural Indonesia. Deutsche Gesellschaft für Internationale
Numminen, S., Lund, P., 2019. Evaluation of the reliability of solar micro-grids in
Zusammenarbeit (GIZ) GmbH. Energising Development (EnDev), Indonesia.
emerging markets – issues and solutions. Energy for Sustainable Development 48,
Hartvigsson, E., Ahlgren, E.O., 2018. Comparison of load profiles in a mini-grid:
34–42.
assessment of performance metrics using measured and interview-based data.
Pedersen, J., et al., 2021. In: Distribution system planning with battery storage using
Energy for Sustainable Development 43, 186–195.
multiperiod optimal power flow. Proceedings of the 14th International Renewable
Hartvigsson, E., Stadler, M., Cardoso, G., 2018. Rural electrification and capacity
Energy Storage Conference 2020 (IRES 2020), Atlantis Highlights in Engineering, 4
expansion with an integrated modeling approach. Renew. Energy 115, 509–520.
February 2021.
Hartvigsson, E., Ahlgren, E.O., Molander, S., 2020. Tackling complexity and problem
Sharma, V., Chandel, S.S., 2013. Performance analysis of a 190 kWp grid interactive
formulation in rural electrification through conceptual modeling in system
solar photovoltaic power plant in India. Energy 55, 476–485.
dynamics. Syst. Res. Behav. Sci. 37, 141–153.
Song, X., et al., 2019. Reliability varying characteristics of PV-ESS-Based standalone
Hartvigsson, E., Ehnberg, J., Ahlgren, E.O., Molander, S., 2021. Linking household and
microgrid. IEEE Access 7, 120872–120883.
productive use of electricity with mini-grid dimensioning and operation. Energy for
Vezzoli, C., et al., 2018. Distributed/decentralized renewable energy systems. In:
Sustainable Development 60, 82–89.
Designing Sustainable Energy for All. Green Energy and Technology. Springer.
IEA, 2020. SDG7: Data and Projections. International Energy Agency, Paris, France. http
Wassie, Y.T., Adaramola, M.S., 2021. Socio-economic and environmental impacts of rural
s://www.iea.org/reports/sdg7-data-and-projections. (Accessed 2 March 2022).
electrification with Solar Photovoltaic systems: evidence from southern Ethiopia.
accessed on.
Energy for Sustain Development 60, 52–66.
IEC, 2016. IEC TS 61724-3:2016. Photovoltaic System Performance - Part 3: Energy
Yin, R.K., 2014. Case Study Research. Design and Methods, fifth ed. Sage publications,
Evaluation Method. International Electrotechnical Commission, Geneva,
Thousand Oaks, CA, USA.
Switzerland.
IEC, 2017. IEC TS 61724-1:2017. Photovoltaic System Performance - Part 1: Monitoring.
International Electrotechnical Commission, Geneva, Switzerland.

14