ArticlePDF Available

Classification and modeling of load profiles of isolated mini-grids in developing countries: A data-driven approach

Authors:

Abstract and Figures

Reaching universal access to electricity by 2030 requires a massive deployment of mini-grids in rural areas of developing countries. Among the many challenges hindering this process, there are the high uncertainties in assessing demand patterns in rural communities, the costs of field survey campaigns, and the absence of ample and reliable datasets coming from existing projects. This paper tries to address these issues by presenting and discussing a database of load profiles from sixty-one off-grid mini-grids from developing countries worldwide, gathered from the literature, private developers and fieldworks, and reported with technical, socio-economic and geographical characterization factors. A clustering procedure led to the identification of five archetypal load profile clusters, which are presented and analyzed together with their load duration curves. Subsequently, the distribution among the clusters of the various characterization factors selected is studied. The proposed approach allows to widen the range of load profiles usually considered, and to seek correlations between the load profile shapes, the peak power and average energy consumption per connection, the number of customers, the age of measurement, geographical position, operator model, type of tariff and generation technologies present. This work establishes a first step in the creation of a shared database for load profiles of rural mini-grids, helping to overcome the lack of available data and difficulties of demand assessment, proposing original insights for researchers to understand load patterns, and contributing to reduce risks and uncertainties for mini-grid developers.
Content may be subject to copyright.
1
Classification and modeling of load profiles of isolated mini-
grids in developing countries: a data-driven approach
Luca Lorenzoni1, Paolo Cherubini1,*, Davide Fioriti1, Davide Poli1, Andrea Micangeli2, Romano Giglioli1
1DESTEC, University of Pisa, Largo Lucio Lazzarino, 56122 Pisa, Italy
2DIMA, "Sapienza" University of Rome, Via Eudossiana 18, 00184 Rome
E-mail addresses: luca.lorenzoni.ll@gmail.com (L. Lorenzoni), paolo.cherubini@ing.unipi.it (P.
Cherubini), davide.fioriti@ing.unipi.it (D. Fioriti), davide.poli@unipi.it (D. Poli),
andrea.micangeli@uniroma1.it (A. Micangeli), romano.giglioli@unipi.it (R. Giglioli)
Abstract
Reaching universal access to electricity by 2030 requires a massive deployment of mini-grids in rural areas
of developing countries. Among the many challenges hindering this process, there are the high uncertainties
in assessing demand patterns in rural communities, the costs of field survey campaigns, and the absence of
ample and reliable datasets coming from existing projects. This paper tries to address these issues by
presenting and discussing a database of load profiles from sixty-one mini-grids from developing countries
worldwide, gathered from the literature, private developers and fieldworks, and reported with technical,
socio-economic and geographical characterization factors. A clusterization procedure led to the identification
of five archetypal load profile clusters, which are presented and analyzed together with their load duration
curves. Subsequently, the distribution among the clusters of the various characterization factors selected is
studied. The proposed approach allows to widen the range of load profiles usually considered, and to seek
correlations between the load profile shapes, the peak power and average energy consumption per
connection, the number of customers, the age of measurement, geographical position, operator model, type
of tariff and generation technologies present. This work establishes a first step in the creation of a shared
database for load profiles of rural mini-grids, helping to overcome the lack of available data and difficulties
of demand assessment, proposing original insights for researchers to understand load patterns, and
contributing to reduce risks and uncertainties for mini-grid developers.
Keywords: microgrid, off-grid, load profile, demand assessment, rural electrification, electric
consumptions
1. Introduction
1.1. Contextualization
Universal access to electricity powered by renewable energy sources is considered a cornerstone of the
transformation of the global energy system [1]. Indeed, reaching universal access to “affordable, reliable,
sustainable and modern” energy has been explicitly recognized as a top priority for the international
community with the United Nations’ Agenda 2030 in 2015 [2], as the seventh of the 17 Sustainable
Development Goals (SDGs). Energy access is recognized to be fundamental for socio-economic development,
and SDG 7 is positively related to most of the other SDGs targets, with synergies outnumbering possible
drawbacks [3]. Access should result in the development of productive uses of energy for their
transformational impact: in fact, a revision of SDG 7 has been proposed to integrate them, which would
increase the interlinkages with other SDGs [4].
2
So far, there has been significant progress towards the achievement of SDG 7, but there were still 840
million people without access to electricity in 2017, mostly concentrated in the rural areas of sub-Saharan
Africa [5]; with current policies and forecasted population growth, it is estimated that more than 600 million
people will still have no access to electricity by 2030 [6].
Extending the national grid into rural areas it is often not economically feasible [7], therefore off-grid
systems will be key in seizing the access gap. This study will focus on mini-grids in particular, which can
provide high levels of service with respect to standalone systems, thus fully enabling productive uses of
energy [8], while also improving energy security, resilience and reliability [8][10].
It is estimated that the installation of 210.000 mini-grids will be the cheapest solution for 490 million
people to gain access by 2030, also thanks to declining technology costs [11]. Nevertheless, upfront costs are
just one of the many barriers currently preventing widespread development of mini-grids, such as
uncertainties and gaps in the regulatory framework, non-cost-reflective tariffs, limited access to finance,
need for capacity building, uncoordinated electrification planning exposing to risk of main grid arrival, lack of
reliable data and the high variability of demand, among others [12][14].
This work aims at contributing to speed up the effective development of mini-grids in developing
countries by addressing specifically the interdependent issues of data availability and demand assessment,
which are crucial for electricity planning and mini-grid sizing, and can be pivotal for the commercial viability
of mini-grids [15].
1.2. Literature Analysis
During the design phase of a single mini-grid, the initial load estimation is a critical task: overestimating
the load profile can jeopardize the project profitability due to the extra capital costs, while an
underestimation of demand leads compromises the reliability of services and dissatisfies end consumers [16],
[17]. Case studies available in the literature, e.g. for Zambia [18] or Mongolia [19], confirm that incorrect
demand assessment and consequent design limitations result in hampering the successful deployment of
mini-grids.
The effect of load profile uncertainties on the optimum sizing of off-grid PV systems has been treated
quantitatively in [20], whereas in [21], the authors calculated that over-estimating the load led to extra costs
of approximately 1.92 to 6.02 US$/Wh, for a case study in Malawi. It is clear that the load assessment is a
critical step to properly size mini-grids, but it is a complex task due to the high uncertainties of rural contexts,
and lack of prior knowledge. This is especially true for communities with no prior access to electricity, where
only the usage of traditional sources can contribute to estimating an equivalent electric load profile [22],
which is then subject to uncertainties at different timescales, going from short term, which affects daily
operations and dispatch strategy, to the long-term load evolution in the lifetime of a project [23]. The long-
term trend of energy demand is also determined by the various actions undertaken by the developer, such
as the level of community engagement [24], [25] and the promotion of productive uses of energy [26], which
is in fact referred to as “demand creation” [27].
To provide a first estimation of the expected load profile, a common practice is to employ energy-use
surveys that evaluate the socio-economic characteristics of the community, which are the inputs of load
assessment methodologies; however, interviews can be subject to huge errors as shown in [28] and [29],
thus the question list has to be well defined and validated.
In order to support the load estimation, several techniques have been developed to simulate human
behaviour [30][33]; however, these bottom-up tools need many detailed inputs concerning appliances
characteristics and consumption patterns. While reliable inputs can be obtained by employing effective data
collection methodologies, as shown in [22], the field missions, required to carry them out, are costly and
time-consuming. It is clear that establishing a solid dataset with real-world data could be used by several
companies to perform a preliminary load assessment and by researchers and practitioners in the field to test
and validate models. By processing the data, useful correlations between socio-economic aspects and
3
electricity consumption can arise to ultimately infer load characteristics of greenfield locations based on
verified similarity patterns with existing mini-grid sites.
A major barrier to the development of an ample and reliable dataset on mini-grids, as needed by
academics and practitioners, is the limited sharing of data from existing projects [34], albeit being often
advocated as a key priority to increase the viability of mini-grids [35]. Industry organizations can overcome
these limitations by gathering data from their associates: the Africa Minigrid Developers Association (AMDA),
for instance, collected operating and financing data from mini-grid sites across 12 countries in sub-Saharan
Africa [36]. However, the database is not publicly available, and the study did not tackle load profiles but only
average monthly consumptions, highlighting marked differences across countries without founding clear
causality to explain them [36].
Concerning public data collection of load profiles, a first attempt has been developed by the Sandia
National Laboratories in 2007 [37], reporting few village profiles. Prinsloo et al. performed a similar review
for few African villages [38], with a focus on single households and thermal energy needs in addition to
electrical ones. The National Renewable Energy Laboratory (NREL) reviewed a few papers reporting load
profiles from sub-Saharan Africa to extrapolate information to construct a ground-up profile for a generic
mini-grid [41]. Another study proposed a load synthetization method, but focusing only on few and similar
sites belonging to a single company concentrated in Kenya and Tanzania [39], which does not go in favor of
generalizability. The same authors performed a load profile clusterization for customer classes, using a data
sample from mini-grids in Tanzania [40]. In conclusion, to the best knowledge of the authors and according
to the reviewed literature, no other paper has proposed a data-driven classification of load profiles for rural
mini-grids in developing countries.
Acknowledging all the above, in this study we propose a first open dataset and a data-driven analysis of
measured load profile of world-wide mini-grids alongside with selected characterization factors of the same
sites, so to lay the foundation of further studies on load assessment and suggest correlations among
parameters. The proposed data-driven classification highlights correlations between the socio-economic
characteristics of the villages and the measured demand, which can help developers and researchers in
developing load assessment methodologies. In this work, the authors have gathered in a single document
real-world data of sixty-one rural mini-grids, reported in the Appendices A and B to the best detail achievable,
with the intent of suggesting correlations between geographical and socio-economic characteristics of the
sites and load profile shapes. This is a work-in-progress activity and this document is the first step towards a
comprehensive classification of archetypal load shapes that could be useful to support the design process of
new mini-grids, hoping it can be effectively employed as a consultative document for this purpose for
researchers and practitioners alike.
2. Methodology
The main objective of this work is to obtain a representative daily load curve for each site identified in
the data collection phase (sec. 2.1), to be then classified into archetypal clusters (sec. 2.2). Possible
correlations between the resulting clusters and selected external factors, including any geographical, socio-
economic and technical information available of the selected communities have been investigated (sec. 2.3).
The overall procedure is summarized in sec. 2.4.
2.1. Data collection method
The data employed in this study have been either collected in the field, thanks to the installation of
energy meters in three Honduran communities [42], attributable to private developers [43], or taken from
the literature. The literature research has been performed focusing on peer-reviewed scientific papers and
grey literature, to include relevant reports drawn up by international agencies or sector organizations. Being
established in the research field [44], the database Scopus has been consulted using structured research
4
strings featuring alternative combinations of all relevant keywords (as an example, load profile, load pattern,
load curve, load diagram) to screen the greatest possible number of papers and select the ones featuring
daily measured load profiles of rural mini-grids in developing countries.
Load profiles may exhibit various degrees of temporal variations, from a weekly basis to a seasonal or
yearly scale, depending on many socio-economic, political and geographical conditions of the local area and
population [17]. In this activity, we aimed at collecting the data with the highest level of detail possible, to
perform the correlation analysis. However, load profiles are usually reported in the literature as a single daily
measurement, sometimes justified by the limited seasonality due to the limited variety of conditions in the
year. Most microgrids in developing countries are located in equatorial regions, where the weather
conditions are similar through the year and social behaviors do not significantly change. In addition, most of
these variations are often very limited while the main interest focuses on identifying the main shape of the
load profile. Moreover, in the very few cases where more than a daily measurement was available, the
relative variations among load profiles resulted negligible with respect to the overall shape, therefore, for
the sake of simplicity, an average profile was considered
1
.
Another key point regards the evolution of shape and consumption over time. Since the mini-grid
installation, consumption tends to increase and the shape of the load profile may vary depending on specific
site features [17], [39]. Although the dynamic evolution of load profiles is not within the scope of this work,
the measurement age (i.e. the period from mini-grid commissioning to load measurement) was included to
weigh the effect of consumer takeoff in mature projects [45].
The data collection focused on gathering load profiles of real mini-grids whatever the generation
technologies are; however the most recurrent are PV, diesel, hydro, wind in few cases, sometimes combined
with battery storage, as reported in Table A-1 Table A-5 of Appendix A . As long as there is enough energy
available, the technologies do not affect the load consumption, while when load shedding occurs, the
selected assets, given the selected operating strategy, affect the load demand. Often in rural microgrids, a
relatively low share of load curtailment is generally accepted as long as it is not too high; however, aiming at
performing a clustering analysis on the load demand not constrained by the assets, the data corresponding
to sites experiencing reported generation constraints or a restriction in the hours of service (i.e. only during
the day) have been removed from the clustering analysis.
2.2. Normalization and clustering procedure
In order to compare mini-grids of different size, the profiles have been normalised to mean hourly
demand, as commonly done such as in [39], allowing a proper comparison of the load profile shapes. After
that, aiming to identify archetypal profiles, a clusterization procedure is proposed using the hierarchical
clustering that has been regarded as the most suited algorithm for the proposed case, due to the size of the
dataset [46] and the possibility to choose the number of clusters a posteriori [47]. Given the review of
different algorithms applied to time series data in [46] and the electrical load analysis in [48] and [49], it
turned out that there is no general algorithm best suited to all applications. Moreover, the most suitable one
depends upon the data setup and our a priori knowledge of the expected output’ [47]. However, drawing
analogies from the results presented in [46] and [47], and after several attempts done using different
clustering algorithms (i.e. k-means), hierarchical clustering turned out to be the most suited, thus providing
the most promising results. In particular, to evaluate the “distance” between the profile and the clusters, the
Euclidean metric [48] and the Ward linkage method, as in [47], resulted to be more suited for the scope of
this analysis, compared to other distances and methods.
1
Only two exceptions were found among the dataset: case MY1, for which the three midweek consecutive
measurements were available, and have been averaged as a single profile, and case OM2 for which the winter and
summer profiles were considered separately as reported in Table A-1 Table A-5 of Appendix A.
5
2.3. Characterization factors for load profiles
As mentioned above, in addition to the electric load profile, other data of the selected communities have
been gathered to suggest possible correlation patterns between them and the archetypal load profile
clusters. These parameters have been selected as a compromise between the level of detail and the
availability of the information. In fact, very detailed information at a community level, such as the
demographic distribution of people, income proportion, etc., would be obtainable only through field surveys,
whose results are very seldomly reported in the literature. However, in some cases, as reported in Table A-1
Table A-5 of Appendix A additional data were gathered from the sources, for example in terms of the
composition of the customer base (i.e. the number of domestic and commercial connections). It would have
been desirable to have a higher level of detail across all cases to conduct more comprehensive analyses.
However, a uniform dataset of representative parameters has been devised to include such a high number
of case studies.
The technical, geographical and socio-economic information retrieved has been divided into quantitative
and qualitative data, as discussed below.
2.3.1. Quantitative indicators
The main quantitative data we considered are the number of connections, the installed power capacity,
the daily load profile and, as mentioned above, the age of profile measurement. From this data, the average
daily energy per connection and the average daily peak power per connection were obtained, as summarized
in Table 1. The table reports also the intervals chosen to discretize the selected indicators, aiming to simplify
the analysis and the notation. For the daily peak power and the daily energy per connection, the ranges are
identified by using an approach similar to the Multi-Tier Framework (MTF) methodology [50]: the same
ranges as the MTF for household connections are used, while custom intervals are employed for the number
of connections and the measurement age, as no reference in the literature have been found according to the
proposed literature review and to the best of authors’ knowledge. The ranges have been selected to best
represent the proposed dataset.
The Multi-tier Framework has been used here only as a reference for the power and energy intervals. It
is worth noticing that MTF proposes separate tier systems for productive applications and community
infrastructure [50], but they couldn't be applied due to lack of disaggregated data on connection types for a
sufficient number of case studies. In this activity, MTF has been used only to define reference average peak
power and daily consumption level through realistic and literature-proof ranges. Hence, both peak power
and average daily energy per connection are directly derived from the load profile by knowing the overall
number of connections, which is also considered as a parameter, as it can have an impact on the shape of
normalized profiles; in fact, as the connections rise, the number of appliances increases as well, but the
likelihood they are switched on at the same time decreases [30]. The measurement age is the time difference
between the date of the commissioning of the microgrid and the date of measurement of the profile. This
information is important to compare the different profiles because during the time the socio-economic
characteristics of the community can change and so the load profile. Since the available measurements are
one-off “snapshots” of a certain phase of a mini-grid, four age classes have been defined, based on the time
between the commissioning of the system and the profile measurement.
Table 1 Tiers for power, energy system size and measurement age.
Tier
Daily peak power
per connection
Tier
Energy per connection
/day
0
x < 3 W
0
x < 12 Wh
1
3 W x < 50 W
1
12 Wh x < 200 Wh
2
50 W x < 200 W
2
200 Wh x < 1 kWh
Number of
Connections
Tier
Measurement
age [months]
x ≤ 35
VY
x ≤ 18
35 < x ≤ 200
Y
18 < x ≤ 36
200 < x ≤
550
MY
36 < x ≤ 60
6
3
200 W x < 800 W
3
1 kWh x < 3.4 kWh
4
800 W x < 2 kW
4
3.4 kWh x < 8.2 kWh
5
x 2 kW
5
x 8.2 kWh
x > 550
O
x > 60
Number of connections: VS, Very Small; S, Small; M, Medium; L, Large.
Measurement age: VY, Very Young; Y, Young; MY, Medium; O, Old
2.3.2. Qualitative indicators
Regarding qualitative data, the following details were investigated.
Climatic zone: by knowing the site location, it is easy to infer the type of climate. According to the
Köppen-Geiger climate classification, for each site, the respective climate zone is found thanks to the
high-resolution map available at [51].
The business model: the classification proposed concerning the business models is based on the
robust study reported in [8], focusing on the operator and considering only the mini-grid
management and not the ownership and financing of the system, which are intentionally neglected.
We refer to utility operators (U) when the energy provider is a state-owned company or a company
in which the state has a controlling interest; in this case, the mini-grid is operated with criteria very
similar to the national electricity network. The management can be done also by private operators
(P), or by local communities in the community-based models (C). Finally, hybrid operator models (H)
are possible, as a combination of the other three models.
The type of tariff: as tariff can influence the way energy is consumed, the proposed classification
also detailed its main characteristics, as done in [52]. The tariffs are categorised in energy-based (E),
depending on the actual energy consumed, and fee-for-service (FFS) when payment does not depend
on energy consumption but on the kind of service used [8] (in the cases investigated, they are flat
monthly charges). Moreover, we refer to a hybrid tariff structure when there is a monthly flat-rate
fee that depends on the scale of energy consumption (see Energy Daily Allowance in [53]).
The classification adopted reflects the actual data availability: for energy-based tariffs, whenever
possible, it is specified whether tariffs are pre-paid (E-pre) or post-paid (E-post) at the time of
measurement [52], or reported simply as E if the payment method was not specified in the source.
The same logic applies to the types FFS and H, for which the pre/post specification has been omitted
due to lack of data.
Human activities: the three principal energy uses in rural contexts are related to basic household
needs, community services and income-generating activities [54] that include both basic services for
small businesses (e.g. refrigeration) and the use of machinery in agriculture or manufacturing (e.g. a
maize mill) [55]. Ideally, the customer base composition, differentiated by the type of user and with
the specific features of productive activities would be needed for a complete characterization effort.
Moreover, the types of lifestyle and activities conducted in a village, the required appliances and
their usage patterns will define the shape of the load profile, which may change according to the
seasonality of energy needs and productive activities, which is then reflected on the ability to pay for
electricity [8]. However, in the literature at most a single hourly load profile for a given day is usually
available, not accompanied by quantitative details on the types of connections and users, neither
alongside precisely documented human activities. Therefore, it was not possible to obtain categories
related to human activities to be correlated with the load profile clusters. Nevertheless, a concise
description of the type of the connections, when possible, has been reported in Table A-1 Table A-5
of Appendix A to best comment the load shape of the reported communities.
7
2.3.3. Procedure
To resume, the procedure used to analyze the gathered information are as follows.
1. A first skimming of the gathered profiles is performed to exclude systems whose production profile
is not representative of the actual demand, like sites where generation has saturated.
2. The electrical profiles are clustered in a number of groups, in order to analyze the common
characteristics of their profile and extrapolate shared features. Then we estimated the duration
curve that best represents the average daily profile since duration curves can be useful to developers
for sizing purposes [56][58].
3. A first correlation analysis between the identified load profile clusters and the technical,
geographical, socio-economic characteristics of the site is developed.
3. Data collection and preliminary load model
3.1. Data collection
According to the review process described in the previous section, sixty-six measured load profiles have
been gathered and the corresponding information are reported in Table A-1 Table A-5 of Appendix A Among
these, five communities have been excluded from the clustering analysis (El Recreo in [42], Basse Santa Su in
[59] and the sites respectively presented in [60], [61], [62]) because, at the measurement moment, the supply
capacity was saturated and load curtailment was regularly performed, thus the measured electrical load did
not correspond to the desired demand.
3.2. Load model by clustering analysis
All the load profiles, once normalized by their mean hourly demand, have been clustered following the
procedure described above. Six major groups have been identified and described in Figure 1: the gray lines
are the typical profiles of the communities falling under the corresponding cluster, while the blue line
represents their average. The clusters peak1 and peak2 presented a similar behavior, as peak2 is basically
delayed. Therefore, given their resemblance, we considered them as a unique group that has been named
peak. In Table 2 the average peak values with their standard deviation are indicated. In Table B-1 to Table
B-5 of Appendix B , all actual values are reported, while in Table C-1 - of Appendix C the average profiles are
shown. By looking at the clusters, it’s clear how from the data emerges a more complex picture with respect
to what is usually intended as a typical load profile of a mini-grid system; in fact, in the literature the
archetype most commonly presented is akin to the step-peak cluster [8], [17], [63][65], while the other four
clusters are not generally represented. This schematization is usually justified by observing that a load profile
is composed of a constant baseload, a load increase during the day due to commercial activities, and an
evening peak due to the return of workers to their dwellings [8]; beyond productive uses, the night load is
attributed to either streetlighting or appliances continuously operating like fridges [17].
In the following, a detailed discussion of each cluster is developed, trying to identify the determinants of
this complex and diverse scenario.
The so-called flat cluster contains most profiles (24), the average curve is nearly constant and the trend
is stationary compared to the others. However, it is necessary to precise that the profiles included in this
cluster are not steadily flat, but they exhibit a predominant baseload. Indeed, typically, the load slightly
increases during the day and, in the evening, a relatively small peak occurs. Nevertheless, few profiles (MY1,
OM1, OM2, HN1, KE5) dissociate from this pattern but a relatively high night load characterises all.
The step cluster contains ten profiles that are characterised by a small load during the night and typically
two peaks. The first one occurs during the day (within 11 a.m. and 4 p.m.) and takes a hilly shape, while the
other one is the classical evening peak. This cluster is characteristic because the first peak is significative and
in two cases (TZ3, UG1) is even higher than the second one.
8
The step-peak cluster contains twelve profiles characterised by a small night load and typically two peaks:
a small one during working-hours and a bigger one during the evening. Conversely to the step cluster, in the
step-peak cluster, the evening peak is significantly higher than the former. Moreover, as shown in Table 2,
the specific values of the peak are higher than in the step case.
The peak cluster contains 11 profiles marked by a sharp evening peak. As visible in Table 2, the peak
values are greater than in the other groups. Generally, these curves are characterised by a little night load
and a slight demand increase during the day till the peak, that is by far higher than the remainder daily load.
The last cluster, named outliers, contains four profiles. The clustering procedure distinguishes them from
the others mainly for their late evening peak (10 ÷ 11 p.m.). Being aware of that, the MO2 profile could be
considered halfway between a step-peak and a peak one, while the other three could be classified as step-
peak.
Figure 1 - The six clusters obtained through the hierarchical procedure (in grey: each profile; in blue: the average curve)
Table 2 - Mean value and standard deviation of the peaks occurring in each cluster
flat
step
step-peak
Peak
outliers
Mean [-]
1.45
2.03
2.43
3.98
2.58
Std [-]
0.26
0.29
0.23
0.46
0.62
In Figure 2 the load duration curves of the chronological profiles shown in Figure 1 are presented. The
curves have been obtained by dividing the normalized profiles presented above by their normalized peak
power. Then, a model to fit each average curve is proposed. The logistic equation (1) has been selected
because it can represent accurately all the cases above mentioned.
  
  
(1)
The power (P) is expressed in per unit, dividing the values by the peak one; the time (t) is expressed in
per unit as well. The two constraints of the model are the peak power and the total energy consumed. In
Table 3, parameters and R2 values are reported.
9
Table 3 - R2 and parameters of the logistic models.
Cluster
R2
a
m
n
τ
flat
0.98
0.304043
-0.68233
-0.90874
131.3405
step
0.98
0.062444
93.19207
4.696574
5.548596
step-peak
0.98
0.008698
4.248138
-0.96195
131.0256
peak
0.97
0.000285
63.36079
-0.98924
132.12
outliers
0.96
0.008202
4.248139
-0.96455
131.0256
Figure 2 - The load duration curves of each cluster (in grey: each profile; in blue: the average curve; in red: the logistic curve)
4. Analysis of characterization factors of load profile clusters
In this section, the five load profile clusters identified in section Errore. L'origine riferimento non è stata
trovata. are compared with the technical, socio-economic and geographic data presented in section 2.3, to
identify for possible trends and relations.
4.1. Quantitative indicators
4.1.1. Daily peak power and energy per connection
Firstly, the distribution of load profile clusters among the tiers presented in Table 1 is presented. In Figure
3 and Figure 4, the peak power and energy per connection are focused, respectively.
10
Figure 3 - Peak power per connection shares across clusters
Figure 4 - Energy per connection shares across clusters
It is immediately evident the difference between the flat cluster and the others: the peak power per
connection and the energy per connection are mainly over tier 2. On the other hand, the step, step-peak and
peak clusters principally include sites characterised by tier 1. Every profile of the step cluster is in Tier 1 of
peak power per connection.
4.1.2. Number of connections
In Figure 5, details related to the number of connections are shown. As it is visible, the flat cluster tends
to include sites with both a large (tier L) and a small number of connections. The peak cluster seems to be
characterised by mini-grids with few users while the step-peak and the outliers ones tend to include sites in
49
100
83
82
25
38
17 9
7542
8
8
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
flat step step-peak peak outliers
0 1 2 3 4 5 na
Tier of peak power
per connection
4
60
75
91
25
8
40 8
9
50
29
17 25
21
29
8
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
flat step step-peak peak outliers
0 1 2 3 4 5 na
Tier of energy
per connection
11
the small interval. Finally, the step cluster represents a more varied range of sizes, since it includes also
medium systems.
As it could be presumed, the tendency that is suggested by the results is that a few connections lead to
a sharper profile while a large number of connections lead to a smoother curve. This has to do with the fact
that the higher the number of connections, the lower the coincidence factor [66]; the consequence is a flatter
profile as the connections become numerous. The latter could be one of the reasons why sites with many
connections mainly characterise the flat cluster. However, some small communities are present, which can
be justified by the tariff arrangement adopted, as discussed in section 4.2.3.
Figure 5 Number of connections shares across clusters
4.1.3. Measurement Age
The measurement age distribution among the clusters is presented in Figure 6. The non-available data
affects significantly this analysis, especially for the flat and outliers clusters. However, it appears that almost
all data are measured within the first five years since the mini-grid installation. Only the flat cluster includes
measurements older than five years (37.5%). Concerning the step, step-peak and peak clusters, the
measurements are mostly taken within the first three years. Moreover, it is visible a light blue ladder in Figure
6: ‘very young profiles’ are increasingly present moving from the flat towards the peak cluster, passing by
the step and the step-peak one.
Essentially, young sites tend to have a low night load and huge peaks, while communities that have been
electrified for several years usually have a flatter profile. Generally, this is justified with the increasing of
connections and appliances as time passes, which leads to a lower coincidence factor, as already explained.
An indication that users in the flat cluster could have more appliances is given by the energy consumption
and the peak power per connection. Indeed, the first one is principally greater or equal to the tier 3 of the
framework, while the second one is greater or equal to the tier 2. These details suggest that the number and
the size of the loads are higher than in the other cases.
12
Figure 6 - Measurements age shares across clusters
On the other hand, the peak profile is the key characteristic of the first three years of life of the mini-
grids operated by private companies and with a tiny number of connections. The energy consumption and
peak-power per connection are mostly within tier 1. These details suggest the presence of a reduced number
of loads of small size (e.g. lights, smartphone charger). Consequently, no anchor loads are present and the
sharp peak after the sunset is probably due to the use of lights and the operation of small activities, as
appliances stores. Indeed, this cluster is not only populated by sites with household-based demand but also
by sites with small business-based demand as visible for KE19, KE20 and KE21 (100%, 80% and 80% of the
demand related to small business, respectively).
4.1.4. Average energy consumption versus mini-grid size
To summarise the principal quantitative correlations, Figure 7 shows the average energy per connection
versus the number of connections for each site, presented as a log-log plot to enhance clarity. In the
rectangular area bordered by the 0.88 kWh line and the 80 connections line (represented by the green
dashed line), the 90.5% of the profiles belong to the flat cluster. The other sites of this cluster that are out of
this area are the HN1, HN2 and MY1 exceptions above mentioned. Moreover, it is visible that the peak
profiles are usually more concentrated in the bottom-left corner of the chart: in the orange dashed area
bordered by the 0.1 kWh line and the 39 connections line, the 70% of the profiles are within the peak cluster.
Regarding the step-peak and step profiles, they are randomly spread out within the two blue dashed lines (1
kWh and 600 connections).
8
30
50 64
13
50
42 27
25
17
20
50
38
25
89
25
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
flat step step-peak peak outliers
VY YMY Ona
Measurement age
13
Figure 7 - Distribution of clusters profiles in a chart reporting number of connections versus average energy consumption.
Concerning the empty area in the top-left corner, it could be populated by energy-kiosks, i.e. central
stations where energy is dispatched to recharge the villagers’ appliances [67]. However, this work addresses
mini-grids and not this kind of community charging stations. The bottom-right corner of the chart
theoretically includes large mini-grid systems with very low energy demand per connection. Further data
collection and analyses are required to ascertain if they are not present in the sample gathered, or if this area
of Figure 7 represents the inception stages of large systems before a ramp-up of demand takes place.
This diagram can be used as a guideline to design the most suitable business model and tariff structure
to guarantee profitability, given the energy sources and generation technologies available. On the other
hand, the findings reported in [36], based on operating and financial data from several mini-grid developers
based in sub-Saharan Africa, show that there is no clear correlation between the average revenues per user
obtained and the relative level of consumption, nor with the installed generation capacity. Therefore, given
the complexity of the analysis, further studies could investigate which area of the graph can be economically
profitable, under given business models local circumstances of the site, including political, social and
geographical conditions that can significantly affect the outcome of the project.
4.2. Qualitative indicators
4.2.1. Geographical position
The geographical locations cover many countries worldwide, as represented in Figure 8, with a
concentration of case studies from East Africa.
For the climate characterization of the mini-grids, the well-established Köppen-Geiger classification is
used, as presented in [51], and through the high-resolution map made available by the same authors. Out of
the 61 case studies, for 44 of them, the geographical coordinates were available, or otherwise, sufficient
information was supplied to identify the climate zone without ambiguities.
The results are reported in Figure 9: more than half of the mini-grids are located in hot arid climates,
about one-third in tropical climates, and a minority of cases in temperate ones, as expected considering that
most of the developing countries considered are located in the tropics.
14
Furthermore, the distribution of climate zones across clusters has been analyzed and presented in Figure
10, considering only the first-order climate classification (tropical, arid, temperate). The analysis did not
reveal clear, easily interpretable trends; the apparent prevalence of step-peak load profiles in arid climates,
for instance, may appear only because of the small sample size (7 step-peak cases with GIS data out of 12
total). Additionally, to establish a link between climatic features and load profiles, data on the implications
of a climate zone characteristics on the actual energy uses of a specific mini-grid would be needed, like water
provisioning, cooling comfort, etc.
Finally, in this first analysis, only the climate zone characteristics of the immediate surroundings of a mini-
grid have been considered, but it has to be considered that also the characteristics of its neighboring areas
may be relevant, e.g., on agricultural and commercial activities, and impact the energy use patterns at a
certain time scale.
While in certain contexts the climate situation is very homogenous, as is for the cases OM1 and OM2, in
the surroundings of the Kenyan-Tanzanian border, for example, there are frequent climate zone alternations
between arid, tropical and temperate. Finally, if the extent of the mini-grid is large enough, as is for the case
ID1, there might be a presence of more sub-climate zones within the domain of a single system.
Figure 8 - Distribution map of the investigated sites with the type of cluster according to Figure 1
15
Figure 9 - Climate zones of the sites investigated according to [51]
Figure 10 Climate zones across clusters
4.2.2. Business model
In Figure 11 and Figure 12, details related to the business model are shown. It is worth noticing that in
the case of a flat profile the energy provider is usually a utility, while in the step, step-peak and peak cases
the operator is typically private. Among the case studies, only six are community-based mini-grids that are
divided among outliers, flat and step clusters.
16
Figure 11 - Operator shares across clusters
The flat profile is typical of mini-grids operated by utilities, and, as seen above, is also common for older
projects with a large customer base. It is then possible to infer that flat profiles are, in part, typical of large,
established electrification projects developed by national utilities, whereas private sector investments are
generally more recent and focused on smaller communities.
4.2.3. Type of tariff
Concerning the tariff, apart from three cases of fee-for-service and one case where energy is for free,
they are all energy-based. It is visible that sites with fee-for-service tariff correspond to a flat profile. In the
cases evaluated, a step, step-peak or peak profile is principally present when consumption is pre-paid.
Figure 12 Tariff arrangement across clusters
17
This consideration can also explain why some small mini-grids exhibit a flat profile, as resulted from
Figure 5. In fact, for some of these cases (TZ1, HN1, HN2), as suggested in [28], [52] and thanks to the
experience of the authors in the sites HN1 and HN2, it can be argued that a monthly fixed fee may affect the
behaviour of the customers that tend to keep the appliances switched on also when they do not need their
services, like during the night. This effect could be even more present for the case MY1, where the electricity
is for free.
A similar argument holds for step-peak and step clusters, since they are both populated by sites
characterised by peak power and energy per connection mostly within tier 1 and, as for the peak cluster, are
mainly operated by private companies and tariffs are energy-based with the prepayment modality.
Conversely, in the flat cluster utilities or communities operate mini-grids, the tariffs are energy-based or fee-
for-service. As a result, it appears that energy-based, prepaid tariffs are more efficient than the other models
that lead to a flatter profile, probably partially caused by the non-attention on switching off devices [52].
5. Final considerations
The results of the clustering exercise confirm that the typical mini-grids load profile shape, given by a low
night demand, some daily activities and a more pronounced evening peak, is well represented across the
dataset, denoted as step-peak cluster. However, three interesting other shapes have been reported and
classified: the step one, in which the daily and evening peak have similar intensity, a flat one, whose demand
has a prevalent baseload component, and a peak one, where limited activities occur during the night,
whereas there is a sharp evening peak.
The flat cluster is the most clearly characterized by the selected quantitative and qualitative indicators,
as it usually occurs in a definite window of high daily energy consumed and large number of connections,
indicating the flattening effect of a large mini-grid size on the load profile. Such kind of systems are generally
installed and operated by national utilities and tend also to be older, suggesting that the measurement
reflects a mature customer take-off and consumption ramp-up phases. Nevertheless, also few small sites
with fee-for-service tariffs fall under this cluster.
The sites in the other clusters have a similar level of energy and power demand (principally 12÷200
Wh/(connection·day) and 3÷50 W/connection). Besides, they are mainly operated by private companies
adopting prepayment methods of purchasing energy. The peak cluster is typical in very small and recently
commissioned sites, suggesting that the two factors of the small consumer base and limited appliance
diffusion over time both concur to a low diurnal baseload and high evening coincidence factor.
The difference between the step and step-peak clusters seems more elusive: on the one hand, the step-
peak cluster is more common in very recent installations, leading to infer that the step cluster can be seen as
a passage towards a flat profile. As for the system size, the step profile is more common among very small
systems, but at the same time featured in medium ones, differently from the step-peak one. Since the core
difference among the step and step-peak clusters lies on the weight of daily and evening activities in the
community, it could be better analysed with a larger dataset including social data, disaggregated load data
among customer classes and information on the types of appliances used and relative productive activities,
which currently is hardly available according to the best knowledge of the authors.
Similarly, the geographical location and climate zone analysis have not shown any particular correlations
with the load shapes, both due to a lack of data - GPS coordinates to establish climate zone were lacking in
28% of the case studies and to lack of basic evidence on how climatological features can affect the energy
need patterns of rural communities.
18
6. Conclusions
Electricity access for all by 2030 is a challenging goal to achieve. The installation of mini-grids in isolated
areas of developing countries is a promising solution for rural electrification, given its advantages in terms of
flexibility, reliability and quality of service, enabling also productive uses of energy. However, the estimation
of the energy demand, required to properly design the energy system, is still a complex task with no
standardized methodology and very limited data are available. This work aims to propose a database of
measured load profiles of rural microgrids and perform a data-driven clustering analysis to identify typical
load patterns and correlations among major technical and non-technical determinants. The dataset has also
been populated with technical, geographic and socio-economic information for each selected site.
As a result, sixty-one measured load profiles are collected and, through a hierarchical clustering
procedure, four reference curves have been obtained and named in order to concisely describe their shape
(flat, step, step-peak, peak). Besides, the related load duration curves, combined with appropriate curve
fitting, are provided. The analyses showcase a heterogeneous picture of different mini-grid projects for
electricity access promoted around the world. For instance, mini-grids with a flat profile are generally larger
and older than the rest, with generally higher average energy and power tiers. More recent and smaller mini-
grids tend to have a more pronounced evening peak; and there is a clear prevalence of privately developed
mini-grids in non-flat clusters.
Therefore, the proposed data-driven archetypal profiles, which is a novelty in the practitioner and
scientific literature, can give developers a preliminary load estimation for new projects, even when no data
are available. The correlation between the load profile and the age of measurement (the time between mini-
grid commitment and the date of measurement) of the load profile also highlights a dynamic evolution of
consumer habits and demand patterns, which means that the first installation at year 0 may need to be
flexible in accommodating variations in both shape and intensity of the profile over time.
This activity proposes a preliminary data-driven analysis on a dataset of sixty-one mini-grids, which is
expected to serve researchers and practitioners for developing further studies, hopefully including a larger
dataset built upon the proposed one. Further studies can investigate the relationship between socio-
economic, political and geographic characteristics of a community and the load demand evolution, which is
difficult to model and validate. In particular, in order to maximize the socio-economic impact, research on
the effective policies to foster local productive uses of electricity and set the optimal tariff scheme is
recommended. To generalize the results and deepen the analysis, a large dataset is required, also
accompanied by detailed socio-economic data, hopefully in close collaboration with private companies.
19
7. References
[1] IRENA, Global Renewables Outlook: Energy Transformation 2050, 2020th ed. Abu Dhabi: International
Renewable Energy Agency, 2020.
[2] High-level political forum on sustainable development, Accelerating SDG7 Achievement. Policy Briefs
in support of the first SDG7 review at the UN high-level politcal forum 2018. United Nations, 2018.
[3] F. Fuso Nerini et al., “Mapping synergies and trade-offs between energy and the Sustainable
Development Goals,” Nat. Energy, vol. 3, no. 1, pp. 1015, 2018, doi: 10.1038/s41560-017-0036-5.
[4] High-level political forum on sustainable development, Accelerating SDG7 Achievement. Policy Briefs
in support of the first SDG7 review at the UN high-level politcal forum 2019. United Nations, 2019.
[5] United Nations, The sustainable development goals report 2019. New York: UN, 2019.
[6] International Energy Agency (IEA), World Energy Outlook 2019. Paris: IEA, 2019.
[7] A. Berizzi, M. Delfanti, D. Falabretti, S. Mandelli, and M. Merlo, “Electrification processes in developing
countries: Grid expansion, microgrids, and regulatory framework,” Proc. IEEE, vol. 107, no. 9, pp.
19811994, 2019, doi: 10.1109/JPROC.2019.2934866.
[8] M. Franz, N. Peterschmidt, M. Rohrer, and G. Kondev, “Mini-grid Policy Toolkit: Policy and Business
Frameworks for Successful Mini-grid Roll-outs,” 2014.
[9] International Renewable Energy Agency (IRENA), Innovation Outlook: Renewable Mini-Grids. 2016.
[10] N. J. Williams, P. Jaramillo, J. Taneja, and T. S. Ustun, “Enabling private sector investment in microgrid-
based rural electrification in developing countries: A review,” Renew. Sustain. Energy Rev., vol. 52, pp.
12681281, 2015, doi: 10.1016/j.rser.2015.07.153.
[11] ESMAP, Mini Grids for Half a Billion People: Market Outlook and Handbook for Decision Makers.
Executive Summary. Washington, DC: Energy Sector Management Assistance Program (ESMAP)
Technical Report 014/19. World Bank, 2019.
[12] IRENA, “Accelerating Off-grid Renewable Energy: Key Findings and Recommendations from IOREC
2016,” IOREC, 2016, pp. 124, 2017, doi: 978-92-95111-07-3.
[13] African Development Bank Group, Sustainable Energy For All - Africa Hub, and Sustainable Energy
Fund for Africa, “Green Mini-Grids in Sub-Saharan Africa: Analysis of Barriers to Growth and the
Potential Role of the African Development Bank in Supporting the Sector,” GMG MDP Doc. Ser. n°1,
no. December, 2016.
[14] D. Manetsgruber, B. Wagemann, B. Kondev, and K. Dziergwa, Risk Management for Mini-Grids: A new
approach to guide mini-grid deployment. Alliance for Rural Electrification (ARE), 2015.
[15] J. Peters, M. Sievert, and M. A. Toman, “Rural electrification through mini-grids: Challenges ahead,”
Energy Policy, vol. 132, pp. 2731, Sep. 2019, doi: 10.1016/j.enpol.2019.05.016.
[16] Energy Sector Management Assistance Program (ESMAP), “Mini-grid design Manual,” 2000.
[17] The German Climate Technology Initiative and GIZ Promotion of Solar-Hybrid Mini-Grids, “What size
shall it be? - A guide to mini-grid sizing and demand forecasting,” 2016.
[18] C. Muhoza and O. W. Johnson, “Exploring household energy transitions in rural Zambia from the user
perspective,” Energy Policy, vol. 121, no. January, pp. 2534, 2018, doi: 10.1016/j.enpol.2018.06.005.
[19] K. Tamir, T. Urmee, and T. Pryor, “Energy for Sustainable Development Issues of small scale renewable
energy systems installed in rural Soum centres in Mongolia,” Energy Sustain. Dev., vol. 27, pp. 19,
2015, doi: 10.1016/j.esd.2015.04.002.
[20] S. Mandelli, C. Brivio, E. Colombo, and M. Merlo, “Effect of load profile uncertainty on the optimum
sizing of off-grid PV systems for rural electrification,” Sustain. Energy Technol. Assessments, vol. 18,
pp. 3447, 2016, doi: 10.1016/j.seta.2016.09.010.
[21] H. Louie and P. Dauenhauer, “Effects of load estimation error on small-scale off-grid photovoltaic
system design, cost and reliability,” Energy Sustain. Dev., vol. 34, pp. 3043, 2016, doi:
10.1016/j.esd.2016.08.002.
[22] V. Gambino, R. Del Citto, P. Cherubini, C. Tacconelli, A. Micangeli, and R. Giglioli, “Methodology for
the Energy Need Assessment to Effectively Design and Deploy Mini-Grids for Rural Electrification,”
Energies 2019, Vol. 12, Page 574, vol. 12, no. 3, p. 574, 2019, doi: 10.3390/EN12030574.
[23] D. Fioriti, R. Giglioli, and D. Poli, “Short-term operation of a hybrid minigrid under load and renewable
20
production uncertainty,” in GHTC 2016 - IEEE Global Humanitarian Technology Conference:
Technology for the Benefit of Humanity, Conference Proceedings, 2016, doi:
10.1109/GHTC.2016.7857317.
[24] S. Feron, R. Cordero, and F. Labbe, “Rural Electrification Efforts Based on Off-Grid Photovoltaic
Systems in the Andean Region : Comparative Assessment of Their Sustainability,” 2017, doi:
10.3390/su9101825.
[25] J. D. D. Niyonteze, F. Zou, and G. N. Osarumwense, “Solar-powered mini-grids and smart metering
systems , the solution to Rwanda energy crisis Solar-powered mini-grids and smart metering systems
, the solution to Rwanda energy crisis,” 2019, doi: 10.1088/1742-6596/1311/1/012002.
[26] H. Ahlborg and M. Sjöstedt, “Small-scale hydropower in Africa: Socio-technical designs for renewable
energy in Tanzanian villages,” Energy Res. Soc. Sci., vol. 5, pp. 2033, Jan. 2015, doi:
10.1016/j.erss.2014.12.017.
[27] A. Pueyo, M. Carreras, and G. Ngoo, “Exploring the linkages between energy , gender , and enterprise :
Evidence from Tanzania,” World Dev., vol. 128, p. 104840, 2020, doi:
10.1016/j.worlddev.2019.104840.
[28] E. Hartvigsson and E. O. Ahlgren, “Comparison of load profiles in a mini-grid: Assessment of
performance metrics using measured and interview-based data,” Energy Sustain. Dev., vol. 43, pp.
186195, 2018, doi: 10.1016/j.esd.2018.01.009.
[29] C. Blodgett, P. Dauenhauer, H. Louie, and L. Kickham, “Accuracy of energy-use surveys in predicting
rural mini-grid user consumption,” Energy Sustain. Dev., vol. 41, pp. 88105, 2017, doi:
10.1016/j.esd.2017.08.002.
[30] P. Boait, V. Advani, and R. Gammon, “Estimation of demand diversity and daily demand profile for off-
grid electrification in developing countries,” Energy Sustain. Dev., vol. 29, pp. 135141, 2015, doi:
10.1016/j.esd.2015.10.009.
[31] S. Mandelli, M. Merlo, and E. Colombo, “Novel procedure to formulate load profiles for off-grid rural
areas,” Energy Sustain. Dev., vol. 31, pp. 130142, 2016, doi: 10.1016/j.esd.2016.01.005.
[32] F. Lombardi, S. Balderrama, S. Quoilin, and E. Colombo, “Generating high-resolution multi-energy load
profiles for remote areas with an open-source stochastic model,” Energy, vol. 177, pp. 433444, 2019,
doi: 10.1016/j.energy.2019.04.097.
[33] N. Narayan et al., “Stochastic load profile construction for the multi-tier framework for household
electricity access using off-grid DC appliances,” Energy Effic., no. Iea 2017, 2018, doi: 10.1007/s12053-
018-9725-6.
[34] X. Li et al., Performance Monitoring of African micro-grids: good practices and operational data, no.
January. NREL, 2020.
[35] A. Yakubu, E. Ayandele, J. Sherwood, A. O. Olu, and S. Graber, Minigrid investment report. Scaling the
Nigerian Market. Abuja: The Nigerian Economic Summit Group (NESG), 2018.
[36] Africa Minigrid Developers Association (AMDA) and Economic Consulting Associates (ECA), “Africa
Minigrid Benchmarking Report,” Nairobi, 2020.
[37] N. Fall, L. Giles, B. Marchionini, and E. G. Skolnik, “Remote Area Power Supply (RAPS) Load and
Resource Profiles A Study for the DOE Energy Storage Program,” SAND Rep. SAND2007-4268 Unltd.
Release Print. July 2007, no. July, p. 47, 2007.
[38] G. J. Prinsloo, R. Dobson, and A. Brent, “Scoping exercise to determine load profile archetype
reference shapes for solar co-generation models in isolated off-grid rural African villages,” J. Energy
South. Africa, vol. 27, no. 3, pp. 1127, 2016, doi: 10.17159/2413-3051/2016/v27i3a1375.
[39] N. J. Williams, P. Jaramillo, B. Cornell, I. Lyons-Galante, and E. Wynn, “Load characteristics of East
African microgrids,” Proc. - 2017 IEEE PES-IAS PowerAfrica Conf. Harnessing Energy, Inf. Commun.
Technol. Afford. Electrif. Africa, PowerAfrica 2017, pp. 236241, 2017, doi:
10.1109/PowerAfrica.2017.7991230.
[40] N. J. Williams, P. Jaramillo, K. Campbell, B. Musanga, and I. Lyons-Galante, “Electricity Consumption
and Load Profile Segmentation Analysis for Rural Micro Grid Customers in Tanzania,” 2018 IEEE
PES/IAS PowerAfrica, PowerAfrica 2018, pp. 360365, 2018, doi: 10.1109/PowerAfrica.2018.8521099.
[41] T. Reber, S. Booth, D. Cutler, X. Li, J. Salasovich, and W. Ratterman, “Tariff Considerations for Micro-
21
Grids in Sub-Saharan Africa,” 2018.
[42] “Own elaborations thanks to installation of three energy meters in the communities of Quinito, El
Dictamo and El Recreo.” 2018.
[43] “Anonymous source.” 2019.
[44] C. Kuster, Y. Rezgui, and M. Mourshed, “Electrical load forecasting models: A critical systematic
review,” Sustain. Cities Soc., vol. 35, no. July, pp. 257270, 2017, doi: 10.1016/j.scs.2017.08.009.
[45] RES4Africa Foundation, Field Studies for Mini-Grid Optimization International Research Group, ERM,
FAO, RINA, and PwC, RE-thinking Access to Energy Business Models. Ways to Walk the Water-Energy-
Food Nexus Talk in Sub-Saharan Africa. Rome: Gangemi Editore, 2019.
[46] X. Wang, K. Smith, and R. Hyndman, “Characteristic-based clustering for time series data,” Data Min.
Knowl. Discov., vol. 13, no. 3, pp. 335364, 2006, doi: 10.1007/s10618-005-0039-x.
[47] G. Le Ray, E. M. Larsen, and P. Pinson, “Evaluating Price-Based Demand Response in Practice - With
Application to the EcoGrid EU Experiment,” IEEE Trans. Smart Grid, vol. 9, no. 3, pp. 23042313, 2018,
doi: 10.1109/TSG.2016.2610518.
[48] G. Chicco, R. Napoli, and F. Piglione, “Comparisons Among Clustering Techniques for Electricity
Customer Classificatio,” vol. 21, no. 2, pp. 933–940, 2006.
[49] G. Chicco, “Overview and performance assessment of the clustering methods for electrical load
pattern grouping,” Energy, vol. 42, no. 1, pp. 6880, 2012, doi: 10.1016/j.energy.2011.12.031.
[50] ESMAP, M. Bhatia, and N. Angelou, “Beyond Connections: Energy Access Redefined,” 2015.
[51] H. E. Beck, N. E. Zimmermann, T. R. McVicar, N. Vergopolan, A. Berg, and E. F. Wood, “Present and
future Köppen-Geiger climate classification maps at 1-km resolution,” Sci. Data, vol. 5, no. 1, p.
180214, Dec. 2018, doi: 10.1038/sdata.2018.214.
[52] B. Tenenbaum, C. Greacen, T. Siyambalapitiya, and J. Knuckles, From the Bottom Up How Small Power
Producers and Mini-Grids Can Deliver Electrification and Renewable Energy in Africa, vol. 17, no. 4.
2014.
[53] Economic-Consulting-Associates, TramaTecnoAmbiental, and Access-Energy, “Project Design Study
on the Renewable Energy Development for Off-Grid Power Supply in Rural Regions of Kenya, Project
no. 30979,” 2014.
[54] S. Mandelli, J. Barbieri, R. Mereu, and E. Colombo, “Off-grid systems for rural electrification in
developing countries: Definitions, classification and a comprehensive literature review,” Renew.
Sustain. Energy Rev., vol. 58, pp. 16211646, 2016, doi: 10.1016/j.rser.2015.12.338.
[55] GIZ and EUEI PDF, “Productive Use of Energy PRODUSE. A Manual for Electrification Practitioners,”
2011.
[56] S. Barsali, P. Di Marco, S. Filippeschi, A. Franco, R. Giglioli, and D. Poli, “Dimostratore di casa attiva,”
2011.
[57] A. S. Jacob, R. Banerjee, and P. C. Ghosh, Sizing of hybrid energy storage system for a PV based
microgrid through design space approach,” Appl. Energy, vol. 212, no. November 2017, pp. 640653,
2018, doi: 10.1016/j.apenergy.2017.12.040.
[58] A. Poulin, M. Dostie, M. Fournier, and S. Sansregret, “Load duration curve: A tool for technico-
economic analysis of energy solutions,” Energy Build., vol. 40, no. 1, pp. 2935, Jan. 2008, doi:
10.1016/j.enbuild.2007.01.020.
[59] F. S. U. C. C. for C. and S. E. Finance, “Renewable Energy in Hybrid Mini-Grids and Isolated Grids:
Economic Benefits and Business Cases,” p. 88, 2015.
[60] M. Solano-peralta, M. Moner-girona, W. G. J. H. M. Van Sark, and X. Vallve, “‘“ Tropicalisation ”’ of
Feed-in Tariffs : A custom-made support scheme for hybrid PV / diesel systems in isolated regions,”
vol. 13, pp. 22792294, 2009, doi: 10.1016/j.rser.2009.06.022.
[61] R. M. Moharil and P. S. Kulkarni, “A case study of solar photovoltaic power system at Sagardeep Island,
India,” Renew. Sustain. Energy Rev., vol. 13, no. 3, pp. 673681, 2009, doi: 10.1016/j.rser.2007.11.016.
[62] H. M. H. Gabler, G. Bopp, F. Haugwitz, Liu Hong, Li Zhiming, “PV village power supply systems in China
- results from a technical monitoring campaign,” no. 1.
[63] G. Léna, Rural Electrification with PV Hybrid Systems. Overview and Recommendations for Further
Deployment. IEA PVPS Task 9, Subtask 4, Report IEA-PVPS T9-13:2013, 2013.
22
[64] Energy 4 Impact and INENSUS, “Billing, Revenue Collection and Metering Models for Mini-Grids,”
2019.
[65] HOMER Energy, “Creating a Synthetic Load from a Profile.” [Online]. Available:
https://www.homerenergy.com/products/pro/docs/latest/creating_a_synthetic_load_from_profile.
html. [Accessed: 25-Oct-2019].
[66] V. Cataliotti, Electrical systems (Vol.1): Generality components. 2005.
[67] H. Louie, M. Shields, S. J. Szablya, L. Makai, and K. Shields, “Design of an off-grid energy kiosk in rural
Zambia,” Proc. 5th IEEE Glob. Humanit. Technol. Conf. GHTC 2015, pp. 16, 2015, doi:
10.1109/GHTC.2015.7343946.
[68] S. L. Balderrama Subieta, A. Tarantino, S. Sabatini, F. Riva, G. Bonamini, and S. Quoilin, “Feasibility
Study of PV & Li-Ion Battery Based Micro-Grids for Bolivian Off-Grid Communities,” Proc. IRES 2017 -
11th Int. Renew. Energy Storage Conf., 2017.
[69] D. Corbus, C. Newcomb, and Z. Yedall, “San Juanico hybrid power system technical and institutional
assessment. To be presented,” World Renew. Energy Congr. VIII, Denver, Color. August 29 - Sept. 3,
2004, no. July, 2004.
[70] E. Hartvigsson, J. Ehnberg, E. Ahlgren, and S. Molander, “Assessment of load profiles in minigrids: A
case in Tanzania,” Proc. Univ. Power Eng. Conf., vol. 2015-Novem, 2015, doi:
10.1109/UPEC.2015.7339818.
[71] A. H. Al-badi, M. Al-Toobi, S. AL-Harthy, Z. Al-Hosni, and A. AL-Harty, “Hybrid systems for decentralized
power generation in Oman,” vol. 6451, no. July, 2016, doi: 10.1080/14786451.2011.590898.
[72] S. Al Mughairi, “The official magazine of rural areas electricity company,” 2014. .
[73] C. L. Azimoh, P. Klintenberg, C. Mbohwa, and F. Wallin, “Replicability and scalability of mini-grid
solution to rural electrification programs in sub-Saharan Africa,” Renew. Energy, vol. 106, pp. 222
231, 2017, doi: 10.1016/j.renene.2017.01.017.
[74] S. Y. Wong and A. Chai, “An Off-grid Solar System for Rural Village in Malaysia,” pp. 0–3, 2012.
[75] Carbon Africa Limited, Trama TecnoAmbiental S.L., Research Solutions Africa Limited, and Energy
research Centre of the Netherlands, “Kenya Market Assessment for Off-Grid Electrification,” 2015.
[76] P. R. Oimeke, “Approval of the schedule of tariffs set by the energy regulatory commission for supply
of electrical energy by the Kenya power and lightining company limited pursuant to section 45 of the
energy act, 2006.” .
[77] Absolute Energy, “Appliances in Kitobo island.” [Online]. Available: http://absolute-
nrg.com/appliances-kitobo/. [Accessed: 15-Oct-2019].
[78] L. Barelli, G. Bidini, P. Cherubini, A. Micangeli, D. Pelosi, and C. Tacconelli, “How Hybridization of
Energy Storage Technologies Can Provide Additional Flexibility and Competitiveness to Microgrids in
the Context of Developing Countries,” Energies, vol. 12, no. 16, p. 3138, 2019, doi:
10.3390/en12163138.
[79] Vulcan and SteamaCo, “Powering Productivity. Early Insights into Mini Grid Operations in Rural
Kenya.,” 2016.
[80] A. B. Bahaj et al., “The impact of an electrical mini-grid on the development of a rural community in
Kenya,” Energies, vol. 12, no. 5, pp. 121, 2019, doi: 10.3390/en12050778.
[81] A. El Fathi et al., “Performance parameters of a standalone PV plant,” Energy Convers. Manag., vol.
86, no. July, pp. 490495, 2014, doi: 10.1016/j.enconman.2014.05.045.
[82] L. Navarte and E. L. Pigueiras, “Experience with PV‐diesel hybrid village power systems in Southern
Morocco,” no. September, 2007, doi: 10.1002/pip.756.
[83] S. Paul Ayeng’o, T. Schirmer, K. P. Kairies, H. Axelsen, and D. Uwe Sauer, “Comparison of off-grid power
supply systems using lead-acid and lithium-ion batteries,” Sol. Energy, vol. 162, no. May 2017, pp.
140152, 2018, doi: 10.1016/j.solener.2017.12.049.
23
In the following Table A-1 Table A-5, data related to the sites investigated are gathered, divided in the columns:
Code: A three-digit code to identify each site. The first two letters indicate the country, followed by a progressive number.
Country: The Country where the site is located.
Village/Area: The name of the village or the area where the mini-grid is located.
Cli. zone: The climate zone according to the Köppen-Geiger classification, as reported in Figure 9.
Size range [kW], Technology: The size of capacity power installed [kW] and the type of technology (PV - Photovoltaic, B - Battery, H - Hydropower, D -
Diesel generator; Wn - Wind turbine).
Op.: Type of operator as introduced in Section 2.3.2.
Tariff: Type of tariff as introduced in Section 2.3.2.
Tier A: Tier of measurement age, as introduced in Table 1.
Tier C.: Tier of the number of connections, as introduced in Table 1.
Tier P.: Tier of peak-power per connection, as introduced in Table 1
Tier E.: Tier of daily energy per connection, as introduced in Table 1.
Brief connections description: A brief description of the type of users connected to the mini-grid, where: HH: Household, B: Business activities.
Sources: Bibliographic reference(s)
In all cases, na stands for non-available.
Table A-1 Dataset of mini-grids in the flat cluster
Code
Country
Village/Area
Cli.
zone
Size range
[kW],
Technology
Op.
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
BO1
Bolivia
El Espino
Bsh
60 PV, 58 D
B
na
E-pre
VY
S
2
3
Mostly HH. In addition, a hospital, a
school and public lighting service.
[32], [68]
MX1
Mexico
San Juanico
BWh
17 PV, 70 Wn,
80 D, B
U
E
MY
S
3
4
Mostly HH. In addition, 5 7 grocery
shops, a few mechanics, 2 restaurants,
2 beer depositories, a small clinic and
street lightning.
[37], [69]
TZ1
Tanzania
southwestern
highlands
Cwb
120 H
U
FFS
O
M
3
4
Mostly HH and stores. In addition, a
hospital, a small college, five mills and
three workshops.
[28], [70]
24
Code
Country
Village/Area
Cli.
zone
Size range
[kW],
Technology
Op.
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
OM1
Oman
Hameed
BWh
100 D
U
E
na
na
na
na
na
[71]
OM2
Oman
Masirah
BWh
10,597 D
U
E
na
L
5
5
HH, stores, hotels, restaurants, cafes
and fishing activities.
[71],[72]
NA1
Namibia
Tsumkwe
Bsh
202 PV, 300 D
U
E
MY
L
2
3
Mostly HH. In addition, B, a clinic, a
water pump, a school, a police station.
[73]
MY1
Malaysia
Long beruan -
Ulu Baram
Af
54 PV, B
C
For
free
VY
S
0
1
Mostly HH. In addition, a church and a
public centre.
[74]
MW1
Malawi
Mulanje
district
na
na
na
na
na
na
na
na
na
[30]
KE1
Kenya
Hola
Bsh
60 PV, 800 D
U
E-
post
MY
L
3
4
HH (881 low-consumers*, 731 medium-
consumers, 4 high-consumers), 340 B
and industrial. 10% of the generated
electricity is destinated to an
agricultural irrigation scheme that
operates day and night.
[59],[75],
[76]
ID1
Indonesia
Nusa Penida
Am
3650 D, (800
Wn, 80 PV
non-working)
U
E
O
L
2
3
Mostly HH; there are only a handful of
shops, small resorts, government
offices, a gas station and a local bank,
adding relatively little demand.
[59]
HN1
Honduras
Quinito
Am
24 H
C
FFS
O
S
2
4
Mostly HH. In addition, 8 B, a school, a
kindergarten, a medical centre, two
churches and a community centre.
[42]
HN2
Honduras
El Dictamo
As
15 H
C
FFS
Y
S
2
2
Mostly HH. In addition, 4 B, a school, a
kindergarten, two churches and a
parish all.
[42]
PH1
Philippines
Busuanga
island
Am
3580 D
H
E
na
L
2
3
HH (89% of customers; 30% of
demand), B (8% of customers; 45% of
demand), public facilities including
streetlights (2% of customers; 14% of
demand), and some industry (0.04% of
customers; 11% of demand).
[59]
25
Code
Country
Village/Area
Cli.
zone
Size range
[kW],
Technology
Op.
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
DO1
Dominican
Rep.
Las Terrenas
Af
9500 D
U
E
na
L
3
5
Low-income HH (62% of customers;
23% of demand), but also high-income
HH, including holiday homes (24% of
customers; 28% of demand), as well as
resorts, hotels, hospitals, banks, and
other B (14% of customers; 48% of
demand).
[59]
CO1
Colombia
Puerto
Leguizamo
Af
4200 D
U
E
na
L
3
5
Mainly HH (93% of customers; 35% of
demand), 200 businesses (hotels,
restaurants, banks, shops) and public
facilities (e.g. hospital, airport; 7% of
customers; 22% of demand). The bulk
of electricity, though, is consumed by
the Navy as a single anchor client (42%
of demand).
[59]
VC1
Saint
Vincent
Bequia
Af
4100 D
U
E
O
L
3
5
87% HH and 13% B. There is no
industrial client; two existing public
streetlight systems only have
neglectable demand.
[59]
KE2
Kenya
Baragoi
Bsh
248 D
U
E-
post
O
M
3
3
HH (104 low-consumers*, 86 medium-
consumers), 40 B and industrial.
[75], [76]
KE3
Kenya
Eldas
BWh
184 D
U
E-
post
Y
S
3
5
HH (36 low-consumers*, 30 medium-
consumers), 14 B and industrial.
[75], [76]
KE4
Kenya
Elwak
BWh
680 D, 60 PV
U
E-
post
O
L
3
4
HH (361 low-consumers*, 300 medium-
consumers, 2 high-consumers), 139 B
and industrial.
[75], [76]
KE5
Kenya
Lodwar
BWh
2740 D, 60 PV
U
E-
post
O
L
2
3
HH (1073 low-consumers*, 889
medium-consumers, 5 high-
consumers), 414 B and industrial.
[75], [76]
KE6
Kenya
Lokichoggio
Bsh
680 D
U
E-
post
Y
S
4
5
HH (75 low-consumers*, 62 medium-
consumers), 29 B and industrial.
[75], [76]
26
Code
Country
Village/Area
Cli.
zone
Size range
[kW],
Technology
Op.
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
KE7
Kenya
Merti
Bsh
240 D, 10 PV
U
E-
post
O
M
2
3
HH (196 low-consumers*, 163 medium-
consumers, 1 high-consumer), 76 B and
industrial.
[75], [76]
KE8
Kenya
Mfangano
Aw
520 D
U
E-
post
MY
S
3
5
HH (54 low-consumers*, 45 medium-
consumers), 21 B and industrial.
[75], [76]
KE9
Kenya
Wajir
BWh
3400 D
U
E-
post
O
L
2
2
HH (1848 low-consumers*, 1532
medium-consumers, 8 high-
consumers), 713 B and industrial.
[75], [76]
Table A-2 Dataset of mini-grids in the step cluster
Code
Country
Village/Area
Cli.
zone
Size range [kW],
Technology
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
KE10
Kenya
na
na
3**
PV, B (back-up D
may be present)
E-pre
Y
VS
1
1
Only B.
[39]
TZ2
Tanzania
na
Cfb
6
PV, B
E-pre
Y
VS
1
1
Only B.
[43]
TZ3
Tanzania
na
Aw
6
PV, B (and not
specified D)
E-pre
VY
S
1
1
na
[43]
TZ4
Tanzania
na
Cwb
15.9 (TZ12 + TZ4)
PV, B
E-pre
VY
S
1
1
na
[43]
TZ5
Tanzania
na
na
7.5 (TZ5 + TZ15)
PV, B
E-pre
MY
S
1
2
Only HH.
[43]
TZ6
Tanzania
na
Cfb
6
PV, B
E-pre
Y
VS
1
1
Only HH.
[43]
UG1
Uganda
Kitobo
Af
228
PV, 80 D, B
E-pre
Y
M
1
2
Mostly HH. In addition, few B and an
anchor load consisting of an ice
machine.
[77],
[78]
KE11
Kenya
Entesopia
Cfb
1.5 ÷ 5.6
PV, B
E-pre
Y
S
1
2
na***
[29],[79]
27
Code
Country
Village/Area
Cli.
zone
Size range [kW],
Technology
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
KE12
Kenya
Kitonyoni
Bsh
13
PV, B
E
MY
S
1
2
Mostly B.
[80]
KE13
Kenya
Enkoireroi
Aw
1.5 ÷ 5.6
PV, B
E-pre
na
VS
1
1
na***
[29],[79]
Table A-3 Dataset of mini-grids in the step-peak cluster
Code
Country
Village/Area
Cli.
zone
Size range [kW],
Technology
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
KE14
Kenya
na
na
3 ÷ 6** PV, B
E-pre
VY
S
1
1
Distribution of connections type
roughly: 83% B, 7% HH, 10% HH + B.
[39]
KE15
Kenya
na
na
3** PV, B
E-pre
Y
VS
1
1
Distribution of connections type
roughly: 69% B, 26% HH, 5% HH + B.
[39]
TZ7
Tanzania
na
na
6 ÷ 9** PV, B
E-pre
VY
S
1
1
Distribution of connections type
roughly: 11% B, 70% HH, 19% HH + B.
[39]
TZ8
Tanzania
na
na
3 ÷ 6** PV, B
E-pre
VY
S
1
1
Distribution connections type roughly:
22% B, 58% HH, 20% HH + B.
[39]
TZ9
Tanzania
na
na
3 ÷ 9** PV, B
E-pre
VY
S
1
2
Distribution connections type roughly:
34% B, 54% HH, 12% HH + B.
[39]
TZ10
Tanzania
na
Aw
6
PV, B
E-pre
Y
S
1
1
Only HH.
[43]
TZ11
Tanzania
na
Bsh
9.5
PV, B
E-pre
Y
S
1
1
na
[43]
TZ12
Tanzania
na
Cwb
15.9 (TZ12 +
TZ4)
PV, B
E-pre
VY
S
1
1
na
[43]
TZ13
Tanzania
na
Aw
6
PV, B
E-pre
Y
S
1
1
na
[43]
MO1
Morocco
Elkaria
Bsh
7.2
PV, B (and not
specified Wn)
na
na
VS
2
3
Only 16 HH.
[81]
28
KE16
Kenya
Merile
Bsh
1.5 ÷ 5.6
PV, B
E-pre
na
VS
1
1
na***
[29],[79]
CV1
Cape
Verde
Monte Trigo
BWh
39 PV, 20 D, B
H
Y
S
2
3
75% of demand for HH, 17% public (e.g.
streetlights) and 9% B including ice
production for fish preservation.
[53]
Table A-4 Dataset of mini-grids in the peak cluster
Code
Country
Village/Area
Cli.
zone
Size range
[kW],
Technology
Tariff
Tier
A.
Tier
C.
Tier
P.
Tier
E.
Brief connections description
Sources
PH2
Philippines
Alaminos
Am
na
na
na
S
2
2
Mostly HH.
[37]
KE17
Kenya
Marti
As
1.5 ÷ 5.6 PV, B
E-pre
na
VS
1
1
na***
[29],[79]
KE18
Kenya
Opiroi
Cfb/Aw
1.5 ÷ 5.6 PV, B
E-pre
na
VS
1
1
na***
[29],[79]
KE19
Kenya
na
na
3** PV, B
E-pre
VY
VS
1
1
Distribution connections type
roughly: 78% B, 5% HH, 17% HH + B.
[39]
KE20
Kenya
na
na
3** PV, B
E-pre
Y
VS
1
1
Distribution connections type
roughly: 79% B, 11% HH, 10% HH + B.
[39]
KE21
Kenya
na
na
3** PV, B
E-pre
Y
VS
1
1
Distribution connections type
roughly: 100% B.
[39]
KE22
Kenya
na
na
3** PV, B
E-pre
VY
VS
0
1
Distribution connections type
roughly: 11% B, 75% HH, 14% HH + B.
[39]
TZ14
Tanzania
na
na
3 ÷ 6** PV, B
E-pre
VY
S
1
1
Distribution connections type
roughly: 65% B, 7% HH, 28% HH + B.
[39]
KE23
Kenya
Barsaloi
Bsh
1.5 ÷ 5.6 PV, B
E-pre
Y
S
1
1
na***
[29],[79]
KE24
Kenya
Namba
Koloo
Aw/Am
1.5 ÷ 5.6 PV, B
E-pre
na
VS
1
1
na***
[29],[79]
KE25
Kenya
Olenarau
Cwb
1.5 ÷ 5.6 PV, B
E-pre
VS
1
1
na***
[29],[79]
Table A-5 Dataset of mini-grids in the outliers cluster
Code
Country
Village/Area
Cli. zone
Size range [kW],
Technology
Tariff
Tier A.
Tier C.
Tier P.
Tier E.
Brief connections description
Sources
TZ15
Tanzania
na
na
7.5 (TZ5 + TZ15)
PV, B
E-pre
Y
S
1
1
na
[43]
29
MO2
Morocco
Iferd
BWh
9 ÷ 24 PV, 12 D, B
E
MY
S
2
2
Mostly HH.
[82]
MO3
Morocco
Idboukhtir
BWh
9 ÷ 24 PV, 12 D, B
E
MY
VS
2
2
Mostly HH and a water pump.
[82]
TZ16
Tanzania
na
na
na
E-pre
na
S
2
3
na
[83]
*Low-consumer: 0÷50 kWh/month; medium-consumer: 50÷1500 kWh/month; high-consumer <1,500 kWh/month [75].
** 3kWp intervals for roughly 50 costumers’ (…) and ‘the system may also include a diesel generator if there is the possibility that demand will exceed supply’
(…) [39].
*** ‘The mini-grids serve a variety of customer types, including households, businesses and social service institutions such as churches, among others’ [29].
However, no specific details for each site are available.
30
Table B-1 - Hourly load profiles of mini-grids classified as cluster flat.
Hourly profile of the mini-grid by code [kW]
Time
BO1
MX1
TZ1
OM1
OM2 w*
OM2 s*
NA1
MY1
MW1
KE1
ID1
HN1
HN2
00:00
10.64
20.00
65.18
43.00
2,980.00
5,703.50
60.29
0.0730
8.07
202.07
1,444.60
9.93
1.44
01:00
10.02
18.65
63.86
42.50
2,914.60
5,733.10
56.58
0.0775
8.89
202.07
1,444.58
9.61
1.28
02:00
9.44
17.48
63.06
41.16
2,794.80
5,601.08
56.24
0.0774
9.17
195.48
1,418.00
9.54
1.31
03:00
9.10
17.48
63.23
39.16
2,704.20
5,562.49
58.23
0.0700
10.28
186.30
1,401.00
9.38
1.31
04:00
8.70
17.30
62.75
36.96
2,673.11
5,402.41
66.31
0.0672
10.98
188.89
1,419.38
9.37
1.42
05:00
6.91
17.42
62.91
34.39
2,671.23
5,089.43
73.80
0.0671
9.92
215.41
1,537.00
9.31
1.59
06:00
7.01
17.98
72.85
26.13
2,672.82
4,353.99
77.72
0.0779
6.88
245.37
1,531.04
8.19
1.76
07:00
7.49
20.49
66.09
21.29
2,700.47
4,178.14
79.90
0.0811
5.95
248.00
1,354.80
7.65
1.64
08:00
7.67
21.47
59.59
23.28
2,622.88
4,055.02
83.84
0.0962
5.95
254.00
1,350.00
6.98
1.04
09:00
8.16
20.98
56.00
24.34
2,670.21
4,078.04
86.08
0.0541
7.16
277.44
1,375.00
7.04
1.51
10:00
8.70
21.29
60.62
25.65
2,553.36
3,997.67
84.95
0.0517
6.61
298.11
1,375.00
7.42
1.43
11:00
8.99
20.86
58.05
28.40
2,558.02
3,984.35
80.93
0.0339
6.91
310.00
1,400.00
7.86
1.52
12:00
9.57
20.55
67.69
31.38
2,559.39
4,138.58
77.67
0.0271
6.96
318.67
1,408.12
8.01
1.67
13:00
9.69
20.06
57.13
36.63
2,557.20
4,564.69
76.12
0.0186
6.97
311.56
1,404.69
8.37
1.74
14:00
10.15
20.49
55.68
40.40
2,747.30
5,574.76
74.64
0.0129
6.52
306.67
1,400.81
8.72
1.89
15:00
9.84
20.43
57.58
42.16
2,760.64
5,768.50
73.82
0.0142
7.00
303.41
1,383.04
8.20
2.04
16:00
9.63
20.86
59.23
39.05
2,798.92
5,298.24
74.18
0.0227
7.56
295.30
1,365.69
7.43
1.81
17:00
10.17
20.74
62.49
34.61
2,716.05
4,309.05
76.86
0.0329
9.77
284.19
1,346.96
7.78
2.27
18:00
15.59
24.17
60.55
26.45
3,270.06
3,709.54
85.36
0.0513
18.49
321.07
1,578.69
6.93
2.63
19:00
18.16
34.36
75.77
24.73
3,550.88
4,089.16
94.29
0.0717
22.79
411.82
2,193.69
7.90
3.24
20:00
17.81
35.09
88.09
28.62
3,363.80
4,441.40
82.08
0.0729
11.56
407.56
2,250.50
8.58
3.10
21:00
16.69
32.45
85.69
30.71
3,367.10
4,594.15
74.65
0.0753
12.59
369.00
2,085.42
10.29
2.50
22:00
13.83
27.98
72.52
33.04
3,330.81
4,858.69
66.95
0.0806
10.83
303.82
1,834.62
10.11
1.69
23:00
11.75
23.19
66.12
22.39
3,322.35
5,551.06
61.57
0.0741
10.42
251.87
1,630.77
9.90
1.56
*s: summer profile (July); w: winter profile (January); both belong to the flat cluster.
31
(continue)
Hourly profile of the mini-grid by code [kW]
Time
PH1
DO1
CO1
VC1
KE2
KE3
KE4
KE5
KE6
KE7
KE8
KE9
00:00
1,466.00
2,707.82
1,017.79
811.48
22.89
32.65
104.23
92.15
173.74
25.90
41.97
126.77
01:00
1,415.89
2,650.21
948.62
781.97
20.73
28.69
97.21
87.33
162.32
23.46
40.30
115.41
02:00
1,365.07
2,563.79
928.85
767.21
20.76
29.34
93.32
80.87
162.43
19.73
42.15
114.63
03:00
1,326.44
2,534.98
918.97
752.46
20.17
31.54
95.09
81.73
159.10
18.22
43.76
108.82
04:00
1,332.89
2,448.56
988.14
745.08
21.01
28.82
97.57
82.13
157.95
19.78
40.27
120.65
05:00
1,371.44
2,390.95
1,057.31
767.21
25.93
32.57
104.65
94.06
159.33
23.66
40.01
143.34
06:00
1,365.33
2,707.82
1,037.55
745.08
27.30
36.86
92.51
96.75
155.91
23.10
40.17
145.69
07:00
1,436.69
3,168.72
1,146.25
789.34
25.17
28.39
91.31
117.37
154.74
30.18
42.13
152.51
08:00
1,522.81
3,341.56
1,274.70
855.74
24.63
36.28
117.34
118.43
174.68
32.33
43.81
167.09
09:00
1,582.85
3,399.18
1,393.28
877.87
30.20
39.64
135.09
113.59
165.37
37.51
42.20
158.45
10:00
1,618.26
3,427.98
1,482.21
885.25
27.08
40.69
132.05
109.52
164.19
35.82
42.26
171.73
11:00
1,606.30
3,370.37
1,551.38
885.25
38.30
44.96
133.74
125.00
169.02
41.07
43.84
161.29
12:00
1,586.11
3,341.56
1,561.26
885.25
35.44
43.40
123.12
113.36
190.77
35.78
42.30
154.37
13:00
1,619.67
3,283.95
1,581.03
877.87
46.48
34.04
125.63
109.33
180.68
33.88
44.26
133.45
14:00
1,607.56
3,341.56
1,600.79
885.25
33.55
38.55
117.08
97.52
192.43
40.91
47.72
143.02
15:00
1,617.00
3,341.56
1,600.79
885.25
32.18
29.13
114.32
106.93
199.05
39.67
46.33
147.67
16:00
1,619.96
3,283.95
1,581.03
870.49
32.37
37.64
106.85
112.98
181.64
38.44
49.72
159.66
17:00
1,733.50
3,658.44
1,541.50
870.49
35.87
38.78
120.35
102.89
182.10
36.21
48.88
141.39
18:00
1,912.00
3,975.31
1,699.60
973.77
45.73
49.85
150.48
103.03
213.84
52.68
58.13
193.87
19:00
1,915.00
3,917.70
1,749.01
1,084.43
49.97
54.68
169.75
110.39
233.19
53.71
65.80
214.36
20:00
1,910.85
3,687.24
1,679.84
1,069.67
43.28
44.35
152.63
104.73
226.77
48.15
65.05
202.35
21:00
1,833.23
3,427.98
1,521.74
1,003.28
37.96
38.03
154.26
92.14
215.70
34.52
50.73
172.94
22:00
1,668.33
3,082.30
1,324.11
944.26
30.91
33.36
135.36
101.02
189.99
30.17
51.22
151.88
23:00
1,527.62
2,880.66
1,116.60
863.11
25.43
32.10
109.61
90.80
182.00
28.13
40.93
139.53
32
Table B-2 - Hourly load profiles of mini-grids classified as cluster step.
Hourly profile of the mini-grid by code [W]
Time
KE10
TZ2
TZ3
TZ4
TZ5
TZ6
UG1
KE11
KE12
KE13
00:00
4.20
[43]
6,718.67
277.20
299.71
37.90
01:00
3.20
5,235.65
250.34
300.09
20.47
02:00
3.10
4,611.33
156.51
300.25
15.57
03:00
3.70
4,248.13
75.93
313.86
10.67
04:00
7.80
4,043.80
63.33
349.63
5.78
05:00
25.70
3,908.47
103.12
318.15
5.31
06:00
45.20
4,240.23
417.79
329.93
9.51
07:00
80.80
4,337.65
599.83
394.44
16.25
08:00
128.00
4,954.48
682.06
479.68
21.29
09:00
169.00
6,788.09
924.11
534.76
28.06
10:00
194.00
8,479.78
1,362.13
635.09
41.17
11:00
198.00
11,776.27
1,487.79
563.27
64.17
12:00
196.00
14,669.83
1,446.68
591.79
87.72
13:00
190.00
14,337.33
1,562.40
707.05
108.57
14:00
184.00
15,304.92
1,240.11
708.24
121.45
15:00
171.00
14,834.10
1,296.81
727.37
125.35
16:00
153.00
14,505.65
1,206.28
805.85
124.72
17:00
165.00
11,480.22
2,425.50
1,440.57
116.54
18:00
235.00
9,535.74
2,377.42
1,208.43
109.31
19:00
235.00
11,282.62
1,954.66
855.62
121.65
20:00
167.00
13,905.00
1,402.58
630.47
129.89
21:00
75.10
13,998.91
945.33
449.66
117.33
22:00
23.30
12,758.69
593.53
342.91
93.39
23:00
9.00
9,178.04
395.57
315.06
64.54
33
Table B-3 - Hourly load profiles of mini-grids classified as cluster step-peak
Hourly profile of the mini-grid by code [W]
Time
KE14
KE15
TZ7
TZ8
TZ9
TZ10
TZ11
TZ12
TZ13
MO1
KE16
CV1
00:00
169.00
17.50
156.00
58.70
609.00
[43]
669.00
33.08
2,691.94
01:00
138.00
15.60
135.00
53.80
483.00
545.25
23.05
2,009.48
02:00
120.00
14.40
124.00
50.80
436.00
599.73
19.92
2,009.48
03:00
109.00
14.20
122.00
50.50
399.00
558.40
16.79
2,009.48
04:00
134.00
15.60
127.00
57.50
376.00
456.43
13.66
2,009.48
05:00
192.00
18.60
140.00
68.10
423.00
697.14
13.51
2,274.88
06:00
184.00
22.30
166.00
83.80
809.00
795.47
15.87
2,881.52
07:00
225.00
31.90
209.00
102.00
1,210.00
1,134.50
24.77
3,222.75
08:00
297.00
43.40
240.00
119.00
1,430.00
908.29
39.20
3,184.83
09:00
365.00
52.20
273.00
185.00
1,670.00
952.47
54.59
2,350.71
10:00
398.00
66.10
315.00
287.00
1,790.00
886.21
63.94
2,388.63
11:00
399.00
73.80
330.00
310.00
1,830.00
1,040.10
66.69
2,199.05
12:00
376.00
74.10
332.00
298.00
1,800.00
954.40
68.46
3,905.21
13:00
351.00
64.60
335.00
235.00
1,800.00
1,049.00
69.57
4,018.96
14:00
346.00
55.60
343.00
189.00
1,910.00
1,144.73
67.91
3,639.81
15:00
330.00
46.10
361.00
171.00
2,090.00
1,156.35
68.39
3,677.73
16:00
343.00
39.60
357.00
168.00
2,080.00
1,027.50
75.56
2,464.45
17:00
578.00
46.50
365.00
154.00
2,020.00
1,056.00
85.61
3,222.75
18:00
973.00
103.00
623.00
288.00
3,310.00
900.50
98.66
5,952.61
19:00
988.00
130.00
797.00
372.00
3,890.00
2,951.75
121.40
7,431.28
20:00
812.00
107.00
786.00
352.00
3,780.00
2,994.08
123.95
7,545.02
21:00
563.00
65.00
584.00
231.00
3,040.00
2,051.00
100.26
6,900.47
22:00
343.00
35.70
344.00
135.00
1,890.00
1,865.20
70.08
5,497.63
23:00
222.00
22.80
209.00
77.40
992.00
997.94
47.49
3,563.98
34
Table B-4 - Hourly load profiles of mini-grids classified as cluster peak
Hourly profile of the mini-grid by code [W]
Time
PH2
KE17
KE18
KE19
KE20
KE21
KE22
TZ14
KE23
KE24
KE25
00:00
416.35
50.40
41.61
18.60
95.50
29.20
21.40
12.50
36.31
11.73
41.61
01:00
202.24
30.37
25.56
10.60
54.50
18.70
14.90
10.90
15.89
9.94
25.56
02:00
192.28
24.43
20.29
7.35
37.80
15.00
11.30
11.70
9.33
9.09
20.29
03:00
202.24
18.49
15.02
6.48
30.10
16.20
8.70
10.90
9.83
8.24
15.02
04:00
249.04
12.56
9.74
11.80
34.20
24.20
7.90
14.10
25.97
7.39
9.74
05:00
513.58
11.92
11.37
19.40
57.10
36.40
9.80
26.40
86.24
8.45
11.37
06:00
1,143.21
16.38
15.45
20.80
57.60
39.30
9.30
26.20
85.49
14.53
15.45
07:00
614.10
25.85
28.37
28.50
44.80
39.60
8.70
18.30
153.83
13.31
28.37
08:00
511.84
35.62
41.08
33.30
41.00
44.30
8.20
33.10
185.60
10.46
41.08
09:00
560.35
38.29
52.35
34.70
46.00
48.40
7.90
23.60
195.43
10.38
52.35
10:00
451.62
44.98
49.36
36.80
53.10
49.10
8.10
25.40
204.26
10.52
49.36
11:00
505.54
43.46
51.28
37.20
57.80
54.70
8.30
24.00
188.12
11.31
51.28
12:00
774.68
41.84
59.20
39.00
66.20
54.80
8.70
25.60
179.55
11.81
59.20
13:00
388.91
43.08
68.07
46.50
76.10
65.30
9.60
30.50
125.58
12.67
68.07
14:00
388.82
44.67
69.01
48.00
86.20
68.20
11.10
31.10
120.29
13.17
69.01
15:00
517.68
48.19
59.29
46.20
98.60
77.10
13.60
41.00
124.57
14.80
59.29
16:00
530.22
48.22
51.96
50.70
124.00
82.40
19.20
39.00
108.94
19.90
51.96
17:00
616.62
55.31
46.14
98.00
219.00
102.00
40.20
34.40
279.16
27.11
46.14
18:00
978.74
75.81
52.36
189.00
487.00
215.00
73.20
148.00
894.21
49.99
52.36
19:00
1,294.76
153.72
185.28
189.00
604.00
306.00
80.70
223.00
1,058.63
79.31
185.28
20:00
3,602.41
261.78
325.43
156.00
576.00
280.00
74.50
173.00
856.63
70.79
325.43
21:00
2,449.84
274.27
267.80
108.00
455.00
208.00
58.60
93.50
479.89
54.38
267.80
22:00
1,695.06
197.68
148.06
62.80
302.00
121.00
41.90
38.20
246.88
33.32
148.06
23:00
1,178.61
102.54
77.33
33.30
176.00
57.80
29.60
19.90
106.92
17.04
77.33
35
Table B-5 - Hourly load profiles of mini-grids classified as cluster step-peak
Hourly profile of the mini-grid by code [W]
Time
TZ15
MO2
MO3
TZ16
00:00
[43]
2,320.00
880.00
5,500.00
01:00
1,215.68
815.23
5,059.12
02:00
749.72
786.56
4,880.34
03:00
606.34
773.81
4,879.25
04:00
577.90
769.80
4,909.29
05:00
759.94
770.41
5,051.39
06:00
1,318.43
729.39
5,326.12
07:00
1,146.61
618.66
5,462.83
08:00
772.34
547.14
4,319.39
09:00
665.99
424.68
4,435.29
10:00
706.49
399.14
4,911.48
11:00
859.69
444.89
5,464.94
12:00
1,104.38
582.57
5,932.91
13:00
1,246.42
843.14
5,820.08
14:00
1,444.06
1,025.49
5,318.20
15:00
2,000.33
1,045.36
4,695.23
16:00
1,976.17
868.78
5,576.54
17:00
1,410.36
684.67
6,537.48
18:00
994.70
582.61
6,996.00
19:00
1,036.38
765.00
6,906.98
20:00
1,738.60
1,808.17
9,366.04
21:00
4,539.97
2,191.89
12,285.26
22:00
5,536.93
1,954.32
12,430.49
23:00
4,800.00
1,280.00
8,758.33
36
Table C-1 - Demand profiles corresponding to those shown in Figure 1: average value and standard deviation.
Mean and standard deviation of each cluster average profile [p.u.]
Time
flat
step
step-peak
Peak
outliers
mean
std
mean
std
mean
std
mean
std
mean
std
00:00
0.934
0.161
0.461
0.287
0.461
0.167
0.494
0.208
1.153
0.241
01:00
0.900
0.184
0.356
0.229
0.366
0.158
0.315
0.148
0.826
0.061
02:00
0.881
0.186
0.295
0.191
0.346
0.158
0.257
0.121
0.633
0.220
03:00
0.871
0.177
0.256
0.187
0.332
0.147
0.219
0.095
0.597
0.247
04:00
0.875
0.159
0.263
0.199
0.356
0.120
0.228
0.083
0.587
0.260
05:00
0.904
0.126
0.302
0.174
0.459
0.148
0.357
0.153
0.616
0.239
06:00
0.899
0.145
0.420
0.148
0.531
0.132
0.469
0.331
0.733
0.204
07:00
0.905
0.160
0.551
0.140
0.640
0.167
0.465
0.142
0.717
0.118
08:00
0.941
0.209
0.708
0.216
0.756
0.101
0.535
0.148
0.601
0.101
09:00
0.956
0.121
0.915
0.291
0.907
0.133
0.565
0.164
0.573
0.164
10:00
0.959
0.126
1.152
0.335
1.091
0.271
0.570
0.153
0.615
0.212
11:00
0.987
0.164
1.320
0.314
1.151
0.317
0.584
0.140
0.683
0.220
12:00
0.980
0.149
1.462
0.290
1.171
0.271
0.642
0.176
0.794
0.183
13:00
0.974
0.193
1.537
0.278
1.131
0.186
0.648
0.188
0.892
0.100
14:00
0.987
0.198
1.541
0.266
1.104
0.139
0.670
0.184
0.962
0.132
15:00
0.985
0.200
1.550
0.212
1.118
0.157
0.715
0.132
1.042
0.223
16:00
0.985
0.158
1.481
0.200
1.086
0.230
0.758
0.146
1.010
0.147
17:00
1.006
0.128
1.626
0.503
1.177
0.248
1.077
0.406
0.917
0.124
18:00
1.194
0.264
1.584
0.484
1.707
0.453
2.246
1.130
0.828
0.268
19:00
1.364
0.328
1.663
0.300
2.391
0.221
3.226
0.953
0.909
0.247
20:00
1.280
0.209
1.556
0.345
2.284
0.233
3.634
0.550
1.527
0.394
21:00
1.190
0.166
1.362
0.440
1.689
0.263
2.737
0.704
2.383
0.352
22:00
1.075
0.127
0.984
0.474
1.095
0.302
1.665
0.565
2.519
0.655
23:00
0.968
0.140
0.654
0.375
0.653
0.193
0.925
0.365
1.887
0.679
37
38
... Load profiles are important when modeling, simulating, and optimizing energy systems because they offer insights into energy consumption patterns [124][125][126]. These profiles inform decisions on system sizing, resource allocation, and operational strategies by detailing variations in energy demand over time. ...
... The importance of load profiles cannot be over-emphasized. However, the development of load profiles, particularly in the global south, faces significant challenges due to limited data collection and sharing, high survey costs, and socioeconomic barriers [124][125][126]. To navigate these challenges, studies typically include small sample sizes, focus on specific components, and rely on average profiles, as summarized in Table 3. Small sample sizes limit the generalizability of the findings to the broader population, while profiles focusing on limited components may not capture the full range of energy consumption behaviours and needs. ...
... The energy consumption is then broken down by explicitly defining the number of people and households in a community, and the components that constitute the off-grid loads such as household devices, health clinic, school, retail shop, and industrial/agricultural energy consumption. These load profiles can then be scaled using the archetypical profiles developed by Lorenzoni et al. [125], as illustrated in Fig. 9 [see supplementary materials 1 and 2 (excel-based load calculation tool)]. ...
... Lorenzoni et al. [43] developed a database of load profiles from sixty-one mini-grid projects in DCs. A clustering identified archetypal profiles, followed by graphical analysis to explore factors influencing load profile shapes. ...
... Household wealth proxies such as "monthly expenditure", "dwelling quality index", and "motorized vehicle ownership" positively impact appliance presence, aligning with findings from Rao and Ummel, Dominguez et al., and Richmond et al. [26,28,40]. [28,[42][43][44]. Higher "education of household head" also correlates with greater appliance ownership, possibly reflecting better electricity awareness or higher income, as suggested by Rao and Ummel [26], though contrary to Kurata et al. [27]. ...
Article
Full-text available
To grant reliable and affordable electricity provision to non-electrified communities, proper system sizing, based on accurate demand estimation is crucial. However, the absence of historical data, and scarce, scattered, and often unreliable pre-electrification surveys, make this process particularly prone to errors. Acquiring data, especially with high quality and detail, is difficult, time-consuming and expensive. Even though, in a few site-specific cases the limited data collected has allowed researchers to develop methodologies to generate synthetic demand profiles based on variegated site-specific socioeconomic information and appliance adoption patterns. However, given the lack of comprehensive datasets of such information, the use of synthetic meth-odologies has been circumscribed to limited regional and socioeconomic scopes. This research proposes the development of a data-driven machine-learning framework for estimating appliance adoption patterns with a subset of relevant socioeconomic indicators, identified throughout a comprehensive literature analysis and data collection across various sources. To successfully train the model, a novel open-access database has been created and populated with socioeconomic information combined with appliance data collected from public and private sources. Finally, a structured logistic regression analysis has been performed, not only to capture the nexus of socioeconomic factors with appliance adoption but also to estimate the most relevant ones. The methodology calibrated with the proposed open-access database has shown 71.7 % accuracy, which represents an important achievement in the field. The study's findings lay the foundations for simplifying the estimation of appliance adoption, which can facilitate the demand estimation for sizing rural energy systems and rural electrification approaches.
... Finally, powering new consumers requires a design phase of the grid, especially in remote areas, where the national grid is absent, making building mini-grids necessary. Therefore, accurate information on the users' patterns and behaviours is of great importance to estimate, for example, the initial load, which is a critical task [32]. In fact, overestimating the load profile will eventually jeopardise the profitability of the project due to extra costs, while underestimating the demand will lead to issues regarding the reliability of provided services and leaves the customers dissatisfied [32][33][34][35]. ...
... Therefore, accurate information on the users' patterns and behaviours is of great importance to estimate, for example, the initial load, which is a critical task [32]. In fact, overestimating the load profile will eventually jeopardise the profitability of the project due to extra costs, while underestimating the demand will lead to issues regarding the reliability of provided services and leaves the customers dissatisfied [32][33][34][35]. ...
Article
Full-text available
In the West African Monetary and Economic Union (UEMOA), information on the characteristics of the users and patterns of electricity end-uses remains hard to find. This study aims to contribute to reducing the gap in research on domestic electricity consumption in the region by unveiling the ownership rates, patterns of use and electricity consumption of domestic appliances in urban households through a city-wide survey. Three categories of urban users were investigated including high, medium and low consumers. Findings demonstrated various ownership rates for appliances, ranging from 100% for lighting fixtures to 0% for washing machines depending on user category. Domestic electricity demonstrated patterns consisting of three peak demand periods, with the main ones occurring in the evening (19:00 to 20:00) and the night (22:00). Other demand characteristics include an average daily electricity use ranging from 0.50 to 6.42 kWh per household, a maximum power demand of between 0.19 and 0.70 kW and a daily load factor between 35 and 58%. Finally, the appliances contributing the most to domestic electricity use include air-conditioners, fans, fridges and freezers, televisions and lighting fixtures, with contributions differing from one category of user to another. Policy implications including review of the appliances’ importations framework and policies, and incentives for purchasing efficient appliances, design of more tailored policies, considering the different backgrounds of the users, education enhancement on energy behaviours for increasing energy efficiency/conservation, and implementation of DSM programs including load levelling, load shifting and load reducing depending on the type of appliance for energy conservation in the domestic buildings were derived. Overall, a large range of stakeholders of the electricity sector, not only in the West African Economic and Monetary Union (UEMOA), but also in other regions and countries sharing common characteristics should be interested in the results of this study.
... This can be achieved through long-or short-term measurements or predictions (Serrano-Guerrero et al., 2018). Previous studies present models for energy profiles prediction based on consumer parameters, or employing surveys, regression analysis, decision trees, and ANN (Abarkan et al., 2013;Blodgett et al., 2017;Lorenzoni et al., 2020;Serrano-Guerrero et al., 2018;Tso & Yau, 2007). ANN has been successful in forecasting household electric energy consumption and load profiles (Rodrigues et al., 2014). ...
Article
Full-text available
This paper presents an approach for sizing a hybrid photovoltaic system for a small-scale peanut oil processing company (Yaye Aissatou, Passy) in rural Senegal using a synthetic load profile. In this study, a predictive model of the electrical load of a service-based plant oil processing company was developed through a diagnosis, to evaluate the extraction process. The mass and energy balance were measured, and the process was implemented into MATLAB Simulink. The simulated load profile was implemented in HOMER Pro and the characteristics of the most profitable hybrid systems were identified. The results showed that the lowest net present cost over 25 years was found with a PV/battery/grid-system with 18.6 kW p solar panels, 16 kWh of storage, and an initial investment of 20,019 €. Compared to a grid-only scenario, this solution reduces the net present cost from an initial 72,163 € to 31,603 €, the operating cost from 3675 € per year to 590 € per year, and the cost of energy from 0.29 to 0.13 €/kWh. The renewable fraction of the proposed system is 90.0 % while the expected payback period is 6.2 years. The study demonstrates the economic feasibility of using solar energy for plant oil processing.
... In many cases, data on electricity demand are not available or not reliable since they are rarely collected, and not reported systematically [4]. The lack of databases with high-resolution data on energy usage from rural areas results in limited historical data to draw from [5]. The most common method for estimating electricity usage currently adopted is through appliance ownership and use data collected via interviews [6]. ...
Article
Energy poverty is a significant barrier to development for millions of people globally in remote areas; developing nations like India still use conventional fuels to meet their energy needs. Microgrids can be a feasible solution for remote electrification by integrating distributed energy resources. The present work investigates the feasibility, planning, and optimal sizing of a standalone microgrid system from a socio-techno-economic and environmental perspective for the electrification of a remote area in an Indian scenario. For the feasibility analysis, a remote village, Dayarthi, in Andhra Pradesh, India, was investigated by considering the daily load profile, which includes domestic loads, community loads, and agriculture loads. The total load demand is 333.53 kWh/day, with a peak load of 45.75 kW. Four potential microgrid configurations are investigated, with various combinations of diesel generator, wind turbine, photovoltaic, and battery storage. A sociotechnical-economic-environmental analysis identifies the best configuration by looking at the different microgrid scenarios that are possible and suggesting the best one with the highest percentage of renewable energy at the lowest net present cost and levelized cost of energy with minimum unmet loads. Furthermore, the optimal scenario cost of energy is compared with the most recent study in the literature.
Conference Paper
In recent years, there has been a growing interest in using renewable energy sources to electrify rural areas in Nigeria. The major motivation for this study is the recognition of the huge potential of renewable energy to improve rural electricity access in the country. One promising option is to use off-grid photovoltaic (PV) mini grid systems. These systems are small, modular, and can be easily installed in remote areas. However, some studies have shown that these systems are posed with certain problems such as its inability to carry loads for prolonged periods of time when the sun is out. This work investigates the feasibility of an optimized off-grid PV mini grid system for electrification of rural areas in Nigeria. The paper made use of the HOMER pro, SAM and MATLAB software. The HOMER pro software was to design a proposed mini grid for a case community in the northern part of Nigeria. The results of the HOMER pro simulations showed that the system generated high NPC and low LCOE. The proposed system for the northern generated an NPC of N= 380,498,512 and an LCOE of 37.01 N= /kWh. The results of the SAM software showed that the PV array of the proposed models was able to generate sufficient electricity for the community. The results of the MATLAB software showed that a short-term load forecasting carried out using the ANN Levenburg-Marquardt algorithm gave results more accurately for the short-term load demand of the community. This exchange rate at the time of this work was N= 769.5 to $1.
Article
Full-text available
Microgrids (MGs) have emerged as a promising solution for providing reliable and sustainable electricity, particularly in underserved communities and remote areas. Integrating diverse renewable energy sources into the grid has further emphasized the need for effective management and sophisticated control strategies. This review explores the crucial role of control strategies in optimizing MG operations and ensuring efficient utilization of distributed energy resources, storage systems, networks, and loads. To maximize energy source utilization and overall system performance, various control strategies are implemented, including demand response, energy storage management, data management, and generation‐load management. Employing artificial intelligence (AI) and optimization techniques further enhances these strategies, leading to improved energy management and performance in MGs. The review delves into the control strategies and their architectures, and highlights the significant contributions of AI and emerging technologies in advancing MG energy management.