Electricity For All The Contribution of Large-Scale Planning Tools To The

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Renewable and Sustainable Energy Reviews 120 (2020) 109624

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews


journal homepage: http://www.elsevier.com/locate/rser

Electricity for all: The contribution of large-scale planning tools to the


energy-access problem
Pedro Ciller *, Sara Lumbreras
Institute for Research in Technology (IIT), ICAI School of Engineering, Comillas Pontifical University, Santa Cruz de Marcenado 26, 28015, Madrid, Spain

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

Keywords: There is a huge need for investment for the electrification of rural areas in developing countries. Any effective
Electricity access approach to this enormous task requires strategic planning that combines diverse electrification modes: grid
Rural electrification extension, mini-grids, and stand-alone systems. Advanced computer tools are necessary to support planners. This
Electrification methodology
paper reviews the existing techniques, software tools, and approaches that can contribute to this job. We propose
Energy technology
Planning model
a comprehensive but compact mathematical formulation of rural electrification planning as an optimization
Software tool problem. This general formulation establishes a common ground for a critical review of the different tools and
solution methods and allows the identification of the primary research needs in this field.

Decision-makers must be aware of different factors that define the


1. Motivation: the challenge of access to electricity problem, such as cost, long-term sustainability, regulatory issues, local
constraints, or specific preferences to determine the best course of action
About 840 million people worldwide lack access to electricity, [4]. These factors strongly relate to each local context and, therefore,
mainly in rural areas in Sub-Saharan Africa [1], and it is systematically each region requires a unique tailor-made solution. In principle, a rural
in the rural areas where the deficit of electrification is the most preva­ electrification plan must search for the best compromise between
lent. There is not a globally accepted definition of access to energy, minimizing the cost of supply and the quality of the service provided.
which may be translated into an erroneous perception of the problem There is a growing consensus on the fact that a plan should be a com­
since a village with 10% of electrified households may be considered to bination of off-grid and on-grid systems [5]. Social preferences and other
be electrified [2]. Indeed, electricity is often used to meet the basic needs human factors should also be considered [6,7]. The value that con­
of households and other facilities (such as schools and health centers), sumers give to the reliability level depends on many factors, as well as
often neglecting the remaining needs. The Sustainable Energy for All their preference for a particular electrification mode or a specific gen­
(SE4All) project introduced a multi-tier framework with several levels of eration technology despite a higher cost. Moreover, the existing regu­
electricity access (or tiers) to consider the different levels of electrifica­ latory framework can profoundly impact the selection of an
tion. This multi-tiered framework [3] has been officially adopted by the electrification plan. Some policies can promote specific renewable en­
World Bank (WB) to get a more accurate representation of electrification ergies such as solar or wind or encourage the reduction of fossil fuels.
in its projects. Besides, the beneficiaries of the electrification project can and should be
Establishing the planning agenda of a large-scale rural electrification involved in the process. Although policymakers and utilities play the
project is a complex issue that requires a multidisciplinary approach. most prominent role in electrification planning, other players such as

Abbreviations: AC, Alternating Current; CIEMAT, Centro de Investigaciones Energ� eticas, Medioambientales y Tecnol�ogicas; DC, Direct Current; DER-CAM,
Distributed Energy Resources - Customer Adoption Model; DL, Distribution Layout; DP, Development Pole; EF, Electric Feasibility; EFOM, Energy Flow Optimization
Model; GIS, Geographic Information Systems; HOMER, Hybrid Optimization Model for Multiple Energy Resources; HV, High Voltage; IED, Innovation Energie
D�
eveloppement; IIT, Instituto de Investigaci�on Tecnol�
ogica; KTH, Royal Institute of Technology; LAPER, Logiciel d’Aide �
a la Planification d’Electrification
� Rurale;
LCOE, Levelized Cost of Electricity; LV, Low Voltage; MARKAL, MARKet Allocation; MIT, Massachusetts Institute of Technology; MILP, Mixed Integer Linear Pro­
gramming; MINLP, Mixed Integer Non-Linear Programming; MV, Medium-Voltage; NGO, Non-Governmental Organization; NL, Network Layout; NSE, Non-Served
Energy; OSeMOSYS, Open Source Energy Modelling System; OnSSET, Open Source Spatial Electrification Tool; REM, Reference Electrification Model; RLI, Renier
Lemoine Institut; RNM, Reference Network Model; SDG 7, Seventh Sustainable Development Goal; SE4All, Sustainable Energy for All; T, Topography; TE, Trans­
mission Expansion; TIMES, The Integrated MARKAL-EFOM System; UR, Upstream Reinforcements; WB, World Bank.
* Corresponding author.
E-mail address: pciller@comillas.edu (P. Ciller).

https://doi.org/10.1016/j.rser.2019.109624
Received 10 June 2019; Received in revised form 21 November 2019; Accepted 24 November 2019
Available online 19 December 2019
1364-0321/© 2019 Elsevier Ltd. All rights reserved.
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

Nomenclature: soc maximum allowed state of charge of the battery [p.u.]


soc minimum allowed state of charge of the battery [p.u.]
SETS profileiht maximum generation profile in node i at hour h for
N supply and demand nodes generation technology t [kWh]
I2N demand nodes (households, schools, health centers, etc.) profile iht minimum generation profile in node i at hour h for
S2N supply nodes (already existing network candidate generation technology t [kWh]
connection points) plimijvc capacity limit of the c-th catalogue line of voltage v
Y off-grid elements (generation technologies and batteries) between nodes i; j [kW]
and grid extension components (MV/LV substations) reactanceij reactance of the candidate line between nodes i; j [Ω]
G2Y off-grid elements (generation technologies and batteries)
sbase base power [kW]
and on grid generation
xijvc reactance of the c-th catalogue line of voltage v between
Z2Y other elements related to the grid-extension designs
nodes i; j [p.u.]
(namely MV/LV substations)
O2G components of the off-grid system that are not generation VARIABLES
technologies (namely batteries) CAPiy positive variable that accounts for the capacity of the y-th
T2G generation elements, both off-grid (namely diesel, solar element installed at node i [kW]
and wind) and on-grid (where the category used will be CDECiyc positive variable used to express the capacity cost CAPiy
grid extension) with a piecewise linear function
C available elements of the catalogue for each generation BINCiyc binary variable that takes the value 1 if CDECiyc (and
technology (including lines for grid extension) therefore CAPiy ) is between capyðc 1Þ and capyc
H hours for the period considered (usually one year: 8760 h) Вiz binary variable that takes the value 1 if an only if there is a
V Voltage levels (namely low voltage and medium voltage) z element (namely an MV/LV substation) located in node i
PARAMETERS CHARGEih positive variable that accounts for how much the battery
annualfactory fraction of the cost of the y-th element that is is charged in node i at hour h
amortized in one fiscal year. It depends on its lifetime, the DISCHARGEih positive variable that accounts for how much the
discount rate considered and its total cost battery is discharged in node i at hour h
capyc capacity of the c-th element of the catalogue associated CINViy positive variable that accounts for the investment cost
with the element y [kW] associated with element ylocated at node i
cinvyc investment cost required to acquire an element y of TOTCINVy positive variable related to the total investment cost
capacity c [$] associated with element y
cfuel market price for diesel [$/liter] CFUELi positive variable related to the fuel cost associated to the
clijvc investment cost of the c-th catalogue line of voltage v diesel generator located at node i
Ξihðt¼gensetÞ binary variable that describes the commitment of a diesel
between nodes i and j [$]
cO&My annual operation and maintenance cost of element y generator (it takes the value 1 if the generator at node i is
expressed as a fraction of investment cost [p.u.] on at hour h, and 0 otherwise)
demandih demand of node i at hour h [kWh] TECHGENihg positive variable related to the real power generated at
emissions diesel emissions per liter [emissions/liter] node i by technology g at hour h [kWh]
εdiesel efficiency of diesel generators [liters/kW] GENih positive variable related to the real power generated at
εnetwork network efficiency [p.u.] node i in hour h, either with a microgrid generator or with
εcharge efficiency of battery charge [p.u.] a grid extension of the distribution network [kWh]
εdischarge efficiency of the battery discharge [p.u.] Pijhvc free variable related to the directed power flow that goes
M big-M parameter (sufficiently large number used in through node i to node j at hour h through the c-th
disjunctive constraints) catalogue line of voltage v [kW]
capacityy maximum capacity of generation, batteries, and Pþ
ijhvc positive variable that takes the value Pi;j;h;v;c if Pi;j;h;v;c is
substations installed [kW] positive and takes the value 0 otherwise
capacity y minimum capacity of generation, batteries, and SOCih positive variable related to the state of charge of the
battery located at node i in hour h [p.u.]
substations installed [kW]
Θih phase angle at node i [radians]
chargei maximum hourly charge for the battery located in node i
XLijvc binary variable that takes the value 1 if nodes i; j are
[kWh]
connected with the c-th catalogue line of voltage v
charge i minimum hourly charge for the battery located in node i
[kWh]

local entrepreneurs and non-governmental organizations (NGOs) may


intervene as well. The adopted business model varies with the structure
of the power sector or the level of participation of private investors in
the typically publicly-owned incumbent distribution utility [8].
Fig. 1 shows the different steps of the rural electrification process.
Although they are represented as independent steps in a sequential
process, in practice, these are usually implemented in a series of itera­
tions during the timeframe of the process, which may take several years. Fig. 1. Rural electrification process. Source: adapted from Ref. [9].
For example, the definition of the problem may change in the middle of

2
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

the process to accommodate to a new political scenario and, although Several reviews deal with specific aspects of rural electrification. As a
this may not affect the second step, it may be necessary to incorporate particular issue that can have important uses in rural electrification, the
new political constraints into the techno-economic solutions provided in use of Geographic Information Systems (GIS) in renewable energy pro­
the third step. jects has been studied [15]. An overview of different types of energy
The first step consists in defining the problem. The main goals of the models is provided in Ref. [16], and a meta-analysis of the most relevant
project are established, defining the region to be analyzed (which may studies that compare both grid-connected and off-grid systems has been
be a large-scale area or even an entire country), the electrification target, published [17].
the available budget and the timeframe of the process. The regulatory A part of the literature addresses the rural electrification planning
framework and political context shape the problem as well. Some problem at a small scale (i.e., focusing on an individual system that
countries may subsidize the use of green energy or penalize emissions; if usually represents a village or a community). Ref. [18] presents the main
this is the case, such constraints should be incorporated into the process. techniques used to find the best generation design for a hybrid system.
The second step is related to collecting the relevant data such as the These techniques are the basis of many generation sizing software tools
location of the consumers and their demand profiles, the location of the [19] that have been applied in numerous projects [20]. Ref. [21] clas­
existing power grid, renewable source availability such as solar irradi­ sifies the main methodologies for off-grid electrification projects.
ance and wind speed, the available components for generation and However, there are still no publications that consider the large-scale
network and the topography of the region. The available components (i.e., aiming at an entire country or region) techno-economic rural
limit the feasible techno-economic solutions in step 3. Data gathering electrification planning problem as a whole, defining it in its most
turns out to be one of the most time-consuming tasks in the whole general scope and linking its different dimensions to the tools that have
process since accurate data are usually scarce and hard to find. Some­ been proposed to tackle it in the academic literature and as software
times it is necessary to cross information from several databases and tools. Ref. [22] tries to establish a framework to analyze energy planning
process it to obtain an estimation of the relevant inputs. The authors tools in the context of developing countries, but it does not perform a
think that the data gathering process has not received enough attention sound technical analysis, and the number of tools and methods it con­
in the literature, being section 3.3 from Ref. [10] an exception. In order siders is reduced. Ref. [23] analyzes several general energy planning
to deal with this issue, some tools exploit Geographic Information Sys­ models and their application to developing countries, but the analysis is
tems (GIS) capabilities to provide instant access to updated geospatial performed from a broader perspective and does not include the specific
information and combine it with, for example, the multi-tier approach tools mentioned in this paper. This paper provides a comprehensive
proposed by the WB for demand estimations [11]. overview of the problem and the tools available to solve it at the
The third step calls for the application of rural electrification plan­ large-scale level, and it points to the most critical areas where additional
ning tools. They assist in finding the best electrification planning solu­ research is needed.
tion for the problem defined in the first step with the data gathered in The contributions of this paper are the following:
the second step. To achieve this, it is necessary to determine which
consumers are better electrified with grid extensions designs and who � It comprehensively defines the techno-economic rural electrification
would benefit more from off-grid systems. problem, including a general formulation that provides a common
Next, the solution is implemented. It is essential to consider the ground to compare tools and methodologies. This formulation was
preferences of the consumers since, for example, some villages could missing in the literature.
prefer to be electrified with mini-grids, although the best techno- � It includes a critical review of academic studies and software tools
economic solution involves grid extensions to electrify them. In prac­ related to the large-scale techno-economic rural electrification
tice, it may be necessary to conduct field interviews for the final planning problem, incorporating considerations about both
implementation. modeling and solution methods.
Universal access to energy is recognized as a key development goal, � As a result of this analysis, this paper provides a list of what the
and as such has been included in the seventh Sustainable Development authors consider the main challenges that should be addressed by the
Goal (SDG 7) established by the United Nations. The SDG 7 urges for rural-electrification community.
universal access by 2030, which implies formidable short and medium-
term investment (about 55 billion dollars per year [12]). Without a The rest of this paper is organized as follows. First, Section 2 gives an
substantial effort, 650 million people will not have access to electricity overview of the techno-economic rural electrification planning problem.
by 2030 [1]. Moreover, the number of people without access to clean Section 3 defines the variables involved and proposes a formulation to
cooking facilities is around 3 billion, and it has not improved substan­ encourage a common basis for discussion. Then, section 4 presents the
tially in recent years. This poses considerable health risks, as the use of most important works and classifies them according to their modeling
polluting technologies to cook causes around 4 million deaths per year complexity and computation speed. Both academic studies and com­
[1]. Air pollution is the most important environmental cause of death in mercial tools are included in this classification. When presenting the
developing countries, identified as the cause of most chronic lung dis­ classification, the main advantages and caveats of each modeling or
eases and respiratory tract cancers [13]. The rural electrification prob­ solution choice are discussed. In addition, a concise summary of this
lem has grown to be one of the most impending challenges in power analysis is presented in the form of summary tables for easy inspection.
systems planning, even expected to grow in importance given the elec­ Section 5 extracts conclusions and presents a list of what, according to
trification targets that should be met in the next few decades. the authors, should be the main future lines of research to be able to rise
The magnitude of the problem justifies the use of tools that can to the challenges that rural electrification will present in the next
support electricity planning initiatives to make the necessary investment decades.
as efficient as possible. Although there are political and social factors
that may condition these plans, the process can be improved with the use 2. The rural electrification problem
of rigorous data and tools [14]. Since incorporating all the relevant as­
pects into a single model is unfeasible, the existing software tools tend to From a strictly techno-economic point of view, the rural electrifica­
focus on the techno-economic part of the problem, trying to find some tion problem consists of determining the combination of stand-alone
least-cost solution. These models undoubtedly facilitate the planning systems, microgrids and grid-extensions, and their corresponding spe­
process, but they do not substitute human intervention, which must be cific designs, which optimizes the trade-off between cost and reliability
able to incorporate into the final solution all the additional factors of supply to some pre-specified ensemble of consumers, subject to any
mentioned above. constraints that may exist. Planning for an isolated region would

3
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

constitute a particular case of this problem since, in this case, grid- are few references related to clustering applications to the rural elec­
extension is not a feasible electrification solution. Even when limiting trification planning problem in the literature such as [24], although this
the scope of the problem to the techno-economic aspect, the number of process has manual steps and cannot be adapted easily to large-scale
decision variables that are needed to characterize an electrification plan cases. Another example is [25], but this reference does not consider
is very high. The most important ones are: the off-grid technologies and their costs in the process. Most tools and
methodologies require the user to define the clusters beforehand and
� The chosen electrification mode for each consumer (grid extension, introduce them as inputs (villages, settlements or cells), although a few
mini-grid or stand-alone) and the set of consumers it belongs to (grid tools such as GEOSIM, the Open Source Spatial Electrification Tool
extension feeder or mini-grid). (OnSSET) and the Reference Electrification Model (REM) include a
� The generation design for each off-grid system (generation and clustering algorithm. Fig. 3 shows a clustering example obtained with
storage technologies, capacities and locations) and its dispatch the REM model. The grid-extension clusters are obtained by grouping
strategy (i.e., which generation technologies are used to satisfy de­ the consumers for candidate grid-extension designs. It is important to
mand at each specific hour). highlight that the final electrification mode of a grid-extension cluster
� The network design of each microgrid and grid extension: voltage, does not need to be a grid extension (REM decides that later with a cost
type, and layout of distribution lines, as well as the location and the comparison between off-grid and grid extension solutions). Consumers
capacity of any new substations. that belong to the same cluster are represented as points that have the
� Upstream reinforcements or additional generation resources that same color in Fig. 3. The location of consumers often follows roads or
might be needed upstream. rivers, which is reflected in their spatial representation.
The off-grid generation design or generation sizing problem consists
Some of the decision variables of this problem, which corresponds to
the third step shown in Fig. 1, are related to the three subproblems
shown in Fig. 2, which include (a) the generation sizing problem, (b) the
network design problem and (c) the clustering problem. Fig. 2 also
shows the most frequent objective functions considered for this problem
in the context of rural electrification. These three problems are inter­
connected: their solutions are not independent of the remaining prob­
lems. For example: in order to determine the generation and the network
of a mini-grid, it is necessary to solve the generation sizing and network
design problems for that specific mini-grid. However, it is necessary to
group the consumers first to determine the candidate mini-grids (clus­
tering problem), but a proper clustering of consumers into mini-grids
needs to consider all the costs related to mini-grids in the process, so
the clustering problem depends on the generation sizing and network
design results. It is important to highlight that the generation sizing and
the network design problems need to be solved a significant number of
times in the large-scale rural electrification planning problem, so a
direct application of methods that were designed to solve a single
instance of these problems may fail for computational reasons.
The clustering problem consists in grouping the consumers into the Fig. 3. REM’s grid extension clusters and the already-existing distribution
best candidate systems (isolated, microgrids, and grid extensions). There network (represented with black lines).

Fig. 2. The rural electrification problem.

4
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

of calculating the best generation design for a given off-grid system and meet the demand during the night.
its corresponding demand. The inputs of the tools that aim at this The network design problem consists of obtaining the best power
problem usually include demand patterns, costs, and technical charac­ distribution network for a given set of consumers. This definition in­
teristics of the generation technologies. There are reviews related to the cludes the calculation of a network for a microgrid and a grid-extension.
methods used to sizing hybrid energy systems [18] as well as their The input of the tools that focus on this problem usually includes the
configurations and control methods [26]. The main software tools that location of the consumers, the costs, and the technical information about
aim at sizing hybrid energy systems are discussed in Ref. [19]. Some of the available network components (lines and transformers). These tools
the most widely-known microgrid generation design tools are the also require information about the already existing network to calculate
Hybrid Optimization Model for Multiple Energy Resources (HOMER) grid-extension designs. A review of the methodologies and tools used to
[27] and the Distributed Energy Resources - Customer Adoption Model calculate the power distribution network has been recently published
(DER-CAM) [28]. Fig. 4 shows an example, obtained with the REM [29]. One of the most powerful existing network design tools is the
model, of the daily dispatch of a mini-grid where a load-following Reference Network Model (RNM) [30], which is used in the REM model
strategy is applied. A load-following strategy is a heuristic dispatch to calculate network designs for grid extensions and mini-grids. Fig. 5
where the resources used to meet the demand always follow the same shows the projection onto Google Earth Pro of a grid-extension design
order: solar, battery, and diesel. If there is solar energy available after calculated with RNM, which calculates the layout of the Medium
meeting the demand, then it will be used to charge the battery. The Voltage (MV) and Low Voltage (LV) lines, and the location of the
diesel generator is not used to charge the battery in the load-following transformers.
strategy. The dispatch shown in Fig. 4 uses solar energy to meet the This review focuses on tools and methodologies that address the
demand and charge the batteries during the day. The energy stored in large-scale techno-economic rural electrification planning problem.
the batteries is used to meet the demand in the evening, and the diesel There are several reasons to include only large-scale tools and methods
generator meets the demand at night. However, the diesel generators in the review. In the first place, some reviews already deal with the
make noise when they operate, which is particularly unpleasant during methods [18] and tools [19] at a small-scale level. In the second place,
the night. A cycle-charging dispatch strategy could be implemented to the nature of the problem is different since, in most cases, there is no
avoid using the diesel generator during the night. The circle-charging need to cluster the consumers when a village or a settlement is electri­
dispatch strategy can operate the diesel generator at its maximum ca­ fied because all the village or settlement is electrified altogether as a
pacity when it is used, and charge the battery with the energy left after single system. Finally, the computational resources needed to solve a
meeting the demand. With the cycle-charging strategy, the diesel small-scale problem are lower, allowing the use of classical optimization
generator could operate at its maximum capacity to meet the demand techniques or computationally intensive procedures that would fail in a
and charge the batteries during the evening, and using the batteries to large-scale problem.

Fig. 4. REM’s daily sample dispatch of a mini-grid. The black line represents its total demand.

5
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

Fig. 5. REM’s network design for a grid extension. The already-existing distribution network is represented with black lines.

The next section introduces a formulation of the techno-economic that go beyond cost, such as environmental impact. Other relevant
rural electrification planning problem, which is used in Section 4 to factors are social impact (measured by the amount of Non-Served En­
classify the tools and methodologies that address this problem at a large- ergy (NSE) or job creation) and regulatory conditions (such as subsidies
scale scope. or tax reduction for certain types of technologies).
In addition, it is vital for the success of a project to consider stake­
3. The rural electrification problem: a comprehensive holder preferences. Often, this is undertaken by generating a small list of
formulation non-dominated, feasible solutions. Then, the preferred solution is
identified as the candidate solution with the highest acceptance among
This section aims to provide a general perspective of the rural elec­ stakeholders [6]. However, Section 4 shows that there is no rural elec­
trification problem, pointing to the main approaches in the literature. trification planning tool or methodology that optimizes a multicriteria
The existing mathematical formulations available describe the problem objective function that involves several of these factors.
only partially as they focus on a single village or community [31,32], or As mentioned in Section 2, there are three different possibilities or
include a reduced level of detail [33]. A general formulation –which was modes when it comes to providing access to electricity: though an
missing in the literature– is very valuable to build comparisons and focus extension of the already existing network, using a microgrid or with
discussions. isolated systems. Fig. 6 shows an example of a mixed rural electrification
Large-scale tools deal with an extensive area such as a region or a plan.
country and consider only a purely economic criterion for computa­ The electrification plan comprises the following decisions:
tional reasons. On a smaller scale, it is possible to consider other factors

Fig. 6. An example of a rural electrification problem and solution.

6
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

� the best electrification mode for each building considered (grid Table 1
extension, microgrid or standalone system); Formulation constraints.
� the distribution network design for each microgrid and grid Group Description
extension;
Distribution network They determine the layout and capacities of lines and
� the generation design and dispatch strategy of each microgrid and constraints transformers included in distribution networks of
stand-alone system (including technology, size, and location); microgrids and grid extensions. They include geometric
� the location and characteristics of any additional MV/LV substations. considerations (constraints from 1 to 5) and electric
criteria (constraints from 6 to 13) that are frequently
considered in the network design problem shown in
The rest of this section comprehensively describes the rural electri­ Fig. 2.
fication problem. The formulation proposed has been developed to serve Generation constraints They determine the capacities of the off-grid equipment
as an illustration for the general statement of the problem and its most (constraints from 14 to 17) and the hourly dispatch
important features. The following simplifications, accurate but useful, (constraints from 18 to 24) of each microgrid and isolated
system. They are related to the generation sizing problem
have been adopted:
shown in Fig. 2.
Cost constraints They model the way that costs are computed (constraints
1. The formulation considers only generation costs and disregards the from 25 to 31) for each isolated system, microgrid, and
cost of other elements such as inverters and charge controllers, which grid extension design. The costs included account for the
are comparatively much cheaper. investment and operation cost of the elements of the
network and generation equipment.
2. Only one line can be installed between each pair of nodes. Parallel
lines are not considered in most projects, as demand profiles are
usually low compared to line capacities, so that parallel lines are not
usually useful. XLmnvc � Вmz 8ðm; n; c; zÞ 2 N 2 CZ; v ¼ fMVg 2 V (3)
3. The cost of losses is calculated in a simplified way.
4. The formulation assumes that upstream impacts are negligible. This XLmnvc � Вnz 8ðm; n; c; zÞ 2 N 2 CZ; v ¼ fMVg 2 V (4)
includes changes in the system-wide dispatch or needs for rein­ If a supply node does not have an MV/LV substation, then it cannot
forcement in the transmission network. be connected to a low-voltage line:
5. Standard network constraints such as voltage drops and reliability X
constraints are not considered given that they increase problem XLsnvc � Вnz ​ ​ ​ 8ðs; nÞ 2 SN; v ¼ fMVg 2 V (5)
complexity considerably. They can be included in subsequent c2C

analyses. The balance of generation minus demand at each node and the power
6. The capacity of the generation elements of the microgrid is consid­ used to charge batteries is equal to the power transmitted to the nodes
ered continuous, and linear piecewise functions approximate the that are connected to it (Kirchhoff’s First Law):
investment cost of these elements.
GENih þ DISCHARGEih CHARGEih demandih þ NSEih
7. Generators installed in a microgrid can be located only in buildings
that belong to the microgrid itself, and not outside of it. εnetwork
X (6)
8. The existing distribution network is discretized into a finite number ¼ Pijhvc 8i; h 2 NH
of candidate connection points, which are considered for grid-
j;h;v;c 2 NHVC

extension designs. Power flow equations must be satisfied (Kirchhoff’s Second Law):
� Θih Θjh
These simplifications are relatively widespread and appear in most plimijvc ⋅ 1 XLijv � Pijhvc ⋅ sbase � plimijvc
xijvc (7)
related works on this subject. �
Subsection 3.1 details the formulation. ⋅ 1 XLijv 8ði; j; h; v; cÞ 2 N 2 HVC

Θih ¼ 0 8h 2 H; i ¼ f1g 2 N (8)


3.1. Formulation
Power flows are bounded by capacity:
It is important to highlight that, although the rural electrification
planning problem has a dynamic and stochastic nature, the authors have plimijvc ⋅ XLijvc � Pijhvc � plimijvc ⋅ XLijvc 8ði; j; v; cÞ 2 N 2 VC (9)
opted for a static formulation, which is consistent with the approach The variable Вiz takes the value 1 if and only if CAPiz is strictly
found in the full list of the literature surveyed: for the time being, no greater than zero:
large-scale tool or methodology can handle multiple time horizons or
include robust optimization methods to effectively cope with un­ CAPnz
� Вnz � M⋅CAPnz 8ðn; zÞ 2 NZ (10)
certainties, which are frequent in developing countries due to the lack of M
reliable data. In this subsection, the constraints are classified into several The capacity of each MV/LV substation (if installed) is greater than
groups, which are described in Table 1, and the objective functions or equal to the power flow that flows from it:
considered are introduced later. X
CAPiz � Pþ
ijhvc M ⋅ ð1 Вiz Þ 8ði; zÞ 2 NZ (11)
j;v;c2NVC
3.1.1. Distribution network constraints
Demand nodes can host generation technologies different from grid
Pþ 2
ijhvc � Pijhvc 8ði; j; h; v; cÞ 2 N HVZ (12)
extension:
Demand at each node bounds NSE:
CAPig ¼ 0 8i 2 I; g ¼ fgrid extensiong 2 G (1)
NSEih � demandih 8ði; hÞ 2 IH (13)
Two nodes can only be connected through one line:
X
XLmnvc � 1 8ðm; nÞ 2 N 2 (2) 3.1.2. Generation constraints
v;c2VC The total generated power is the sum of the power generated by
If a node does not have an MV/LV substation installed, then it cannot technology:
belong to the medium voltage distribution network:

7
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

X
GENih ¼ TECHGENiht 8ði; hÞ 2 IH (14)
t2T

Limits for the capacity of the generation elements and the battery:

capacity g � CAPig � capacityg 8ði; gÞ 2 IG (15)

The maximum generation allowed for each technology depends on


the capacity installed and a generation profile that depends on technical
parameters of the generation technology and resource availability (such
as solar irradiance or wind speed, among others). In the cases of MV/LV
substations and diesel generators, the generation profile may be
constant: Fig. 7. Linear interpolation of the investment costs.

TECHGENiht � CAPit ⋅ profileiht 8ði; h; tÞ 2 IHT (16)


X
The minimum generation allowed for each generation technology BINCiyc ¼ 1 8ði; gÞ 2 IðYnfgrid extensiongÞ (25)
depends on the capacity installed and a minimum generation profile.
c2C

This bound might not be useful if there is no actual minimum generation Only the capacity that is related to the active cost segment is
for a capacity: effective:
CAPit ⋅ profile iht � TECHGENiht 8ði; h; tÞ 2 IHðTnfgensetgÞ (17) cinvyðc 1Þ ⋅ BINCiyc � CDECiyc � cinvyc ⋅BINCiyc 8ði; y; cÞ

The state of charge of the battery at any hour depends on the state of 2 IðYnfgrid extensiongÞC (26)
charge of the previous hour and on how much the battery has been The capacity of an element is equal to its active capacity:
charged or discharged in the previous hour: X
CAPiy ¼ CDECiyc 8ði; mÞ 2 IðYnfgrid extensiongÞ (27)
SOCih ¼ SOCiðh 1Þ c2C

εcharge ⋅CHARGEiðh 1Þ þ DISCHARGEiðh εdischarge


1Þ =
The investment cost of an element is calculated using linear inter­
þ 8ði; hÞ
maxcapacityo polation between the nearest two elements of the catalogue:
2 IðHnf1gÞ; o ¼ fbatteryg 2 O (18)
X �
Constraints on the state of charge of the batteries: CINViy ¼ cinvyðc 1Þ ⋅BINCiyc þ CDECiyc cinvyðc 1Þ ⋅BINCiyc

(28)
c2C
!
soc � SOCih � soc 8ði; hÞ 2 IH (19) cinvyc cinvyðc 1Þ
⋅ 8ði; yÞ 2 IðYnfgrid extensiong Þ
Limits for the speed of charge/discharge: capyc capyðc 1Þ

charge i ⋅CAPig � CHARGEih �chargei ⋅CAPig 8ði;hÞ 2IH;g ¼fbatteryg 2 G The total investment cost associated with an element is the sum of
(20) the corresponding investment costs through all nodes:
X
TOTCINVy ¼ CINViy 8y 2 Ynfgrid extensiong (29)
discharge i ⋅ CAPig � DISCHARGEih � dischargei ⋅ CAPig 8ði; hÞ 2 IH; g i2I
¼ fbatteryg 2 G
The investment cost of the lines is calculated using the binary vari­
(21) ables related to their connections:
If a diesel generator is installed, there is a minimum load restriction X �
CINVy ¼ XLijvc ⋅ Clijvc ; y ¼ fgrid extensiong 2 G (30)
that must be met if the generator is on: m;n;v;c2 N 2 VC

TECHGENtih � CAPit ⋅ profile iht Mð1 Ξiht Þ 8ði; hÞ 2 IH; t ¼ fgensetg 2 T The diesel cost for a diesel generator depends on the fuel cost, the
(22) efficiency of the generator and how much power is generated by the
diesel generator:
TECHGENiht � M⋅Ξiht 8ði; hÞ 2 IH; t ¼ fgensetg 2 T (23) �X �
CFUELi ¼ TECHGENiht ⋅cfuel⋅εdiesel 8i 2 I; t ¼ fgensetg 2 T (31)
The diesel generator produces emissions when it operates: h2H

DEih ¼ GENtih ⋅εdiesel⋅emissions 8ði; hÞ 2 IH; t ¼ fgensetg 2 T (24)


3.1.4. Objective functions
3.1.3. Cost constraints Regarding the objective functions, this formulation considers three
As shown in Fig. 7, the investment cost related to generation tech­ main possible terms in the objective function, describing the economic,
nologies and batteries is interpolated using linear piecewise functions. social, and environmental factors associated with the electrification
These piecewise linear functions should reflect the economies of scale solutions. The economic term in the objective function minimizes the
associated with larger capacities in generation technologies. total investment and running costs of the electrification plan. Since the
Where cinvyc is the investment cost of the y-th generation technology useful life of the elements can be very different (for example, the useful
that has capacity c (which is the element capyc ). The formulation uses a life of a line can be over 20 years whereas the useful life of a battery is
binary variable BINCiyc that takes the value 1 if y-th generation tech­ usually around 5) the economic parameter that is minimized in this
nology installed at node i has a capacity that lies between capyðc 1Þ and formulation is the total annualized cost, which is composed of invest­
ment, operation and maintenance and fuel costs (which result from the
capyc . The formulation also uses a positive variable CDECiyc that is equal
use of diesel generators).
to the capacity of the y-th generation technology installed at node i if
that capacity lies between capyðc 1Þ and capyc , and takes the value zero in X CINVy X X
f1 ¼ þ CINVy ⋅cO&My þ CFUELi (32)
the remaining cases. Specifically, the equations are: y2Y
annualfactory y2Y i2I
Exactly one of the cost segments is active:

8
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

The social term weights in NSE: dimension and their corresponding weights.
X Regarding generation sizing, the most frequently considered tech­
f2 ¼ NSEnh (33) nologies include solar panels, wind, diesel generator, hydro, and
biomass. Here it is also valued how easy it is to incorporate a new
n;h2NH

Finally, the environmental term incorporates the effects of emissions: generation technology into the tool or methodology. The generation
X sizing methods range from not sizing it but calculating the LCOE to
f3 ¼ DEnh (34) classical optimization methods that guarantee the optimum solution at
n;h2NH
the expense of a high computation cost and the possibility of considering
These functions are the ones usually considered in small-scale several architectures (that could be in Alternating Current (AC) and
problems. Although they could also be used in a large-scale project, Direct Current (DC)) or customizing it is very valuable. The generation
finding a subset of the Pareto frontier to decide the final solution usually equipment may be located in several spots or a single place, and the most
requires solving several mono-objective problems and this task could common criteria considered when solving this problem are cost, NSE,
require an unmanageable amount of computation time for a large-scale and emissions.
case. Regarding network design, network layout calculation methods
In addition, it should be noted that many large-scale tools and range from not calculating it and just providing an LCOE of the network
methodologies use the Levelized Cost of Electricity (LCOE) for deter­ cost to using classical optimization techniques that ensure optimality at
mining the best electrification mode in a simplified manner. The the expense of a very high computation cost. Some tools consider usual
formulation presented here does not use this concept but instead cal­ electric constraints such as maximum voltage drop allowed when
culates a total cost comprehensively. designing the network of a mini-grid or a grid extension, and they ensure
This general formulation of the problem can serve as a common the electric feasibility of the network performing power flows. Others
ground to discuss the different perspectives of the problem, as well as rely entirely on geometric calculations, but their network designs would
providing a starting point to describe tools and methodologies for rural need to be checked before implementation since they could be electri­
electrification planning in Section 4. cally unfeasible. Most tools only consider the impact of grid extensions
on the distribution network, but connecting large regions to the current
4. Classification of tools and methodologies distribution network could imply reinforcing the transmission network
and installing new generation power plants. Although upstream re­
There is an array of tools and methodologies that provide valuable inforcements may be neglected in small-scale projects, they may have a
assistance in rural electrification projects. The adequate tool or combi­ significant impact on the electrification solution of a large-scale rural
nation of tools for a specific project depends on its scope. This section electrification plan. Similarly, the viability and cost of a network design
analyses the main features of rural electrification planning tools and may be heavily influenced by the terrain slopes, and forbidden zones
methodologies to provide useful insights to the readers. such as the lakes and mountains of the analysis region. In addition,
In this section, the tools and methodologies are classified according future expenses may be avoided if the networks of mini-grids are grid-
to their modeling complexity. They range from first-pass tools, which compatible. The most common criterion considered when solving this
offer information for a pre-feasibility analysis with a short computation problem is cost.
time, to the detailed analysis tools, which provide a rural electrification Regarding the objective function, there are two ways of dealing with
plan with electric designs for all the grid extensions and microgrids. a multicriteria problem. The first one consists of defining a scalar
Fig. 8 provides an overview of the complexity levels. Although the objective function that includes all the criteria involved; an example
proposed classification is undoubtedly useful, the boundaries among would be the weighted sum method. The second one involves working in
levels are not clear-cut, so there could be cases where a tool or meth­ a multidimensional solutions space where each axis relates to an
odology could arguably belong to one or another level. It is important to objective function. Here, multicriteria optimization techniques are
highlight that a higher modeling complexity comes at the expense of applied to obtain the Pareto frontier or set of non-dominated solutions,
higher computation time. and the user eventually selects one of these solutions. Some tools such as
For each level, a classification framework based on the formulation Logiciel d’Aide � a la Planification d’Electrification
� Rurale (LAPER) and
introduced in Section 3 is proposed, which is coherent with the problem GEOSIM include terms associated with non-techno-economic factors
definition established in Section 2. This framework applies an expert- such as political and development criteria for villages, allowing the user
assigned score to each tool and methodology in three different di­ to perform sensitivity analyses and measure the impact of these factors
mensions associated with its generation sizing procedure, its network in the electrification solution. Most tools and methodologies perform a
design algorithm, and its objective function. Then, each tool is ranked on cost minimization, but other criteria such as minimizing emissions and
each dimension with an average weighted sum of the corresponding NSE are relevant too.
criteria which is scaled to the interval [0, 1], where the value zero A crucial point that should be highlighted is that all existing rural
corresponds to the worst possible score, and the value one corresponds electrification planning models fail to deal with uncertainty, although it
to a perfect score. Table 2 shows the criteria considered for each
Table 2
The classification framework.
Dimension Aspects Weight

Generation sizing Generation technologies considered 5


Generation sizing method 5
Location of the generation site 3
Mini-grid architecture 3
Criteria considered 3
Network design Layout calculation method 5
Network levels considered 3
Topography 2
Criteria considered 3
Objective function Multicriteria approach 2
User-defined terms 2
Criteria considered 3
Fig. 8. Classification of tools by modeling complexity.

9
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

plays a critical role in energy planning [34]. This limitation can be related to the expressions involved in the LCOEs calculations. Recently,
mitigated using scenario analysis, which involves defining several sce­ OnSSET has been updated to include a clustering algorithm that merges
narios where the uncertain parameters take several values and analyzing nearby cells and adapted its grid-extension cost estimations to deal with
the corresponding impact on the planning solution. However, this is only clusters [44].
a partial solution [35]. The OnSSET methodology has been applied to Sub-Saharan Africa
The critical uncertainty in the problem is introduced by demand. [45], Nigeria [46], Ethiopia [47]. This model has also been applied in
Demand is hard to estimate in developing countries, as it depends on combination with a more general energy modeling tool, Open Source
several socio-economic factors such as population growth. An inaccurate Energy Modeling System (OSeMOSYS), to produce more detailed results
forecast of demand could translate into an oversizing of the systems if in Sub-Saharan Africa [48] and Kenya [49], and the Division of Energy
the estimated demand is higher than the real one. On the other hand, it System Analysis of KTH has carried out projects applying it to Nigeria,
could also lead to and undersizing of the systems, which will have lower Tanzania, Zambia, Afghanistan [50] and Benin [51] in cooperation with
reliability than expected [36]. the WB, among others [52].
Other relevant parameters are related to off-grid technologies used in Another relevant tool that belongs to this level is IntiGIS [53], which
mini-grid. Regarding the diesel generator, there is uncertainty related to was developed at Centro de Investigaciones Energ� eticas, Medi­
the fuel cost and fuel transportation from the nearest supply point to the oambientales y Tecnolo �gicas (CIEMAT). Preliminary versions of this tool
mini-grid [37]. Wind profiles are very locational and can be modeled are SOLARGIS [54] and SOLARGIS2 [55]. Ref. [56] shows an applica­
following a Weibull distribution, whereas solar irradiance can be tion in Lorca (Spain). IntiGIS is a more advanced version with enhanced
modeled following a Beta distribution [38]. There is plenty of literature capabilities, which also groups consumers into cells and estimates their
that deals with these uncertainties when optimizing the generation size LCOE with analytic expressions. This model considers wind, solar, and
of a single mini-grid [39], but those techniques are yet to be imple­ diesel as off-grid technologies and the grid extension possibility as an
mented into large-scale rural electrification planning tools. Uncertainty on-grid solution. An application of IntiGIS to Latin America is presented
has been thoroughly studied in the related problems of distribution in Ref. [57]. An upgraded version, IntiGIS 2.0, has been developed
planning [40] or transmission planning [41,42]. These methods could recently [58,59], but the authors have not found publications in the
also be adapted to rural electrification planning tools. literature related to it. The IntiGIS tool includes several
The remaining of this section is devoted to analyzing the features of hybrid-configurations as off-grid alternatives and applies analytic ex­
tools and methodologies that belong to each level with the proposed pressions to size their components.
framework. As the modeling complexity increases, so does the number LAPER [60] also belongs to this level since the user has to input the
of constraints considered from the comprehensive formulation. initial grid-extension network manually. Afterwards, the model starts
disconnecting villages based on cost comparisons. LAPER also estab­
4.1. Pre-feasibility studies lishes an electrification schedule, ranking villages according to several
criteria that are weighted with user preferences. Ref. [61] shows an
Pre-feasibility tools offer a first-pass information grouping the con­ application of the LAPER model in Morroco.
sumers into villages or cells and estimating the lowest-cost electrifica­ The Rapid Electrification Screening Tool (REST) was created to
tion mode for each client using approximated LCOEs or similar provide a quick first-pass estimation that includes the detailed cost of
economic indicators. Their main advantages are low computation time several electrification solutions. Although it was developed to work at a
and usability. Moreover, some of them exploit the benefits of GIS tech­ local level (focusing on a single village or community), this tool has been
nologies to access the necessary data, which could include the location included in this review as it can be applied iteratively to produce results
of power plants, the layout of the power grid, solar irradiance, wind similar to the ones provided by other tools included in this section. The
speed, hydro potential, the layout of the roads and population density, REST tool was developed with the aim of fighting against energy poverty
among others. This feature allows the user to obtain immediate esti­ in Uganda, Tanzania, and Kenya [62].
mations for large-scale areas and perform a sensitivity analysis on the There are also several methodologies in the literature that apply
spot. This feature is the one that adds more value for some of these tools similar procedures. Electrification planning methods based on spatial
since it eliminates the need to gather specific data from different analysis have been applied to Burkina Faso [63] and Africa [64,65].
sources. Ref. [66] studies the influence of several factors in the best electrifica­
These tools can virtually consider any off-grid technology or com­ tion mode, including the distance to the grid and the life cycle cost of
bination of technologies if there are analytic expressions that allow the system. A summary of these tools is provided in Table 3. All of them
economic calculations and input data for the case study although they minimize a cost-related function to obtain the solution, which is usually
provide an estimation of the cells that should be electrified with grid the LCOE although it may be a net present value or an annuity.
extensions, they do not consider the possibility of electrifying several Fig. 9 shows the pre-feasibility classification according to the
cells in the same grid extension. They generally do not provide a grid framework proposed in Table 2. LAPER is the most complete tool in
extension layout or generation designs for the off-grid systems. terms of network design and objective functions, although it does not
These tools do not consider most features of the formulation intro­ provide geospatial capabilities that facilitate data acquisition. IntiGIS is
duced in Section 3. Specifically, they only consider constraint (1) from the most remarkable one regarding generation designs since it includes
the distribution network constraints, and they do not include any gen­ several hybrid configurations among its off-grid candidate solutions, and
eration constraint, being IntiGIS an exception since it sizes off-grid it uses analytic expressions to size off-grid components, going one step
generation with analytic expressions. beyond other pre-feasibility tools.
Perhaps the most advanced tool of this category is OnSSET [43], Fig. 9 shows that there is room for improvement in the objective-
which was developed at the Royal Institute of Technology (KTH) as an function dimension. There is a trade-off between accuracy and compu­
open-source tool. OnSSET can estimate the electrified consumers using tation time in the generation sizing and network design dimensions, but
GIS data, group the unelectrified consumers in square cells of 1 km2 and the addition of user-defined criteria in a similar manner as LAPER and
calculate the LCOE of off-grid technologies that include solar, wind, Ref. [66] should improve the tools without a significant increase in
hydro, and diesel. This model also incorporates the effect of the topog­ computation time. However, the planer can also consider those criteria
raphy when estimates the LCOE of electrifying a cell with grid extension, later, post-processing the electrification solution that the tools provide.
adding cost penalties on areas of high elevation and slope gradients.
However, it does not consider alternative routes between the network
and a cell that could lower cost. Ref. [11] shows detailed information

10
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

Table 3
Pre-feasibility tools and methodologies. NL¼Network Layout, T ¼ Topograhy
Year Tool Granularity Level Network Design Generation Design Objective Function Additional references

NL T Solar Wind Hydro Diesel Biomass

2017 OnSSET [23] Cell No Yes Yes Yes Yes Yes No Cost [11,44,46,47,49]
2016 [63] Cell No No Yes No Yes Yes No Cost –
2013 [64] Cell No No Yes No Yes Yes No Cost [65]
2012 [66] Village No No Yes No No No Yes Costb –
2011 IntiGIS [33] Cell No No Yes Yes No Yes No Cost [56–59]
2008 REST [62] Village No No Yes Yes No Yes No Cost –
2001 LAPER [40] Village Yes No Yes Yes Yes Yes No Costa [61]
a
LAPER includes user-defined terms that measure political, financial resources and development criteria to calculate an electrification schedule, but the final
electrification mode of each village is calculated solely with a cost minimization criteria.
b
The impact of carbon emissions is considered, including cost terms in the calculations of the LCOEs.

Fig. 9. Pre-feasibility tools classification according to the classification framework.

4.2. Intermediate analysis tools previously developed similar (albeit less sophisticated) methodologies
and applied them in Nigeria [68] and Sub-Saharan Africa [69].
Intermediate analysis tools operate similarly, but they apply more Network Planner [70], which was developed by the University of
complicated calculations to network designs, calculating the layout of Columbia, is an open-source tool that belongs to this category. When
the grid extension solution usually with heuristic methods based on calculating off-grid generation designs for villages, it considers diesel,
minimum-spanning-tree algorithms. However, the network design al­ solar and batteries as off-grid technologies, and the model sizes the
gorithms that these tools apply are limited to distance-based constraints, generation with rules of thumb such as the capacity of the battery is a
which implies that a proposed design may not be viable from an electric multiple of the capacity of the solar panel, which is sized for peak
point of view. On the other hand, a tool that calculated off-grid gener­ demand.
ation designs with an algorithm that went beyond rules of thumb would On the other hand, Network Planner uses an iterative algorithm to
belong to this category too. Unfortunately, the authors have only iden­ calculate grid extension layouts [71]. This algorithm compares the best
tified one methodology with this level of complexity in generation de­ off-grid solution for a village with its internal grid-extension cost, which
signs [67], even though incorporating an iterative application of consists of a transformer and the low-voltage network cost, but it does
generation sizing algorithms seems reasonably straightforward [18,26]. not include the cost of the medium-voltage line that connects the village
As these tools are at an intermediate level, there is a partial corre­ to the power grid or another grid-electrified village. For those villages
spondence between the aspects they consider and the formulation with an internal network cost lower than its best off-grid cost, the model
introduced in Section 3. Specifically, they usually consider distribution- calculates the maximum length that this medium-voltage line could
network constraints (1–5) and generation constraints (14, 15) plus a have, so that grid extension is the best option. Then, the model iterates
stylized version of constraints (16) and (17), which bound the genera­ connecting villages where the maximum length is larger than its dis­
tion capacity between zero and peak demand) and (18) (that ensures tance to the network or a village that was connected in a previous
that the demand is met and enables the possibility of transferring energy iteration. Network Planner has been applied in Nigeria [72,73], Liberia
from one node to another). On the other hand, most of the tools in this [74] and Ghana [75].
section focus on cost minimization and consider neither NSE nor The iterative procedure that Network Planner uses to connect vil­
emissions. lages to the grid is also applied in Ref. [76], which was one of the first
Ref. [67] presents the most dominant method in this category, electrification planning methodologies proposed in the literature. A
developed by researchers from the Renier Lemoine Institut (RLI). It similar method is presented in Ref. [77], which introduces an algorithm
generates a raster map to determine the layout of the network. The raster that tries to emulate the development of the power systems to determine
map is created considering topography information such as terrain the expansion of the power grid.
elevation, slopes, and location of protected areas and forests. This Another important tool that belongs to this level is GEOSIM, which
method applies an iterative process to determine the generation design was developed by Innovation Energie D�eveloppement (IED). This tool
of each mini-grid, simulating one year on an hourly basis for each labels some villages as Development Poles (DPs) according to their inner
candidate generation design. The method also includes an initial clus­ potential, which is calculated based on health, local economy and ed­
tering algorithm, and it can divide the plan into several implementation ucation indicators, as well as the distances to the remaining villages.
phases to facilitate project execution. Researchers from the RLI Then, it creates clusters of villages that are electrified together around

11
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

DPs using an algorithm based on the Huff model [78]. The idea of

Intermediate analysis tools and methodologies. NL¼Network Layout, TE ¼ Transmission Expansion, DL ¼ Distribution Layout, UR¼Upstream Reinforcements, EF ¼ Electric Feasibility, T ¼ Topography, HV¼High Voltage,

This method includes a cost input parameter, in $/kWh, to account for power generation and transmission costs. This parameter may be used to include an estimation of the reinforcement costs in an approximate way.
clustering villages is entirely coherent when working with

Additional
references
low-populated settlements that are close to each other and it is not

[71–75]
[68,69]

[80,81]
present in most tools of this category or the previous one.
GEOSIM has been applied in several countries [79], developing


projects and selling licenses. Particularly, it was applied in Benin, Bur­
kina Faso, Cameroon, Ethiopia [80], Madagascar, Mali, Niger, Tanzania,

Composite
Cambodia, and Lao People’s Democratic Republic [81].

Objective

functionb
objective
Function
A summary of the tools and methodologies that correspond to this

Cost

Cost

Cost

Cost

Cost

Cost
Cost
section is provided in Table 4. It is worth noticing that some interme­
diate analysis tools use classical optimization to obtain the layout of the
network, but this comes at the expense of high computation times and a

Sizing Method

Optimization
calculations

calculations

calculations

calculations
Expressions
reduced number of villages that the methodology can process. One of

Exogenous

Exogenous

Exogenous

Exogenous
Classical
Iterative

Analytic
process
these methodologies [82] even considers three different objective

Noc
functions using hierarchical lexicographic programming (the first level
is to maximize the aggregated demand electrified with grid extension,
the second level is to minimize the sum of the distances between grid

Biodiesel
extension villages and the power grid, and the third level minimizes the

Yes
No

No

No

No

No

No
No
inter-distance among grid extension villages). However, in this meth­

Maximize demand covered by the grid, minimize distance among villages and the grid, and minimize dispersion among villages connected to the grid.
odology, it is necessary to know the number of villages that are elec­

Biomass
trified with grid extension beforehand or perform a sensitivity analysis

Yes
No

No

No

No

No

No

No
modifying these parameters. Another methodology that applies opti­
mization to calculate the network layout is [83], minimizing a cost

Diesel
function and comparing the results with the heuristic presented in

Yes

Yes

Yes

Yes

Yes
No

No

No
Ref. [71]. However, this methodology only considers one point to con­
nect villages to the existing grid. Another methodology that uses clas­

Hydro
sical optimization is [33], which applies a Mixed Integer Linear

Yes
No

No

No

No

No

No

No
Programming (MILP) model to determine which cells should be elec­
trified extending the transmission network and which ones should have Generation Design

Wind

Yes

Yes
No

No

No

No

No
No
solar panels. However, the size of the cells is considerably large (2000

It estimates the LCOE to provide the best off-grid supply alternative, but it does not provide optimal generation sizes.
km2), and the off-grid generation is limited to solar panels. There are
Solar

also generic models that use classical optimization techniques such as


Yes

Yes

Yes

Yes

Yes

Yes

Yes
Yes
Energy Flow Optimization Model (EFOM), MARKet ALlocation (MAR­
KAL) or The Integrated MARKAL-EFOM System (TIMES) that were

Classical Optimization (MILP)


developed to provide assistance in a wide range of energy planning programming) þ MST-based
(hierarchical lexicographic

problems and allow the user to customize specific parts to develop a


MST-based calculations

MST-based calculations

MST-based calculations

MST-based calculations
MST-based calculations
Classical Optimization

Classical Optimization
tailormade application for their specific problem. Although some of
Layout calculation

them could be adapted to deal with the techno-economic rural electri­


fication planning up to a certain extent [84], the authors have not found
calculations

any large-scale application in this regard in the literature, although they


(MINLP)

can be combined with the tools described in this review as in the case of
OSeMOSYS and OnSSET [48,49] mentioned in Section 4.1. Generic
models have been applied to small-scale problems (i.e., a problem with a
Yes

No

No

No

No

No

No
No
single village or settlement) [85], although large-scale applications that
T

address this problem have not appeared.


No

No

No

No

No

No

No
No
EF

There are network design algorithms in the literature [86] that could
be extended to consider off-grid electrification solutions and become
Yes
UR

rural electrification planning tools but, since they have not been applied
No

No

No

No

No

No
No
a

to a rural electrification planning problem considering off-grid solutions


so far, they are not included in Table 4.
HV/
MV

MV

MV

MV

MV
MV
MV
No
DL
Network Design

Fig. 10 shows the intermediate analysis classification according to


MINLP ¼ Mixed Integer Non-Linear Programming.

the framework proposed in Table 2. Attempts to deal with this problem


Yes

Yes
No

No

No

No

No
No
TE

with more accuracy when calculating network designs come at the


expense of less detail when computing the generation design. It is worth
Yes

Yes

Yes

Yes

Yes

Yes

Yes
Yes
NL

to notice that some intermediate analysis tools rank higher in network


design than pre-feasibility tools, but this does not happen in generation
Granularity

designs.
Village

Village

Village

Village

Village
Village
Level

Cell

Cell

4.3. Detailed generation and network designs

Some other tools attempt to incorporate the maximum possible level


Network

GEOSIM
Planner

of granularity, offering detailed consumer-by-consumer network designs


Tool

[67]

[82]

[83]

[33]

[77]

[76]

that consider the usual electric constraints (such as maximum voltage


drop allowed) and generation designs based on weather conditions such
Table 4

2019

2016

2016

2015

2013

2011

2010
2000

as solar irradiance levels. This level of detail is obtained, however, at the


Year

expense of much longer computation times and a steeper learning curve


b
a

12
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

Fig. 10. Intermediate analysis tools classification according to the classification framework.

for the user. Besides, these models require a significant amount of input optimizes the network considering catalogues with several lines and
data which are considerably difficult to obtain. transformers, which is a capability that no other large-scale rural elec­
The REM model, which was developed by the Massachusetts Institute trification planning tool has so far.
of Technology (MIT)/Comillas- Instituto de Investigacio �n Tecnolo �gica REM also goes one step further than other tools when incorporating
(IIT) Universal Energy Access Lab as a commercial tool, is the only tool topography, introducing topographical considerations when clustering
that arguably can be classified at this level [10]. consumers and when calculating detailed network layouts with RNM
This tool is the closest match to the formulation provided in Section [88]. RNM also incorporates forbidden zones (areas that lines cannot
3. All distribution-network and generation constraints are considered trespass) in its calculations.
except equation (24), which incorporates diesel emissions. Regarding Another drawback of REM is that RNM was not designed for rural
the objective functions, REM minimizes the total investment and oper­ electrification planning but to help regulators of developed countries
ation cost plus a penalty for NSE. Therefore, although it is a cost- estimate the distribution network cost of large regions. Hence, its al­
minimization tool, it also considers the impact of NSE in the final gorithms could produce results that are not optimal when designing the
solution. network of small microgrids, for example. The network design methods
The core of this tool has three blocks that perform sequential tasks. with a focus on rural electrification described in Section 4.2, however,
The first block calculates generation designs for a set of representative are not as sophisticated as RNM.
candidate off-grid systems using an ad-hoc adaptation of the Hooke and In addition, when connecting many consumers to the existing
Jeeves heuristic algorithm [87] and simulating the behavior of the sys­ network, it may be necessary to reinforce the upstream distribution and
tem for each candidate solution to obtain the amount of NSE and the transmission network. It may even be necessary to consider new power
investment and operation cost. The information related to the remaining plants to satisfy the expected demand in a developing country. There is
generation designs, if needed, is interpolated linearly. plenty of literature related to the problems of reinforcing the network
One of its drawbacks is that the generation technologies considered and expanding the generation, but this problem has not been studied in
are limited to solar and diesel. Adding wind or hydro in REM could the context of rural electrification. Although in a preliminary stage,
imply a significative increase in computation time since it implies several strategies related to considering upstream reinforcements in
increasing the number of search dimensions in the generation sizing REM have already been considered [89]. At this point, no rural elec­
heuristic that REM uses to obtain quasi-optimal generation designs. trification tool that was mentioned in Sections 4.1 And 4.2 incorporates
The second block groups the consumers into clusters that represent an accurate calculation of costs related to these reinforcements.
the candidate systems (stand-alone systems, microgrids, and grid ex­ REM has been applied to case studies in the Vaishali district in India
tensions) for the final electrification solution. REM builds off-grid clus­ [90], Kenya and Colombia [91,92], Rwanda [93], Uganda [94] and Peru
ters considering only off-grid electrification solutions and, in a later [95].
stage, the model considers the possibility of grouping off-grid clusters Fig. 11 shows the detailed generation and network designs tool
forming the grid extension clusters by evaluating the cost of different classification according to the framework proposed in Table 2.
configurations (this time, the possibility of grid extension is considered.
By doing this, REM creates a hierarchical structure of clusters where one
grid extension cluster may contain several off-grid clusters, each of them
having one or several consumers.
The third block evaluates the costs of the clusters and determines the
best electrification solution. Specifically, REM compares the cost of a
grid extension design for a grid extension cluster with the cost of off-grid
solutions (either solar-home-systems or a microgrid) for the off-grid
clusters that belong to this grid extension cluster. This implies that a
grid extension cluster could be electrified with off-grid systems if their
electrification cost is lower, and a combination of off-grid clusters could
be electrified with a grid extension if the cost of their grid extension
cluster is lower. REM also allows an exhaustive evaluation of clusters
that considers, for example, the possibility of electrifying off-grid clus­
ters with grid extensions [10]. REM calculates accurate network designs
for each candidate microgrid and grid extension design to perform
cost-comparisons among clusters. Specifically, REM calculates network Fig. 11. Detailed generation and network design tools classification according
designs using RNM [30]. RNM considers electric constraints and to the classification framework.

13
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

Although REM calculates generation designs more accurately than with a limited number of generation technologies (solar panels,
most tools analyzed here (the procedure presented in Ref. [67] also batteries, and a diesel generator). It would be interesting to upgrade
simulates the behavior of the system for a whole year), this comes at the them so that they include more generation technologies, although
expense of a limited number of generation technologies that could only this could increase significantly the computation time needed in a
be increased with a significant burden of computation time so, in the large-scale planning case.
framework, this tool does not outrank the remaining ones studied in � The network design algorithms that have been incorporated into
Section 4 regarding generation designs. However, there is a significant rural electrification tools either neglect electric constraints or were
difference when it comes to network designs since REM considers created for developed countries (this is the case of RNM, which is
electric constraints and optimizes the layout, considering the topo­ used in REM). It would be interesting to develop network design
graphical features of the terrain. algorithms that are specifically aimed at rural electrification projects
that can cope with voltage regulations and power-flow constraints.
5. Conclusions and needs for further research � Future large-scale rural electrification planning tools should be bet­
ter equipped to deal with uncertainty as there is a scarcity of data in
After presenting a comprehensive but concise definition of the rural developing countries, and the existing information is not always
electrification planning problem and a comprehensive formulation that reliable. However, the addition of robust optimization or stochastic
was missing in the literature, this paper has reviewed the main software techniques could be a challenge as it would involve a significant
tools and methodologies that have been applied to solve it. This paper increase in computation time. Although there are already methods in
classifies them using three levels, based on modeling complexity and the literature that successfully address the uncertainty of specific
computation speed. parameters, they were designed to deal with a single mini-grid or
The first level contains pre-feasibility tools, which usually group the grid-extension. In the large-scale rural electrification planning
consumers into cells and calculate their LCOE with several off-grid problem, however, tools may deal with a vast number of mini-grids
technologies as well as with a grid extension solution. These tools and grid-extensions so a straightforward application of these
offer valuable first-pass information, but they do not calculate a network methods may fail for computational reasons.
layout for the grid extension, and they do not size off-grid generation.
The second level contains intermediate analysis tools, which generally The authors hope that these outstanding issues are tackled soon by
calculate the grid extension layout using minimum-spanning-tree stra­ new research works so that the challenge of universal access to elec­
tegies. Some of them also size the generation of off-grid systems with tricity can be better addressed efficiently and comprehensively.
analytic expressions and iterative methods. The third level contains
detailed generation and network analysis tools. They operate at the
Declaration of competing interests
maximum level of granularity (consumer by consumer) instead of
grouping consumers into villages or cells, and they calculate accurate
The authors declare that they have no known competing financial
generation designs using optimization and simulating the hourly
interests or personal relationships that could have appeared to influence
dispatch of the system. They also calculate network designs considering
the work reported in this paper.
the usual electric constraints, the effect of the topography when calcu­
lating the best layout, and the possibility of having lines and trans­
Acknowledgements
formers of different sizes.
As shown in this review, many tools are currently emerging to tackle
The authors thank Ignacio P�
erez Arriaga for reviewing a preliminary
the global electrification challenge. However, there are several spots
version of this paper.
where these developments are lagging, and several needs for research
are identified.
References
� Most rural electrification planning tools focus solely on economic [1] International Energy Agency, , International Renewable Energy Agency, United
parameters when designing the best rural electrification plan for a Nations Statistics Division, World Bank, and World Health Organization [Online].
given region. Some of these indirectly consider other factors into the Available:. Tracking SDG7. The Energy Progress Report 2019. World Bank; 2019.
https://trackingsdg7.esmap.org/data/files/download-documents/2019-Tracking%
global cost, such as a penalty for NSE. However, there are no tools
20SDG7-Full%20Report.pdf [Accessed: 10-Aug-2019].
that perform multiobjective optimization (for instance, considering [2] Livemint. The problem of lack of rural electricity demand [Online]Available:
cost, emissions, or NSE) allowing them to analyze their trade-offs. https://www.livemint.com/Opinion/nrgFBVsILzEtbJf7pwfH3H/The-problem-of-
� Existing rural electrification planning tools and methodologies are lack-of-rural-electricity-demand.html [Accessed: 10-Aug-2019].
[3] International Energy Agency and World Bank. Sustainable energy for all
static and do not consider the effects that a multi-stage rural elec­ 2015—progress toward sustainable energy 2015: global tracking framework
trification plan could have on the final solutions. However, doing this report. 2015 [Online]. Available: https://www.seforall.org/sites/default/files
would be relevant, given that the electricity demand in rural areas /GTF-2105-Full-Report.pdf. [Accessed 17 August 2019].
[4] Urpelainen J. Grid and off-grid electrification: an integrated model with
can be expected to increase through time. Moreover, none of them applications to India. Energy Sustain. Dev. Apr. 2014;19:66–71.
considers grid-compatible microgrids in their calculations since they [5] Tenenbaum B, Greacen C, Siyambalapitiya T, Knuckles J. From the bottom up: how
only consider microgrids or mini-grids that operate on islanded small power producers and mini-grids can deliver electrification and renewable
energy in Africa. World Bank; 2014 [Online]. Available: http://elibrary.worldbank
mode, although in the long term these microgrids could be replaced .org/doi/book/10.1596/978-1-4648-0093-1. [Accessed 11 June 2017].
with grid-extension designs. [6] Santos P�erez FJ. Metodología de ayuda a la decisi� on para la electrificaci�
on rural
� No rural electrification tool or methodology can estimate accurately apropiada en países en vías de desarrollo. PhD thesis (in Spanish). Universidad
Pontificia Comillas, Madrid. School of Engineering.; 2015.
the costs of the upstream reinforcements that would be necessary to [7] Borofsky Y. Towards a transdisciplinary approach to rural electrification planning
implement in the power network or the generation system if many for universal access in India. Master thesis. Massachusetts Institute of Technology.
consumers are connected to the grid. These reinforcements are Department of Urban Studies and Planning.; 2015.
[8] European Commission and Energy Facility. Sustainability - business models for
negligible in small-scale projects, but their effect is certainly worth
rural electrification. 2012 [Online]. Available: https://europa.eu/capacity4dev/fi
considering in large-scale projects. New developments are necessary le/10582/download?token¼yK25eQn9, . [Accessed 18 April 2018].
to incorporate the effects of rural electrification on the transmission [9] Gonz�alez-García A, P�erez-Arriaga I. Sistemas integrados de suministro el�ectrico
network. aislado y conectado a la red: innovaci�on y gobernanza organizativa para el acceso
universal. Econ Ind 2018;408:101–12 (in Spanish).
� The existing generation sizing algorithms that optimize the design by [10] MIT & IITComillas Universal Energy Access Lab. Computer-aided electrification
performing a detailed simulation of the entire year can only cope planning in developing countries: The Reference Electrification Model (REM).

14
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

Working paper. Ref. IIT WP 18-112-A.; 2018 [Online]. Available: https://www.iit. [41] Tsamasphyrou P, Renaud A, Carpentier P. Optimization. In: Nguyen VH, Strodiot J-
comillas.edu/docs/IIT-18-112A.pdf [Accessed: 10-Oct-2019]. J, Tossings P, editors. Transmission network planning under uncertainty with
[11] Nerini FF, Broad O, Mentis D, Welsch M, Bazilian M, Howells M. A cost comparison benders decomposition, vol. 481. Berlin, Heidelberg: Springer Berlin Heidelberg;
of technology approaches for improving access to electricity services. Energy Jan. 2000. p. 457–72 [Online]. Available: http://link.springer.com/10.1007/9
2016;95:255–65. 78-3-642-57014-8_30. [Accessed 13 August 2019].
[12] International Energy Agency. World Energy Outlook 2018. International Energy [42] Lawson A, Goldstein M, Dent CJ. Bayesian framework for power network planning
Agency; 2018 [Online]. Available: https://www.oecd-ilibrary.org/content/publ under uncertainty. Sustain. Energy Grids Netw. Sep. 2016;7:47–57.
ication/weo-2018-en. [43] Mentis D. Spatially explicit electrification modelling insights. PhD Thesis. Sweden:
[13] Gordon SB, et al. Respiratory risks from household air pollution in low and middle KTH Royal Institute of Technology. Division of Energy Systems Analysis.; 2017.
income countries. Lancet Respir. Med. Oct. 2014;2(10):823–60. [44] Korkovelos A, Khavari B, Sahlberg A, Howells M, Arderne C. The role of open
[14] Chattopadhyay D, Kitchlu R, Jordan RL. Planning for electricity access. Live wire, access data in geospatial electrification planning and the achievement of SDG7. An
2014/35. World Bank; 2014 [Online]. Available: https://openknowledge.wor OnSSET-based case study for Malawi. Energies Apr. 2019;12(7):1395.
ldbank.org/handle/10986/20596. [Accessed 28 April 2018]. [45] Mentis D, et al. Lighting the World: the first application of an open source, spatial
[15] Resch B, et al. GIS-based planning and modeling for renewable energy: challenges electrification tool (OnSSET) on Sub-Saharan Africa. Environ Res Lett Aug. 2017;
and future research avenues. ISPRS Int J Geo-Inf May 2014;3(2):662–92. 12(8):085003.
[16] Jebaraj S, Iniyan S. A review of energy models. Renew Sustain Energy Rev Aug. [46] Mentis D, et al. A GIS-based approach for electrification planning—a case study on
2006;10(4):281–311. Nigeria. Energy Sustain. Dev. Dec. 2015;29:142–50.
[17] Kaundinya DP, Balachandra P, Ravindranath NH. Grid-connected versus stand- [47] Mentis D, et al. The benefits of geospatial planning in energy access – a case study
alone energy systems for decentralized power—a review of literature. Renew on Ethiopia. Appl Geogr Jul. 2016;72:1–13.
Sustain Energy Rev Oct. 2009;13(8):2041–50. [48] Mentis D, et al. Linking the open source, spatial electrification tool (ONSSET) and
[18] Luna-Rubio R, Trejo-Perea M, Vargas-V� azquez D, Ríos-Moreno GJ. Optimal sizing the open source energy modelling system (OSeMOSYS), with a focus on Sub-
of renewable hybrids energy systems: a review of methodologies. Sol Energy Apr. Saharan Africa. Presented at the EGU General Assembly Conference Abstracts
2012;86(4):1077–88. 2017;19:15758 [Online]. Available: http://adsabs.harvard.edu/abs/2017EGUGA..
[19] Sinha S, Chandel SS. Review of software tools for hybrid renewable energy 1915758M. [Accessed 27 March 2018].
systems. Renew Sustain Energy Rev Apr. 2014;32:192–205. [49] Moksnes N, Korkovelos A, Mentis D, Howells M. Electrification pathways for
[20] Neves D, Silva CA, Connors S. Design and implementation of hybrid renewable Kenya–linking spatial electrification analysis and medium to long term energy
energy systems on micro-communities: a review on case studies. Renew Sustain planning. Environ Res Lett Sep. 2017;12(9):095008.
Energy Rev Mar. 2014;31:935–46. [50] Korkovelos A, Bazilian M, Mentis D, Howells M. A GIS approach to planning
[21] Bhattacharyya SC. Review of alternative methodologies for analysing off-grid electrification in Afghanistan.. 2017 [Online]. Available: https://energypedia.info
electricity supply. Renew Sustain Energy Rev Jan. 2012;16(1):677–94. /images/d/d6/A_GIS_approach_to_electrification_planning_in_Afghanistan.pdf.
[22] Moner-Girona M, Puig D, Mulugetta Y, Kougias I, AbdulRahman J, Szab� o S. Next [Accessed 16 August 2019].
generation interactive tool as a backbone for universal access to electricity. Wiley [51] KTH Royal Institute of Technology and SNV Netherlands Development
Interdiscip. Rev. Energy Environ. Nov. 2018;7(6):e305. Organisation. Electrification pathways to Benin.. 2018 [Online]. Available: http://
[23] Debnath KB, Mourshed M. Challenges and gaps for energy planning models in the www.onsset.org/electrification-pathways-for-benin.html. [Accessed 16 August
developing-world context. Nat. Energy Mar. 2018;3(3):172–84. 2019].
[24] Parreno Jr RP, Del Mundo R. Clustering algorithm as A planning support tool for [52] ‘OnSSET projects’, OnSSET projects. 27-Mar-2018 [Online]. Available: http://
rural electrification optimization. Int. J. Sci. Technol. Res. Dec. 2015;4(12): www.onsset.org/projects.html. [Accessed 27 March 2018].
112–20. [53] Pinedo Pascua I. Intigis: propuesta metodol� ogica para la evaluaci� on de alternativas
[25] Govender KM, Meyer AS, Dwolatzky B. The automatic clustering of uniformly de electrificaci�on rural basada en SIG. PhD thesis (in Spanish). Madrid: Universidad
distributed loads for the use in rural electrification planning. In: Transmission and Polit�ecnica de Madrid. School of Agricultural Engineering; 2012.
distribution conference and exposition, 2001 IEEE/PES, vol. 1; 2001. p. 450–5 [54] Monteiro C, Saraiva JT, Miranda V. Evaluation of electrification alternatives in
[Online]. Available: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumbe developing countries-the SOLARGIS tool. In: Electrotechnical conference, 1998.
r¼971276. [Accessed 18 December 2016]. MELECON 98., 9th mediterranean, vol. 2; 1998. p. 1037–41.
[26] Upadhyay S, Sharma MP. A review on configurations, control and sizing [55] Domínguez Bravo J, Pinedo Pascua I, Augusto Gonz� alez J. Herramientas SIG para
methodologies of hybrid energy systems. Renew Sustain Energy Rev Oct. 2014;38: la integraci� on de energías renovables en electrificaci�
on rural. IntiGIS., presented at
47–63. the XIII Congreso Nacional de Tecnologías de la Informaci� on Geogr� afica (in
[27] Lambert T, Gilman P, Lilienthal P. Micropower system modeling with HOMER. Spanish). Las Palmas de Gran Canaria; 2008.
Integr. Altern. Sources Energy 2006;1(15):379–418. [56] Amador J, Domínguez J. Application of geographical information systems to rural
[28] Distributed energy resources customer adoption model (DER-CAM) | building electrification with renewable energy sources. Renew Energy Oct. 2005;30(12):
microgrid. 2017 [Online]. Available: https://building-microgrid.lbl.gov/project 1897–912.
s/der-cam. [Accessed 19 February 2017]. [57] Domínguez J, Pinedo-Pascua I. GIS tool for rural electrification with renewable
[29] Georgilakis PS, Hatziargyriou ND. A review of power distribution planning in the energies in Latin America. Cancun, Mexico, 2009. In: International conference on
modern power systems era: models, methods and future research. Electr Power Syst advanced geographic information systems & web services; 2009. p. 171–6
Res Apr. 2015;121:89–100. [Online]. Available: http://ieeexplore.ieee.org/document/4782711/. [Accessed 15
[30] Mateo Domingo C, G� omez San Rom� an T, Sanchez-Miralles A,
� Peco Gonzalez JP, August 2019].
Candela Martinez A. A reference network model for large-scale distribution [58] Page Arias A. IntiGIS 2.0: objetivos y propuesta metodol� ogica. An� alisis de
planning with automatic street map generation. IEEE Trans Power Syst Feb. 2011; competitividad tecnol� ogica en Ghana., Master thesis (in Spanish). Madrid:
26(1):190–7. Universidad Complutense. Faculty of Geography and History; 2015.
[31] Ferrer-Martí L, Domenech B, García-Villoria A, Pastor R. A MILP model to design [59] Romero Otero L. Sistemas de informaci� on geogr�afica y electrificaci�
on rural.
hybrid wind–photovoltaic isolated rural electrification projects in developing An� alisis, desarrollo y estudio de caso con IntiGIS., Master thesis (in Spanish).
countries. Eur J Oper Res Apr. 2013;226(2):293–300. Madrid: Universidad Complutense. Faculty of Geography and History; 2016.
[32] Triad�o-Aymerich J, Ferrer-Martí L, García-Villoria A, Pastor R. MILP-based [60] Fronius Rainer, Gratton Marc. Rural electrification planning software (LAPER).
heuristics for the design of rural community electrification projects. Comput Oper Electricit�e de France - Research and Development; 2001.
Res Jul. 2016;71:90–9. [61] Soler R, Thomas F, Dhaiby N-E, Bakri M. Optimizing the place of PV systems in
[33] Zeyringer M, Pachauri S, Schmid E, Schmidt J, Worrell E, Morawetz UB. Analyzing rural electrification planning in Morocco. In: Photovoltaic energy conversion,
grid extension and stand-alone photovoltaic systems for the cost-effective 2003. Proceedings of 3rd World conference on, vol. 3; 2003. p. 2574–7 [Online].
electrification of Kenya. Energy Sustain. Dev. Apr. 2015;25:75–86. Available: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber¼1305116.
[34] Mirakyan A, De Guio R. Modelling and uncertainties in integrated energy planning. [Accessed 18 December 2016].
Renew Sustain Energy Rev Jun. 2015;46:62–9. [62] Syngellakis K, Ribeiro A, Arvidson A, Nordstr€ om M, Marcel J-C. ENABLE project
[35] DeCarolis J, et al. Formalizing best practice for energy system optimization summary report. Apr-2008 [Online]. Available: https://ec.europa.eu/energy/inte
modelling. Appl Energy May 2017;194:184–98. lligent/projects/sites/iee-projects/files/projects/documents/enable_final_report.
[36] Louie H, Dauenhauer P. Effects of load estimation error on small-scale off-grid pdf. [Accessed 28 April 2018].
photovoltaic system design, cost and reliability. Energy Sustain. Dev. Oct. 2016;34: [63] Moner-Girona M, B� odis K, Huld T, Kougias I, Szab� o S. Universal access to
30–43. electricity in Burkina Faso: scaling-up renewable energy technologies. Environ Res
[37] Fioriti D, et al. Stochastic sizing of isolated rural mini-grids, including effects of Lett Aug. 2016;11(8):084010.
fuel procurement and operational strategies. Electr Power Syst Res Jul. 2018;160: [64] Szab� o S, B�odis K, Huld T, Moner-Girona M. Sustainable energy planning:
419–28. leapfrogging the energy poverty gap in Africa. Renew Sustain Energy Rev Dec.
[38] Khatod DK, Pant V, Sharma J. Analytical approach for well-being assessment of 2013;28:500–9.
small autonomous power systems with solar and wind energy sources. IEEE Trans [65] Szab� o S, B�odis K, Huld T, Moner-Girona M. Energy solutions in rural Africa:
Energy Convers Jun. 2010;25(2):535–45. mapping electrification costs of distributed solar and diesel generation versus grid
[39] Sharma N, Varun, Siddhartha. Stochastic techniques used for optimization in solar extension. Environ Res Lett Jul. 2011;6(3):034002.
systems: a review. Renew Sustain Energy Rev Apr. 2012;16(3):1399–411. [66] Mahapatra S, Dasappa S. Rural electrification: optimising the choice between
[40] Gholizadeh-Roshanagh R, Najafi-Ravadanegh S, Hosseinian SH. Optimal robust decentralised renewable energy sources and grid extension. Energy Sustain. Dev.
integrated power distribution network planning under load demand uncertainty. Jun. 2012;16(2):146–54.
J. Power Technol. 2016;96(2):115–23.

15
P. Ciller and S. Lumbreras Renewable and Sustainable Energy Reviews 120 (2020) 109624

[67] Blechinger P, Cader C, Bertheau P. Least-Cost electrification modeling and [82] Abdul-Salam Y, Phimister E. The politico-economics of electricity planning in
planning–A case study for five Nigerian federal states. Proc IEEE Sep. 2019;107(9): developing countries: a case study of Ghana. Energy Policy Jan. 2016;88:299–309.
1923–40. [83] Abdul-Salam Y, Phimister E. How effective are heuristic solutions for electricity
[68] Bertheau P, Cader C, Blechinger P. Electrification modelling for Nigeria. Energy planning in developing countries. Socioecon. Plann. Sci. Sep. 2016;55:14–24.
Procedia Aug. 2016;93:108–12. [84] Howells MI, Alfstad T, Victor DG, Goldstein G, Remme U. A model of household
[69] Bertheau P, Oyewo A, Cader C, Breyer C, Blechinger P. Visualizing national energy services in a low-income rural African village. Energy Policy Sep. 2005;33
electrification scenarios for sub-saharan african countries. Energies Nov. 2017;10 (14):1833–51.
(11):1899. [85] Fuso Nerini F, Dargaville R, Howells M, Bazilian M. Estimating the cost of energy
[70] Network planner website. 2017 [Online]. Available: http://qsel.columbia.edu/net access: the case of the village of Suro Craic in Timor Leste. Energy Jan. 2015;79:
work-planner/. [Accessed 29 April 2018]. 385–97.
[71] Parshall L, Pillai D, Mohan S, Sanoh A, Modi V. National electricity planning in [86] Kocaman AS, Huh WT, Modi V. Initial layout of power distribution systems for
settings with low pre-existing grid coverage: development of a spatial model and rural electrification: a heuristic algorithm for multilevel network design. Appl
case study of Kenya. Energy Policy Jun. 2009;37(6):2395–410. Energy Aug. 2012;96:302–15.
[72] Ohiare S. Expanding electricity access to all in Nigeria: a spatial planning and cost [87] Hooke R, Jeeves TA. “Direct search” solution of numerical and statistical problems.
analysis. Energy Sustain. Soc. Dec. 2015;5(1):1–18. J ACM Apr. 1961;8(2):212–29.
[73] World Bank. Achieving universal access in the Kaduna electric service area. 2016 [88] Drouin C. Geospatial Cost Drivers in Computer-aided Electrification Planning: The
[Online]. Available: http://documents.worldbank.org/curated/en/78249148 Case of Rwanda. Master thesis. Massachusetts Institute of Technology. Department
7138851242/Achieving-universal-access-in-the-Kaduna-electric-service-area. of Mechanical engineering; 2018.
[Accessed 28 April 2018]. [89] Cotterman T. Enhanced Techniques to Plan Rural Electrical Networks Using the
[74] Modi V, Adkins E, Carbajal J, Shepa S. Liberia power sector capacity building and Reference Electrification Model. Master thesis. Massachusetts Institute of
energy master planning. Final Report. Phase 4: National Electrification Master Plan. Technology. Institute for Data, Systems, and Society.; 2017.
2013 [Online]. Available, http://sel.columbia.edu/assets/uploads/blog/2013/09/ [90] Ellman D. The Reference Electrification Model: A Computer Model for Planning
LiberiaEnergySectorReform_Phase4Report-Final_2013-08.pdf. [Accessed 18 Rural Electricity Access. Master thesis. Massachusetts Institute of Technology.
December 2016]. Engineering Systems Division.; 2015.
[75] Kemausuor F, Adkins E, Adu-Poku I, Brew-Hammond A, Modi V. Electrification [91] Mwalenga L, Amatya R, Gonz� alez-García A, Stoner R, P�erez-Arriaga JI.
planning using Network Planner tool: the case of Ghana. Energy Sustain. Dev. Apr. A Comprehensive Computer-Aided Planning Approach for Universal Energy
2014;19:92–101. Access. Case study of Kaloleni, Kilifi County, Kenya. Enel Foundation Working
[76] Banks DI, Mocke F, Jonck EC, Labuschagne E, Eberhard R. Electrification planning Papers. Enel Foundation; 2016 [Online]. Available: http://universalaccess.mit.
decision support tool. In: Domestic use of energy conference; 2000 [Online]. edu/#/publications [Accessed: 04-Apr-2018].
Available: http://citeseerx.ist.psu.edu/viewdoc/download?doi¼10.1.1.195.21 [92] REM. Test cases Colombia and Kenya. 2016 [Online]. Available: https://www.iit.
37&rep¼rep1&type¼pdf. [Accessed 19 January 2017]. comillas.edu/proyectos/mostrar_proyecto.php.es?nombre_abreviado¼ENEL%20
[77] Deichmann U, Meisner C, Murray S, Wheeler D. The economics of renewable Kenya-Colombia%20Test%20Cases. [Accessed 4 April 2018].
energy expansion in rural Sub-Saharan Africa. Energy Policy Jan. 2011;39(1): [93] Assistance to the government of Rwanda to develop a national electrification
215–27. strategy (NES) and related investment plans. 2017 [Online]. Available: http
[78] Huff DL. A probabilistic analysis of shopping center trade areas. Land Econ 1963; s://www.iit.comillas.edu/proyectos/mostrar_proyecto.php.es?nombre_abreviado
39(1):81–90. ¼Rwanda%20NES%2017. [Accessed 4 April 2018].
[79] GEOSIM projects. 2018 [Online]. Available: http://www.geosim.fr/index.php? [94] Promotion of mini-grids for rural electrification. 2017 [Online]. Available: htt
page¼references-en. [Accessed 29 April 2018]. ps://www.iit.comillas.edu/proyectos/mostrar_proyecto.php.es?nombre_abreviado
[80] Innovation Energie D� eveloppement. Capacity building for Off-grid rural ¼2017%20GIZ%20Uganda. [Accessed 4 April 2018].
electrification planning at the regional level in Ethiopia. 2007 [Online]. Available: [95] Gonzalez-Garcia A, Amatya R, Stoner R, Perez-Arriaga JI. Evaluation of universal
http://www.euei-pdf.org/en/seads/capacity-building/capacity-building-for-off-gri access to modern energy services in Peru. Case study of scenarios for Electricity
d-rural-electrification-planning. [Accessed 28 April 2018]. Access in Cajamarca. Enel Foundation Working Papers. Enel Foundation; 2016
[81] Innovation Energie D� eveloppement. CAP-REDEO results report. 2010 [Online]. [Online]. Available: http://universalaccess.mit.edu/#/publications.
Available: https://ec.europa.eu/energy/intelligent/projects/en/projects/cap
-redeo. [Accessed 28 April 2018].

16

You might also like