Energy Policy 120 (2018) 312–323

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

Decentralised electric power delivery for rural electrification in Pakistan T

a,⁎ b a a
Hassan Abbas Khan , Husnain Fateh Ahmad , Mashood Nasir , Muhammad Fatiq Nadeem ,
Nauman Ahmed Zaffara
a
Lahore University of Management Sciences (LUMS), Pakistan
b
Sewanee: The University of the South, United States

A R T I C LE I N FO A B S T R A C T

Keywords: The paper evaluates solar powered microgrids as a candidate solution for rural electrification in Pakistan where
Energy policy over 51 million people still live off-grid. Microgrids can significantly reduce the cost of providing basic elec-
Rural electrification trification and may also be scaled up to provide higher levels of services efficiently through energy and cost
Solar power sharing. A census of 6 off-grid villages highlights the demand for electrification. On average household report a
Decentralised generation
willingness to pay of USD 1.78 per month for the provision of high quality lighting and USD 3.24 per month for
the addition of a fan. Furthermore, households are willing to pay an additional USD 0.89 per month for a
communal load. First order cost calculations show that this demand can be met through the use of various solar
topologies, in particular decentralised microgrids by local entrepreneurs with suitable public sector subsidies.
Recommendations to modify the current legal framework are also presented, so as to provide an enabling en-
vironment for rural electrification through solar microgrids.

1. Introduction multiple households with electricity through a DC cable network. While

Access to electricity, even simply the provision of high quality
lighting alone, has been shown to increase productivity and provide
opportunities for economic development (Alstone et al., 2015;
Herington et al., 2017). According to the International Energy Agency
(IEA), more than 440 million inhabitants of developing Asia (China,
India, Pakistan and Bangladesh) and a further 580 million in Africa
have no access to any form of electricity (World Energy Outlook, 2017).
Most of those who live off-grid do not have a choice in this regard, and
must rely on unreliable and unhealthy alternatives, like kerosene oil
(Lam et al., 2012; Baul et al., 2018). The major source of electricity i.e.
the national grid is unviable for many of these isolated communities, as
the large upfront costs of electrification through the national grid
makes expansion prohibitively expensive for governments in devel-
oping countries (Thornburg et al., 2016; Ubilla et al., 2014). Therefore,
a paradigm shift towards powering these villages through low cost (and
consequently low-power) distributed renewable resources such as solar
photovoltaics (PV) has been seen in recent years (Ubilla et al., 2014;
Shenai et al., 2016; Madduri et al., 2016).
A recent innovation in the field of decentralised generation is the
solar direct current (DC) microgrid (Gandini and de Almeida, 2017;
Ramchandran et al., 2016; Nasir et al., 2018a). A microgrid is generally
built around a centralised solar generation mechanism that provides
it requires up-front setup costs, a microgrid allows the provision of basic
electrification (defined as high quality lighting and charging a mobile
phone), to multiple households in a single community at a significantly
lower long run cost compared to traditional power provision mechan-
isms. It is also a promising alternative to standalone solar systems and
fossil fuel generation, as it presents a low cost, sustainable and green
alternative.
Prominent practical implementations of microgrids include setups
in India and Africa (Mishra and Ray, 2014; Palit and Sarangi, 2014;
Palit et al., 2014). The most common commercial scale implementation
is the “Mera Gao Power (MGP)” project in India which provides 5 W of
DC electricity, enough to alternately power an LED light and a mobile-
phone charging point, for each subscribing household in a village for
about 8-h per day. MGP has reportedly connected over 100,000
households spread across 400 villages (Palit and Sarangi, 2014;
Urpelainen, 2016). In 2012, the government of Uttar Pradesh, India and
the Renewable Energy Development Agency, installed 1 kW DC mi-
crogrids in 11 districts covering around 4000 houses (Srivastava, 2013).
Other, recent successful deployments include those in Cameroon and
Papua New Guinea, that typically provide up to 10 W of power per
household for 8-h daily operation (Loomba et al., 2016). Other small
container based solar solutions on 12 V and 24 V are also being readily
utilized in Africa (Azimoh et al., 2016; Xu et al., 2016; Blyden and Lee,

⁎
Corresponding author.
E-mail address: hassan.khan@lums.edu.pk (H.A. Khan).

https://doi.org/10.1016/j.enpol.2018.05.054
Received 20 February 2018; Received in revised form 20 May 2018; Accepted 22 May 2018
0301-4215/ © 2018 Elsevier Ltd. All rights reserved.
H.A. Khan et al.
2006). However, none of these systems provide a 24/7 supply to rural
communities due to large costs required for generation and, in parti-
cular, storage requirements (Fossati et al., 2015).
Microgrids generally have been a successful solution for providing
basic electrification in off-grid communities, however we have not seen
an influx of these in Pakistan (Khan and Pervaiz, 2013). Pakistan is the
sixth largest nation in the world with around 51 million people (24% of
the population) living off grid with no access to electricity (World
Energy Outlook, 2017). Given Pakistan's well documented power crisis
and its historical similarities with India, which is itself at the forefront
of the microgrid movement (Urpelainen, 2016; Harish et al., 2014;
Chaurey and Kandpal, 2010), the lack of microgrids in the country is
quite surprising. In this paper, we establish that there is both a demand
for electrification in off-grid areas, and this demand can be met through
new decentralised solar microgrid architectures. By conducting a
census of multiple off-grid communities, we determine the demand for
rural electrification through survey-based elicitation. We find that not
only does demand exist for basic electrification; there is significant
demand for services beyond high quality light and mobile charging. We
combine these results with a detailed cost analysis of different solar
solutions for electrification to show that this demand can be met fea-
sibly through the use of solar microgrids. In particular, three different
architectures: standalone solar PV systems, conventional microgrids
and new decentralised microgrids (an innovative new architecture that
can meet this more sophisticated demand (Nasir et al., 2018a)) are
evaluated from their viability and long term levelized costs perspective.
We find that decentralised solar microgrids, provide an efficient solu-
tion for rural electrification.
Finally, we make policy recommendations based on our findings
and the current legal and economic conditions in Pakistan. In particular
we identify the need for the governments to provide an enabling en-
vironment for private parties to provide electrification through the use
of microgrids. This can be done by introducing more flexibility in the
law governing the generation and distribution of electricity in the
country, which is the largest hurdle in the adoption of microgrids in
Pakistan.
Given the interdisciplinary nature of the paper, for clarity of ex-
position, the remainder of the paper is organised as follows. We first
detail the methodology employed in both the demand and supply side
analysis. Results from both are then presented together, to allow for
easy comparison and to highlight the existence of unmet positive sur-
plus, i.e. the existence of demand and viable costs. Finally, in light of
our results, we conclude the paper with policy recommendations in
both the Pakistani and global context.
2. Methodology
We establish the feasibility of solar electrification (through decen-
tralised solar microgrids) by conducting both demand and supply side
analysis. Demand is established through willingness to pay elicitation,
while feasibility of supply is determined through detailed system cost
analysis.
2.1. Demand assessment methodology for rural electrification
To ascertain the demand for micro-grids, we elicited willingness to
pay by conducting a household census across 6 villages in two clusters
in the Multan district of Punjab. These areas are characterised by “bad
line coverage” due to their proximity to the Chenab River and its reg-
ular flooding, making a reliable connection to the national grid un-
feasible. Fig. 2.1 and its corresponding table lists the villages covered by
our census and provide their geographical location
The census instrument covered all 138 households in the area that
were designated as off-grid, i.e. they are not covered by the national
electricity grid. We elicited willingness to pay for different bundles of
electricity services, to gauge the demand for such services, while also
313
Energy Policy 120 (2018) 312–323
recording basic socio-economic characteristics. Participation in the
census was voluntary and respondents could choose to not answer any
or all questions.
2.1.1. Willingness to pay elicitation
The primary purpose of our census was to measure the demand of
electrification. We relied on survey-elicitation to determine willingness
to pay (WTP). Respondents were asked if they were willing to pay a pre-
selected price for a set of services. If respondents responded in the
negative, they were asked the maximum they were willing to pay.
Households were asked about three different levels of service, which
were chosen to replicate services provided by existing commercial low-
power microgrids, such as Mera Gao Power, and those made available
using decentralised microgrids architecture. The prices at which these
services were offered were randomised between three rate plans, which
presented increasing prices for each level of service. This was done to
control for any anchoring effects caused by the initial price offered. The
literature on WTP highlights the existence of strong reference depen-
dence, where the initial price offered acts as a “reference point” ef-
fecting the respondents underlying value for a good or service
(Johnston et al., 2017).
Due to the proximity of households inside each village, prices were
randomly allocated at the village level, instead of the household level.
Table 2.1 provides details of the level of services and their respective
prices per month under each price plan.
2.1.2. Socio-economic characteristics of households in our data
Table 2.2 provides summary statistics for the socio-economic in-
dicators of the households in our sample. As can be seen from the table,
households in our data are generally low income, with an average per
capita income of USD 26.2 per month, which is less than a dollar a day
per person.1 A possible caveat, as highlighted by past literature, is that
self-reported income is typically underreported (Debowicz et al., 2013;
Bank, 2007). To control for this, we follow the literature and requested
self-reported monthly expenditure as well. We find that the expenditure
on food and other general expenses, which are typically used as proxies
for income, closely match income. Similarly, instances for meat and
fruit consumption are also low, coming out to about once a week for
both on average.
In addition to self-reported data on household characteristics, we
collected information on the structure of the house. This was for two
reasons; first the development literature suggests that the structure of
the house is a good proxy for wealth. Second, decentralised microgrids
allow for some of the resources to be deployed at each house, and the
structure of the house is important from a deployment aspect. We found
that the vast majority of houses (92, or 66.2%) had temporary foun-
dations (mud-based houses), locally referred to as kacha houses
(loosely, ad hoc). The remaining houses (bar 1 pakka or permanent
house) were a mixture of modern building materials and temporary
foundations. These results further strengthen the low income and low
wealth results from Table 2.1. This result should however not be sur-
prising. House location, vis-à-vis grid electrification, should be con-
sidered a function of household income and wealth, as property rates in
localities with easy access to basic utilities would be higher.
2.2. Supply assessment methodology for rural electrification
To assess the feasibility of meeting the demand for rural elec-
trification, we analyse the long run levelized cost of multiple solar so-
lutions. We first present the details of three solar architectures of
1
Responses were collected in Pakistani Rupees (PKR). All PKR values used in the
analysis throughout the paper, have been converted to US Dollars using the exchange rate
of USD 1 to PKR 105. This was the average exchange rate during the census period.
Furthermore, the use of a single exchange rate throughout the analysis is an affine
transformation, and therefore does not affect any results.
H.A. Khan et al. Energy Policy 120 (2018) 312–323

Village Number of
Number2 Households
Cluster 1
1 29
2 22
3 40
Cluster 2
4 16
5 13
6 18

Fig. 2.1. Geographic location of villages in Multan district (South Punjab) along with the number of households in each village. Map image source: Google 2018.
Village names have been dropped to maintain anonymity.

Table 2.1 systems would require very high upfront costs (Goel and Sharma, 2017;
Services offered and their prices by plan. Ali and Shahnia, 2017).
Level of service Prices PKR (USD) Centralised microgrid implementations (Fig. 2.2b), on the other
hand, are more energy efficient than standalone systems due to re-
Price plan Price plan Price plan source sharing capabilities. Subscribers share resources that are both
1 2 3 generated and stored at a central facility. Some flexibility, gained
1 24/7 provision of high quality light (3 150 250 450 (4.29)
through diverse usage across households, yields smaller sizing of the
LED lights), and a mobile charging (1.43) (2.38) overall system with equivalent power delivery when compared to
point. standalone systems. For low-cost deployments, traditional microgrid
2 Services in Bundle 1 and a Fan. 300 500 900 (8.58) systems (e.g Mera Gao Power) rely on low voltage DC distribution.
(2.86) (4.76)
These largely self-sustained implementations typically allow for up to
3 Services in Bundle 2 and shared 450 750 1050 (10)
communal load (water pump). (4.29) (7.14) 8 h of electricity provisions per day (Ferris, 2014).
Scaling-up such a system to provide 24-h service is impractical due
to the higher costs associated with higher power generation and storage
interest, and then provide a unified cost assessment methodology. capacity. The low voltage distribution (12 V/24 V) typically used by
such systems limit both higher power load provision and larger grid
2.2.1. Solar PV architectures for rural electrification sizes. The primary deterrent to higher provisions and lack of scalability
To analyse the feasibility of rural electrification through solar, we are the considerable line losses at low voltages (Nasir et al., 2016;
distinguish between three distinct systems, namely: 1) standalone or Hamza et al., 2017). Furthermore, even if efficiencies are accounted for,
isolated solar home systems, 2) traditional low voltage microgrids the addition of new users on the microgrid is not as simple as increasing
(central generation e.g. Mera Gao Power), and 3) New distributed and power generation and storage. Mismatch in panel generation and sto-
decentralised microgrids (with decentralised resources i.e. decen- rage that is exasperated with age of the system, makes the system rigid
tralised generation and/or storage). in terms of operational expansion.
The general schematic diagram for standalone implementation is An alternative to traditional microgrids are decentralised microgrids
shown in Fig. 2.2a. Standalone systems are generally suboptimal for (Fig. 2.3) which can cater to many of the issues with traditional mi-
rural deployment due to generation and consumption profile of most crogrids (Nasir et al., 2016). Decentralised microgrid systems rely on
rural communities. Solar panels produce most power around noon, the distribution of resources in terms of generation as well as storage
whereas consumption is likely to be higher in the early mornings and where most of the power produced is consumed locally with any surplus
late evenings or nights. These systems have a high levelized cost of power shared between neighbours, therefore allowing for the possibi-
electricity due to limited storage and wastage due to non-matching of lity of powering a shared communal load. Such a system has the in-
generation and consumption. Higher autonomy and larger loads may be herent tendency of resource sharing to extract benefits of usage di-
met through the use of increased storage capacity, however such versity thereby lowering wastage and increasing efficiency. The extra

Table 2.2
Household Characteristics. Any person under the age of 15 is defined as a child.
Variable Mean Standard Deviation Number of observations Median 10th Percentile 90th Percentile

Household size 6.7 3.0 138 6.0 3.0 10.0

Number of children 3.0 2.1 137 3.0 0.0 6.0
Monthly income (USD) 140.6 131.6 137 95.2 47.6 285.7
Monthly income per capita (USD) 26.2 36.1 137 17.7 6.3 57.14
Monthly Expenditure on Food (USD) 90.5 68.3 138 76.2 38.1 142.9
Monthly expenditure on schooling (USD) 15.2 23.2 105 9.5 0 28.6
Weekly Meat consumption (Average number of times a week) 1.2 0.8 137 1 0 3
Fruit consumption (Average number of times a week) 1.4 0.9 138 1 9 4

314
H.A. Khan et al. Energy Policy 120 (2018) 312–323

Fig. 2.2. General Schematic Diagrams for traditional solar architectures used for rural electrification.

power from each household may also be aggregated to run a commu-
nity load, e.g. a water filtration plant, medical equipment of a local
hospital or computing load of a school.
Some decentralised microgrids (Madduri et al., 2016; Inam et al.,
2015) rely on communication among the distributed resources which
requires an extra sensing, monitoring and communication layer, which
increases the cost and complexity of the system. From a rural elec-
trification perspective, such a high cost system may not be economic-
ally viable. Some newer decentralised architectures having distributed
resources (generation as well as storage) allow higher resource utili-
zation along with the capability to aggregate individual resources for
powering a high power community load (Nasir et al., 2018a).
2.2.2. Cost estimation methodology for solar PV systems
Given the differences in system dynamics, in particular those per-
taining to power sharing and high load capabilities, it is not surprising
that all three systems would yield different costs. Here we describe the
general parametric cost estimation methodology, which can be im-
plemented for all three systems by selecting system appropriate para-
meters.
Consider a typical village with N households and allowable power
provision of Ph watts per household for T hours and a communal load of

Fig. 2.3. Schematic diagram of proposed decentralised solar microgrid.

315
H.A. Khan et al.
Pc watts for t hours. Then the total number of required energy units at
the output EO of system is given by
Eo = (NPh T )(1 − Ud ) + Pc t . (1)
where, Ud is the usage diversity factor that captures the inter-household
usage diversity in energy expenditure at any given time. Intuitively, it
captures the fact that the power demand of each household will be less
than or equal to allowable provision and will be different from other
households in the village. Stand-alone systems do not have the provi-
sion/capability of resource sharing at a village scale, that is to say that
if there is any extra production at the household level, it cannot be
shared and is wasted. Therefore, for standalone systems, Ud = 0 , by
construction. Alternatively, microgrids allow for power sharing and
usage diversity and so for any microgrid implementation it is typically
assumed that household consumer below their allowed provision and so
Ud∈(0,1) . For example, it is reasonable to assume from a design per-
spective that in the day time when sunlight is available, lighting load at
each household will be reduced. Similarly, in winters fan loads will be
reduced due to seasonal variation in temperature and associated
cooling requirements.
While there exists the possibility of simultaneous peak load demand
at each household, battery storage acts as buffer for providing these
instantaneous peak demands. In traditional centralized architectures,
since the battery resources are placed at a central location, power de-
livery can be adjusted from the central control point. In case of de-
centralised microgrid a decentralised control schematic is generally
employed for the coordinated operation of distributed storage resources
in the form of batteries at individual households to provide the buffer
for these instantaneous peak demands and to extract the benefit of
usage diversity at a village scale (Nasir et al., 2018b).
Given the energy requirement E0 of a system, system characteristics
such as solar PV panel sizing required to produce the requisite amount
of power must be calculated. These must consider the incident irra-
diance, effect of temperature degradation and various losses including
wiring losses, converter losses, and storage losses during charging/
discharging cycle losses.
To account for various energy losses and degradation in a system,
the amount of energy needed to be produced, Ep , is given by
E0
EP = . exploit the richness of our data to understand the determinants of
ηT ηB ηC ηD (2)
where, ηT , ηB , ηC and ηD are inefficiencies due to temperature, sto-
rage, convertors, and line losses, respectively. The resulting panel size
PPV needed per household can be expressed as
Ep
PPV = . We establish feasibility by estimating both the average willingness
EN (3)
where, E are the peak equivalent sun hours for the region. Similarly,
the first order cost calculation model for a battery system, the battery
energy capacity EB is determined by the total energy that battery must
supply when the sun is not available along with the extra energy that is
dissipated during charging/discharging. Moreover, to extend the bat-
tery life, generally there is a limit on minimum discharging state SOCmin
(%), which again tends to increase the required battery capacity. The
overall battery requirement for a microgrid then is given by
((24 − E ) Ph)(1 + SOCmin )
EB = . pay very close to those offered in Plan 1, though they are statistically
NηB (4)
In addition to storage and solar panel costs, there are other costs
which include the cost of converters, system protection equipment and
conductor (wiring) length. The price of converters is generally pro-
portional to their power processing requirements, loading levels and
current carrying capacity. Therefore, for simplicity, it can be taken as a
fixed percentage, λ c , of the total cost of PV panel. Similarly, the cost of
protection is proportional to the power loading level and short circuit
current capacity, therefore, in the current analysis it can be taken as a
316
Energy Policy 120 (2018) 312–323
fixed percentage as well (λp ). Finally, the cost of total conductor length
(l ) is given by C3 (USD/m).
Considering PV panel cost as C1 (USD/watt) and battery cost as C2
(USD/watt-hour), total system upfront cost is given by
CU = ((1 + λ C )(C1 PPV ) + (C2 EB ) + (lC3))*(1 + λP ). (5)
Generally, life time of a solar panel is 25 years, while the life time of
battery can be taken as NB years and life time of power electronic
converters is given by NC years. Therefore, for a typical 25 year system,
the operation and maintenance cost, along with the number of battery
and power converters replacement is calculated and added with the
capital cost to find the overall lifetime cost of the system. The total
lifetime cost CLT of the system is given by
25 25
CLT = CU + ⎛⎜ ⎛⎜ − 1⎞ ((C2 EB )) + ⎛
⎟ ⎜ − 1⎞ C1 PPV λ C ⎟⎞ (1 + λP ).
⎟
⎝ ⎝ NB ⎠ ⎝ Nc ⎠ ⎠ (6)
Further, it is important to evaluate the effective levelized cost of
electricity, LCOE (USD/kWh) of the system, given by Eq. (7), which
effectively calculates the cost of each unit produced by the microgrid
over its 25-year operation in comparison to its lifecycle costs.
CLT
LCOE =
EO (365)(0.001)(25) (7)
where, EO is the energy produced per day which is multiplied by 365
(days in a year) and 25 (operational lifetime of the system). 0.001 (in
the denominator) gives LCOE in price per kWh as kWh is standard unit
for electricity production/consumption.
3. Results and discussion
Given the data on reported willingness to pay and a clear metho-
dology for estimating the cost of meeting this demand, we now show
that there is surplus to be exploited. That is, the value respondents put
on these services (captured through willingness to pay), is higher than
the cost of providing it using solar systems. This surplus if met, would
not only by definition be welfare increasing, but could also be met by
local entrepreneurs, with minimal government interventions. After es-
tablishing the feasibility of solar solutions to rural electrification, we
willingness to pay.
3.1. Feasibility of decentralised solar microgrids as a solution for rural
electrification
to pay in our data, and by calculating the cost for providing services
using both tradition solar architectures and decentralised microgrids.
Table 3.1 reports prices respondents were willing to pay for services,
both overall and separated by price plan offered. As aforementioned,
they could either accept a price from a randomised price plan or reject
and report their own maximum. Note that these are by construction
truncated values for willingness to pay, as any price reported after re-
jecting the quoted price must be lower than the initial price offered by
the enumerator.
We find that those who rejected Price Plan 2, reported willingness to
lower for highest level of service. Finally, the reported willingness to
pay under Price Plan 3 suggests the effects of anchoring based on the
initial price quoted. All prices are statistically and economically sig-
nificant, suggesting a clear demand for services at all levels, regardless
of the price plan used by the enumerator.
Not only is demand for the level of services significant in our sample
villages, we also find that demand for higher levels of service is sig-
nificantly higher than for basic electrification. Table 3.2 reports the
absolute and relative difference in prices respondents reported they
H.A. Khan et al. Energy Policy 120 (2018) 312–323

Table 3.1 Table 3.3

Average reported willingness to pay in USD for each level of service. List of parameters and their corresponding values.
Average WTP (standard error) Parameter Standalone Centralised Decentralised
Microgrid Microgrid
Number of Lights Only Lights and Lights, fan and
HH Fan communal load Usage Diversity, Ud 0 0.3 (Nasir et al., 2018a)
PV panel price, C1 0.8 USD./Wp (Bao et al., 2017; Palm, 2018)
Overall 122a 1.78 3.24 4.14 Battery price, C2 0.1 USD./Wh (Lead Acid Battery) (Matteson and
(0.09) (0.16) (0.17) Williams, 2015)
Price plan 22 1.43 2.86 4.29 Distribution cost (village 0 450 USD (Nasir, 2018)
1b (0) (0) (0) scale with 40 houses
Price plan 2 56 1.46 2.80 3.62 10 m apart), C3
(0.07) (0.12) (0.12) Converter cost factor, λC 0.3 (Horowitz et al., 2017)
Price Plan 3 44 2.37 3.99 4.73 Protection cost factor, λP 0.05 (Horowitz et al., 2017)
(0.22) (0.39) (0.42) Peak sunshine hours, E 6 h/day (Shahzad et al., 2017; Zafar et al., 2017)
(average irradiance for
a
A village in the census refused to answer questions regarding service most of Southern Punjab,
through solar panels. The village was expecting a grid connection in the coming Sindh and Baluchistan
months and interpreted the solar service as an alternative. which includes major off-
b grid areas)
All respondents in Village 2 accepted the quoted prices under plan 1,
yielding a standard error of 0.

years. Therefore, for 6 years break-even plan, one battery replacement

Table 3.2
Relative and absolute difference in reported prices (USD) with respect to base
plan (lights only). Standard error in parenthesis. Absolute difference is the ar-
ithmetic difference in prices of two packages (Py – Px), while relative difference
is their ratio (Py/Px).
Number of Absolute
Observations difference
Lights and Fan vs. Lights 122 1.46
(0.11)
Lights, fan and 122 2.36
communal load vs. (0.11)
Lights
were willing to pay with respect to the base level of electrification
(lights only).
We find that there is significant difference for all services beyond
those provided by traditional microgrids (lights and mobile charging).
Both the absolute and relative differences are statistically and eco-
nomically significant. Households are willing to pay almost twice as
much for the addition of a fan, and about 2.5 times more for both a fan
and communal load.
Additionally, the marginal change between the two highest levels of
service is also found to be significant, with household willing to pay and
additional USD 0.89 per month (PKR. 94, which is about 1.5 times the
average daily expenditure on food per capita in our sample) for a shared
village water pump (communal load).
In summary, we can see that there is significant demand for rural
electrification. Even when considering the lowest estimates, we see that
on average respondents are willing to pay at least USD 1.43 (PKR 150)
per month for 24/7 basic electrification and USD 2.80 (PKR 294) and
USD 3.62 (PKR 380) per month with the addition of a fan, and a fan and
communal load, respectively.
We now use the proposed first-order cost model from the previous
section to estimate the cost of all three available solar systems.
Table 3.3 summarises the parameters of choice used in the estimation.
Parameters of interest are those that differ between systems. In
particular, when comparing standalone systems to microgrids, the
former cannot benefit from power sharing, captured by a usage di-
versity of zero, however it also does not have any cost associated with
distribution. Considering all these factors, LCOE, along with flat rate
tariff plan for the proposed and existing schemes of electrification of a
40-household village is calculated over 25 year project life. It should be
noted that various component of the system may need replacement
within the 25 years of system life. For instance, batteries would be
replaced every 4 years and power converters/charge controllers every 8
is considered, for 9 years break-even plan, one battery and one charge
controller replacement are considered.
Table 3.4 presents the estimated costs of all three topologies at
varying per household loads. It reports the upfront cost (capital cost)
and total cost over a system life of 25 years. Due to large up-front cost,
Relative
three possible scenarios of financing the upfront costs are incorporated;
Difference
payment upfront (no financing or 0% financing), financing through
1.97 current government subsidised small business loans, and financing at an
(0.67) average commercial small business loan rate (SME-Bank(Ltd), 2018;
2.57
National Bank of Pakistan, 2018).
(0.07)
Our calculations show that solar power, in particular decentralised
microgrids, present a viable alternative to grid electricity even for loads
beyond high quality light. Assuming a non-profit loan structure (0%), a
system can be made viable at nominal fixed monthly prices. Given the
average willingness to pay results from households in our data
(Table 3.1), the system can break even within 9 years for a system with
lights and fans.
Furthermore, recall that the decentralised microgrid is sized to
provide electrification well beyond a traditional setup. The decen-
tralised microgrid has the capacity to provide each household multiple
lights and a fan for up 24 h, where most traditional setups only provide
basic lighting for only 8 h a day.
Therefore, while decentralised microgrids may not be attractive to
entrepreneurs under the current cost structure, they offer a low-cost
system to governments and non-profits, who can with very low levels of
subsidies, make this an enticing proposition for local entrepreneurs.
Table 3.5 reports the years to break even for entrepreneurs based on
different levels of government subsidies, assuming monthly subscrip-
tion charges well below the average reported willingness to pay (PKR.
250 (USD 2.38) for lights and fan and PKR. 350 (USD 3.33) with a
communal load). It is important to note that the cost of electrification of
remote villages through national grids is typically very high due to the
large infrastructural cost required to set up poles, distribution trans-
formers and other costs.
A 75% subsidy, i.e. approx. 6800 USD (PKR. 750,000) per village is
orders of magnitude lower than the electrification costs through the
grid in most areas. Further, the need for the utility to add the requisite
generation is no longer there if microgrids based on solar PV are set up
in these regions. Therefore, investment in local microgrids, powered
through renewables which are operated and maintained locally should
be the way forward for many regions in South Asia and Africa.
Finally, it is important to highlight that the major component to the
cost of any off-grid solar PV system is the cost of storage. Batteries are
both expensive, have short life spans and are generally inefficient. As an
example, for a case of a decentralised microgrid, we show that the cost

317
H.A. Khan et al. Energy Policy 120 (2018) 312–323

Table 3.4
Estimated cost of competing solar generation implementations for a system of 40 households. Monthly charges calculated for paying back upfront cost financed
through no financing (or 0% financing), government scheme/loan-based financing (8% APY through the Prime Minister's Youth Small Business Loan, 2017) and
private financing (15% APY which is slightly higher than the average commercial rate of KIBOR + 6.5%, approx. 13–14% APY).
System Topology Load per house (24/7 provision to Capital Cost Capital + 25 years O Subscription Charges per user per month for payback in
subscribers) (USD.) &M Cost (USD.) (USD./Month)

3 years 6 years 9 years

APY APY APY
0%, 8%,15% 0%, 8%,15% 0%, 8%,15%

Tradition low voltage microgrida (e.g 1 light and mobile phone charging 2020 4550 1.4, 1.6, 1.7 0.8, 1.1,1.3 0.6,0.9,1.1
Mera Gao Power) unit (5 W).
Decentralized Microgrid 3 Lights, 9110 24900 6.3, 7.1, 7.8 4.2, 5.3, 6.3 3.1, 4.4, 5.5
1 fan, charging unit (30 W per
house)
3 Lights, 9525 25510 6.6,7.5, 8.2 4.5, 5.7, 6.7 3.3, 4.7, 5.9
1 fan, charging unit and Communal
load (30 W per house +500 W)
Standalone Production and 3 Lights, 10310 27905 7.1, 8, 8.8 4.9, 6.2, 7.3 3.3, 4.7, 5.9
Consumption (No grid) 1 fan, charging unit (30 Wper house)
3 Lights, 11100 29815 7.7, 8.7, 9.5 5.3, 6.7, 7.9 3.8, 5.3, 6.7
1 fan, charging unit and Communal
load (30W + 500 W)

a
Recall that due to low distribution voltage a traditional central microgrid (Mera Gao) cannot be scaled up for higher power provisioning.

Table 3.5 Table 3.6

Years to break even with 0% financing and subsidy on upfront costs. Multivariate analysis of determinants of willing to pay.
Level of Service Number of years to break even with Government (1) (2) (3)
subsidy on upfront cost Lights only Lights and Fan Lights, Fan and
Communal load
25% 50% 75% 100%
Monthly Household −0.00632*** −0.00953** −0.0129***
Lights and Fan 10 7 2 <1 Income (USD)
Upfront cost USD 9110 (0.00180) (0.00392) (0.00383)
Monthly subscription charges HH Income Squared 9.07e-06*** 1.70e-05*** 2.40e-05***
PKR 250 (USD 2.38) (2.94e-06) (5.73e-06) (5.73e-06)
Lights Fan and Communal 7 6 2 <1 Respondent Education 0.0668** 0.0972* 0.103*
Load (0.0298) (0.0571) (0.0604)
Upfront cost = USD 9525 Weekly Fruit −0.403** −0.618* −0.715*
Monthly subscription charges Consumption
PKR 350 (USD 3.33) (0.175) (0.333) (0.413)
HH in Cluster 2 −1.075** −1.736 −2.997**
(0.434) (1.565) (1.500)
Price Plan offered −0.641*** −1.165*** −1.985***
(0.171) (0.324) (0.350)
Price Plan offered 0.911*** 1.102*** 0.304
(0.207) (0.318) (0.334)
Household level controls Yes Yes Yes
Respondent level Yes Yes Yes
controls
Constant 3.713*** 6.037*** 9.039***
(0.707) (1.790) (1.764)
Observations 108 108 108
R-squared 0.504 0.369 0.429

Base is a household in Cluster 1, offered price plan 1. Robust standard errors in

parentheses.
*** p < 001.
Fig. 3.1. Life Time (25 years) Operation Cost Break-up of a Distributed ** p < 005.
Microgrid. * p < 0.1.

Table 3.7
of storage highly dominates the overall cost of the system in its lifespan
Average willingness to pay (USD) by cluster, and average difference across
of 25 years (Fig. 3.1), at current market prices. It dominates the overall clusters, with standard errors reported in parenthesis.
cost segment of the system. However, recent developments in battery
Number of Lights Only Lights Lights, fan and
technology suggest that the overall cost of such a system is likely to
HH and Fan communal load
come down in future (Schmidt et al., 2017), which will make these
microgrids (or other solar PV backup applications) even more viable. Cluster 1 29 2.31 4.09 5.36
(0.30) (0.54) (0.57)
Cluster 2 91 1.63 3.00 3.79
3.2. Note on the determinants of wiliness to pay
(0.07) (0.11) (0.10)
Difference (Cluster – 0.67 1.09 1.57
Given the richness of our data, we are also able to analyse the de- 1 – Cluster 2) (0.31) (0.56) (0.58)
terminants of demand for households in our data. Table 3.6 reports the

318
H.A. Khan et al.
results of our fully specified model, where we control for respondent,
household, and village level characteristics, as well as for anchoring
effects, by controlling for the price plan that was offered.2
The villages in our sample were both relatively small and geo-
graphically clustered together. This in effect, led to the division of the
sample into two clusters, with villages inside clusters separated from
each other by a couple of acres (Fig. 2.1). Our specification therefore
also controls for any cluster effects.
We find that for households in our data, the willingness to pay
varies with income, respondent education and by standard of living
indicators. Interestingly, we find that the willingness to pay is de-
creasing in monthly income regardless of the level of service offered,
though it decreases at a decreasing rate. Similarly, increases in the
consumption of fruit also lowers willingness to pay across the board.
Although these results might be counter intuitive; as we use fruit con-
sumption as a proxy for standard of living and wealth. However, this
may suggest that those with higher levels of income and wealth have
more readily available access to alternative sources of energy.3
We also find that respondents with more years of education were
more open to the use of solar technology, reflected in their higher
willingness to pay. This suggests the need for familiarity with new
technology, and the need for demonstrations and free trials as part of
any on the ground intervention. Similarly, we can confirm the existence
of anchoring by price plan, as households offered the highest price plan
(3), reported higher willingness to pay, both practically and statisti-
cally.
Finally, we find evidence of cluster effects, as villages in our base
cluster, on average, reported significantly higher willingness to pay
than those in cluster 2. Given the small size of village clusters that are
off-grid, this finding suggests that any intervention first account for
geographical variation. Table 3.7 presents the average willingness to
pay in each cluster and reports the difference. As can be seen, while the
willingness to pay in both clusters are practically and statistically sig-
nificant, so is the difference.
4. Conclusion and policy recommendations
The current study is part of a broader research agenda which seeks
to ascertain the demand for electrification in rural off-grid regions of
Pakistan, and design and implement systems that can meet these needs
in a cost-effective manner. Our findings indicate that decentralised
solar microgrids present a promising route to rural electrification,
especially in areas where grid expansion may be prohibitively ex-
pensive. They have the capacity and scalability to provide electricity
beyond those offered by tradition microgrids. In countries like Pakistan,
that are already facing major crises in supply, they present a low-cost
solution to not just the distribution problem, but also the problem of
generation. Given that the average willingness to pay for basic elec-
trification in our sample is around USD 1.78 (PKR 187) for basic elec-
trification and USD 3.24 (PKR 340) for the additional provision of a fan,
cost analysis suggest that the government must subsidize these systems
for better economic viability for widespread deployments.
The next steps in this agenda would be the roll out pilots of de-
centralised microgrids in off-grid areas, and to follow that up with a
randomised control trial across multiple villages to pin down the re-
vealed demand for electrification, and determine what characteristics
and interventions increase or decrease willingness to pay. Another
natural area for expansion would be to study the appropriateness of
microgrids as a source of back up electricity in on grid areas,
2 Furthermore, such entities should require a single licence from a
All results reported here are robust to changes in specification. Alternative specifi-
cations were not reproduced for the sake of readability. Interested readers may consult
the appendix.
3
Our instrument also asked respondents to list their alternative sources of energy,
however we found that nearly all respondents were unwilling to volunteer this in-
formation.
319
Energy Policy 120 (2018) 312–323
experiencing high number of rolling black outs.
4.1. Broader policy implications
Decentralised microgrids present a low cost and sustainable solution
to the problem of rural electrification. They also provide an opportunity
for local entrepreneurs and communities to lift themselves out of energy
poverty. The only hurdle is the relatively high upfront cost of such a
system, which requires very long periods of time before investors can
break even.
In such a situation, we recommend that governments play a more
active role in helping finance microgrids through possible subsidies and
tax breaks. Under the current operation schemes, there is no import tax
on solar panels, but a 16% sales tax is still applicable for energy pro-
duced in a commercial framework of Independent Power Producers
(IPP) or Energy Service Companies (ESCO). Further, helping small en-
trepreneurs and communities through subsidies and technical kno-
whow, or collaborating with larger players through public-private
partnership can yield a sustainable long term solution to the needs of
their populations, without the need to invest in large and significantly
more expensive grid based solutions.
4.2. Policy implications for Pakistan
We find that microgrids present a viable solution to the rural elec-
trification problem in Pakistan. According to our estimates of both
demand and cost, we find that even traditional microgrid setups would
fare well in Pakistan. The lack of pre-existing implementations may
then be attributed to the inflexible nature of laws governing electricity
generation and distribution in the country. A review of the current legal
framework shows that it is illegal for private parties to sell electricity to
other private agents.4 Exceptions to the law exist, but do not apply to a
typical microgrid setup. A private entity cannot set up a distribution
system for a rural community to sell electricity without a prior licence
for generation, as well as approval from the regional distribution
company. The system is in its current formulation too complicated for
local entrepreneurs or even larger entities to be able to operate mi-
crogrids at the community level. While net metering is now allowed
where individual consumers are only allowed to ‘sell back’ to the grid
but this is not allowed to individual consumers. Therefore, a major
recommendation is to allow private parties (microgrid owners) to sell
electricity to consumers in parallel with distribution companies. This
should be allowed by law and the requirement for a separate distribu-
tion licence should not be imposed to facilitate this process.
Considering our findings and the current policy environment in
Pakistan, we recommend that Pakistan provide an enabling environ-
ment to entrepreneurs and organisations that wish to provide rural
electrification through renewable energy. In the case of Pakistan, we
recommend that the law be altered to allow private parties to generate
and distribute electricity through renewable sources up to 100 kW in
off-grid and bad grid areas at a village level. We recommend the upper
bound of 100 kW to minimize potential legal challenges from dis-
tribution companies (DISCOs) in Pakistan which have the sole right to
distribute power in the current regulatory regime. The bound is esti-
mated as an upper limit to rural needs where DISCOs would not be
threatened in their operations in suburbs and cities. Since we are
talking about the potential law change, then necessary protection must
be given to existing entities while ensuring that new players (en-
trepreneurs who setup local microgrids) have a pathway to entry (this is
not possible under the current regulatory framework).
local authority (e.g the Union Council), instead of the multiple licences
required from multiple parties, at various levels of government. In the
4
See NEPRA ACT NO. XL OF 1997.
H.A. Khan et al. Energy Policy 120 (2018) 312–323

current setup, a microgrid operator in addition to a generation licence Acknowledgements

from the National Electric Power Regulatory Authority (NEPRA), may
also need approval from the regional distribution company, which
presents a case of conflict of interest, as the latter may view the mi-
crogrid as competition.
Microgrids provide a sustainable route out of electricity poverty in
the region. Therefore, governments should do all they can to enable
their implementation.
The authors will like to thank the International Growth Centre
(IGC), UK for their support of this work through Grant no. 37315. We
would also like to thank Sheikh Abdullah for research assistantship and
stakeholders at the Department of Energy, Government of Punjab for
insightful conversations and comments.

Appendix A

Robustness checks

In the paper we report results for only a single model, highlighting the important variables that correlate with willingness to pay. In this Appendix
A, we perform robustness checks by building to our model and also adding more variables to it, to show the robustness of our results to various
specifications.
We start with the base model from theory. It is assumed that WTP is a function of income (proxy for budget) and from the literature we control for
anchoring. In addition, we expect villages to differ based on geographical cluster, so we control for it as well. We then add controls for respondents
(2), Household characteristics (3) and both (4). As a final check, we incorporate non-linear effect of education (5)
Table A1 shows the results for willingness to pay for 24/7 provision of lights only, Tables A2 and A3 show results for lights and fan and lights, fan
and communal load respectively. In each table, specification 5 is the one which has been reported in the main text.
Finally, note that while not reproduced here, results from a Tobit regression are consistent with the ones found using OLS.

Table A1
Robustness checks for Lights only. Robust standard errors in parentheses.
WTP for Lights only (1) (2) (3) (4) (5)
Base Model Base Model and Base model and HH Reported Model (4) and Education squared.
Respondent Controls controls Base and all Controls

Monthly Household Income (USD) −0.00471* −0.00481** −0.00554** −0.00632*** −0.00574***

(0.00244) (0.00225) (0.00217) (0.00180) (0.00180)
Household Income Squared (USD^2) 5.30e-06 4.35e-06 6.76e-06** 9.07e-06*** 8.04e-06***
(3.45e-06) (3.08e-06) (3.12e-06) (2.94e-06) (3.05e-06)
Age 0.00349 −0.00475 −0.00353
(0.00513) (0.00566) (0.00581)
Respondent is head of Household −0.392** −0.285 −0.307*
(0.188) (0.184) (0.184)
Respondent years of Education 0.0612** 0.0668** 0.142**
(0.0288) (0.0298) (0.0596)
Education Squared −0.00872*
(0.00504)
Structure of the house is semi-pakka −0.205 −0.285 −0.330
(0.350) (0.397) (0.364)
Household size 0.0451 −0.0112 −0.0148
(0.0521) (0.0476) (0.0462)
Number of Children 0.0185 0.0293 0.0304
(0.0485) (0.0555) (0.0573)
HH members employed 0.146 0.211** 0.197**
(0.0888) (0.0865) (0.0821)
Average weekly consumption of meat −0.0723 −0.0633
(0.137) (0.142)
Average weekly consumption of fruit −0.403** −0.393**
(0.175) (0.185)
HH in cluster 2 (=1) −0.774** −0.591* −0.825** −1.075** −1.090***
(0.306) (0.355) (0.399) (0.434) (0.403)
Price plan 2 offered −0.367** −0.593*** −0.363* −0.641*** −0.663***
(0.175) (0.216) (0.188) (0.171) (0.170)
Price plan 3 offered 0.920*** 0.844*** 0.904*** 0.911*** 0.874***
(0.211) (0.193) (0.207) (0.207) (0.208)
2.703*** 2.641*** 2.311*** 3.713*** 3.648***
Constant (0.512) (0.750) (0.637) (0.707) (0.688)
Observations 114 110 113 108 108
R-squared 0.326 0.393 0.373 0.504 0.514

*** p < 0.01.

** p < 0.05.
* p < 0.1.

320
H.A. Khan et al. Energy Policy 120 (2018) 312–323

Table A2
Robustness checks for Lights and Fan. Robust standard errors in parentheses.
WTP for Lights and Fan (1) (2) (3) (4) (5)
Base Model Base Model and Base model and HH Reported Model (4) and Education squared.
Respondent Controls controls Base and all Controls

Monthly Household Income (USD) −0.00637 −0.00690 −0.00729* −0.00953** −0.00885**

(0.00500) (0.00488) (0.00438) (0.00392) (0.00405)
Household Income Squared (USD^2) 1.02e-05 8.97e-06 1.21e-05** 1.70e-05*** 1.58e-05**
(6.86e-06) (6.34e-06) (5.96e-06) (5.73e-06) (6.07e-06)
Age 0.00475 0.00141 0.00283
(0.00961) (0.00918) (0.00939)
Respondent is head of Household −0.320 −0.276 −0.302
(0.359) (0.376) (0.373)
Respondent years of Education 0.112** 0.0972* 0.185*
(0.0543) (0.0571) (0.108)
Education Squared −0.0102
(0.00955)
Structure of the house is semi-pakka 0.360 0.177 0.125
(1.456) (1.567) (1.537)
Household size 0.183* 0.0861 0.0819
(0.0993) (0.0967) (0.0954)
Number of Children −0.0578 −0.0136 −0.0123
(0.107) (0.115) (0.117)
HH members employed −0.286 −0.201 −0.217
(0.269) (0.295) (0.286)
Average weekly consumption of meat −0.119 −0.108
(0.292) (0.298)
Average weekly consumption of fruit −0.618* −0.607*
(0.333) (0.345)
HH in cluster 2 (=1) −1.595** −1.197* −1.283 −1.736 −1.753
(0.639) (0.718) (1.486) (1.565) (1.535)
Price plan 2 offered −0.690* −0.943** −0.874** −1.165*** −1.190***
(0.351) (0.416) (0.354) (0.324) (0.324)
Price plan 3 offered 0.903*** 0.839*** 0.956*** 1.102*** 1.059***
(0.311) (0.312) (0.303) (0.318) (0.319)
5.046*** 4.673*** 4.239** 6.037*** 5.961***
Constant (1.058) (1.388) (1.684) (1.790) (1.761)
Observations 114 110 113 108 108
R-squared 0.256 0.278 0.297 0.369 0.374

*** p < 0.01.

** p < 0.05.
* p < 0.1.

Table A3
Robustness checks for Lights and Fan. Robust standard errors in parentheses.
WTP for Lights and Fan (1) (2) (3) (4) (5)
Base Model Base Model and Base model and HH Reported Model (4) and Education squared.
Respondent Controls controls Base and all Controls

Monthly Household Income (USD) −0.00915* −0.00990** −0.0101** −0.0129*** −0.0119***

(0.00532) (0.00493) (0.00481) (0.00383) (0.00385)
Household Income Squared (USD^2) 1.61e-05** 1.52e-05** 1.80e-05*** 2.40e-05*** 2.22e-05***
(7.55e-06) (6.45e-06) (6.79e-06) (5.73e-06) (5.98e-06)
Age 0.00709 0.00312 0.00529
(0.0105) (0.0107) (0.0109)
Respondent is head of Household −0.436 −0.379 −0.418
(0.385) (0.396) (0.392)
Respondent years of Education 0.109* 0.103* 0.236**
(0.0568) (0.0604) (0.102)
Education Squared −0.0155*
(0.00908)
Structure of the house is semi-pakka −0.222 −0.415 −0.496
(1.355) (1.498) (1.442)
Household size 0.157 0.0403 0.0339
(0.0993) (0.0963) (0.0939)
Number of Children −0.0264 0.0378 0.0397
(0.105) (0.114) (0.116)
HH members employed −0.193 −0.106 −0.130
(0.247) (0.271) (0.258)
Average weekly consumption of meat −0.0713 −0.0553
(0.318) (0.325)
Average weekly consumption of fruit −0.715* −0.698
(0.413) (0.430)
HH in cluster 2 (=1) −2.222*** −1.888** −2.407* −2.997** −3.022**
(0.683) (0.777) (1.400) (1.500) (1.448)
(continued on next page)

321
H.A. Khan et al. Energy Policy 120 (2018) 312–323

Table A3 (continued)

WTP for Lights and Fan (1) (2) (3) (4) (5)
Base Model Base Model and Base model and HH Reported Model (4) and Education squared.
Respondent Controls controls Base and all Controls

Price plan 2 offered 1.535*** −1.818*** −1.653*** −1.985*** −2.023***

(0.369) (0.459) (0.393) (0.350) (0.350)
Price plan 3 offered 0.0861 0.0122 0.131 0.304 0.239
(0.322) (0.308) (0.313) (0.334) (0.332)
7.320*** 7.015*** 6.955*** 9.039*** 8.924***
Constant (1.127) (1.589) (1.704) (1.764) (1.717)
Observations 114 110 113 108 108
R-squared 0.312 0.336 0.343 0.429 0.439

*** p < 0.01.

** p < 0.05.
* p < 0.1.

References Resour. Econ. 4, 319–405.

Ali, L., Shahnia, F., 2017. Determination of an economically-suitable and sustainable
standalone power system for an off-grid town in Western Australia. Renew. Energy
106, 243–254 (2017/06/01/).
Alstone, P., Gershenson, D., Kammen, D.M., 2015. Decentralized energy systems for clean
electricity access. Nat. Clim. Change 5, 305.
Azimoh, C.L., Klintenberg, P., Wallin, F., Karlsson, B., Mbohwa, C., 2016. Electricity for
development: mini-grid solution for rural electrification in South Africa. Energy
Convers. Manag. 110, 268–277.
Bank, T., 2007. Agriculture for Development.
Bao, Q., Honda, T., El Ferik, S., Shaukat, M.M., Yang, M.C., 2017. Understanding the role
of visual appeal in consumer preference for residential solar panels (2017/12/01/).
Renew. Energy 113, 1569–1579 (2017/12/01/).
Baul, T.K., Datta, D., Alam, A., 2018. A comparative study on household level energy
consumption and related emissions from renewable (biomass) and non-renewable
energy sources in Bangladesh. Energy Policy 114, 598–608 (2018/03/01/).
Blyden, B.K., Lee, W.-.J. 2006. Modified microgrid concept for rural electrification in
Africa, in: Proceedings of 2006 IEEE Power Engineering Society General Meeting,
p. 5.
Chaurey, A., Kandpal, T.C., 2010. A techno-economic comparison of rural electrification
based on solar home systems and PV microgrids (2010/06/01/). Energy Policy 38,
3118–3129.
Debowicz, D., Dorosh, P., Haider, H.S., Robinson, S., 2013. A disaggregated and macro-
consistent social accounting matrix for Pakistan. J. Econ. Struct. 2, 4.
Ferris, D., 2014. Indian micro-grids aim to bring millions out of darkness. Appropr.
Technol. 41, 58.
Fossati, J.P., Galarza, A., Martín-Villate, A., Fontán, L., 2015. A method for optimal sizing
energy storage systems for microgrids. Renew. Energy 77, 539–549 (2015/05/01/).
Gandini, D., de Almeida, A.T., 2017. Direct current microgrids based on solar power
systems and storage optimization, as a tool for cost-effective rural electrification.
Renew. Energy 111, 275–283 (2017/10/01/).
Goel, S., Sharma, R., 2017. Performance evaluation of stand alone, grid connected and
hybrid renewable energy systems for rural application: a comparative review. Renew.
Sustain. Energy Rev. 78, 1378–1389 (2017/10/01/).
Hamza, M., Shehroz, M., Fazal, S., Nasir, M., Khan, H.A. 2017. Design and analysis of
solar PV based low-power low-voltage DC microgrid architectures for rural elec-
trification, In: Proceedings of IEEE Power & Energy Society General Meeting, 2017,
pp. 1–5.
Harish, S.M., Morgan, G.M., Subrahmanian, E., 2014. When does unreliable grid supply
become unacceptable policy? Costs of power supply and outages in rural India.
Energy Policy 68, 158–169 (2014/05/01/).
Herington, M.J., van de Fliert, E., Smart, S., Greig, C., Lant, P.A., 2017. Rural energy
planning remains out-of-step with contemporary paradigms of energy access and
development. Renew. Sustain. Energy Rev. 67, 1412–1419 (2017/01/01/).
Horowitz, K.A.W., Fu, R., Silverman, T., Woodhouse, M., Sun, X., Alam, M.A. 2017. An
analysis of the cost and performance of photovoltaic systems as a function of module
area, National Renewable Energy Lab. (NREL), Golden, CO (United States) NREL/TP-
6A20-67006 United States 10.2172/1351153 NREL English.
Inam, W., Strawser, D., Afridi, K.K., Ram, R.J., Perreault, D.J. 2015. Architecture and
system analysis of microgrids with peer-to-peer electricity sharing to create a mar-
ketplace which enables energy access, In: Proceedings of the 9th International
Conference on Power Electronics and ECCE Asia (ICPE-ECCE Asia), 2015, pp.
464–469.
Johnston, R.J., Boyle, K.J., Adamowicz, W., Bennett, J., Brouwer, R., Cameron, T.A.,
Hanemann, W.M., Hanley, N., Ryan, M., Scarpa, R., Tourangeau, R., Vossler, C.A.,
2017. Contemporary guidance for stated preference studies. J. Assoc. Environ.
Khan, H.A., Pervaiz, S., 2013. Technological review on solar PV in Pakistan: scope,
practices and recommendations for optimized system design. Renew. Sustain. Energy
Rev. 23, 147–154.
Lam, N.L., Smith, K.R., Gauthier, A., Bates, M.N., 2012. Kerosene: a review of household
uses and their hazards in low-and middle-income countries. J. Toxicol. Environ.
Health Part B 15, 396–432.
Loomba, P., Asgotraa, S., Podmore, R. 2016. DC solar microgrids – a successful tech-
nology for rural sustainable development, in: Proceedings of 2016 IEEE PES
PowerAfrica, pp. 204-x2208.
Madduri, P.A., Poon, J., Rosa, J., Podolsky, M., Brewer, E., Sanders, S.R., 2016. Scalable
DC microgrids for rural electrification in emerging regions. IEEE J. Emerg. Sel. Top.
Power Electron (vol. PP, pp. 1-1).
Matteson, S., Williams, E., 2015. Residual learning rates in lead-acid batteries: effects on
emerging technologies. Energy Policy 85, 71–79 (2015/10/01/).
Mishra, S., Ray, O. 2014. Advances in nanogrid technology and its integration into rural
electrification in India, In: Proceedings of Power Electronics Conference (IPEC-
Hiroshima 2014-ECCE-ASIA), 2014 International, pp. 2707–2713.
Nasir, M., 2018. Scalable DC Microgrid for Rural Electrification(Ph.D. thesis). Electrical
Engineering, Lahore University Of Managmement Sciences, LUMS.
Nasir, M., Zaffar, N.A., Khan, H.A. 2016. Analysis on central and distributed architectures
of solar powered DC microgrids, In: Proceedings of Power Systems Conference (PSC),
Clemson University, pp. 1–6.
Nasir, M., Khan, H.A., Hussain, A., Mateen, L., Zaffar, N.A., 2018a. Solar PV-based scal-
able dc microgrid for rural electrification in developing regions. IEEE Trans. Sustain.
Energy 9, 390–399.
Nasir, M., Jin, Z., Khan, H., Zaffar, N., Vasquez, J., Guerrero, J.M., 2018b. A decentralized
control architecture applied to DC nanogrid clusters for rural electrification in de-
veloping regions. IEEE Trans. Power Electron.
National Bank of Pakistan, 2018. Prime Minister’s Youth Business Load Available 〈http://
smebank.org/products/loan-facilities/tentative-lending-rates/〉.
Palit, D., Sarangi, G.K. 2014. Renewable energy based mini-grids for enhancing electricity
access: experiences and lessons from India, In: Proceedings of International
Conference and Utility Exhibition on Green Energy for Sustainable Development
(ICUE), 19–21 March, pp. 1–8.
Palit, D., Sarangi, G.K., Krithika, P., 2014. Energising Rural India Using Distributed
Generation: The Case of Solar Mini-Grids in Chhattisgarh State, India. Mini-Grids for
Rural Electrification of Developing Countries. Springer, pp. 313–342.
Palm, J., 2018. Household installation of solar panels – motives and barriers in a 10-year
perspective. Energy Policy 113, 1–8 (2018/02/01/).
Ramchandran, N., Pai, R., Parihar, A.K.S., 2016. Feasibility assessment of anchor-busi-
ness-community model for off-grid rural electrification in India. Renew. Energy 97,
197–209 (2016/11/01/).
Schmidt, O., Hawkes, A., Gambhir, A., Staffell, I., 2017. The future cost of electrical
energy storage based on experience rates. Nat. Energy 2, 17110.
Shahzad, M.K., Zahid, A., ur Rashid, T., Rehan, M.A., Ali, M., Ahmad, M., 2017. Techno-
economic feasibility analysis of a solar-biomass off grid system for the electrification
of remote rural areas in Pakistan using HOMER software. Renew. Energy 106,
264–273 (2017/06/01/).
Shenai, K., Jhunjhunwala, A., Kaur, P., 2016. Electrifying India: using solar dc microgrids.
IEEE Power Electron. Mag. 3, 42–48.
SME-Bank(Ltd), 2018. Lending Rates 〈http://smebank.org/products/loan-facilities/
tentative-lending-rates/〉.
Srivastava, A.K., 2013. Solar minigrids in rural areas of Uttar Pradesh. Akshay Urja 4,
16–17.
Thornburg, J., Ustun, T.S., Krogh, B., 2016. Smart microgrid operation simulator for
management and electrification planning, In: Proceedings of PowerAfrica, 2016 IEEE
PES, pp. 1–5.

322
H.A. Khan et al.
Ubilla, K., Jiménez-Estévez, G.A., Hernádez, R., Reyes-Chamorro, L., Irigoyen, C.H.,
Severino, B., Palma-Behnke, R., 2014. Smart microgrids as a solution for rural elec-
trification: ensuring long-term sustainability through cadastre and business models.
IEEE Trans. Sustain. Energy 5, 1310–1318.
Urpelainen, J., 2016. Energy poverty and perceptions of solar power in marginalized
communities: survey evidence from Uttar Pradesh, India. Renew. Energy 85,
534–539.
World Energy Outlook, 2017. (WEO, Electricity Access Database [Online]. Available:
323
Energy Policy 120 (2018) 312–323
〈https://www.iea.org/publications/freepublications/publication/
WEO2017SpecialReport_EnergyAccessOutlook.pdf〉.
Xu, Z., Nthontho, M., Chowdhury, S., 2016. Rural electrification implementation strate-
gies through microgrid approach in South African context. Int. J. Electr. Power
Energy Syst. 82, 452–465.
Zafar, J., Zafar, T., Zafar, K., Zafar, H., Gibson, A.A.P., 2017. Technical case for a mega-
scale solar energy installation for the national power grid of Pakistan. J. Renew.
Sustain. Energy 9, 043502.