ESMAP TECHNICAL PAPER
005
Field Performance Evaluation of Amorphous
Silicon (a-Si) Photovoltaic Systems in Kenya: Methods
and Measurements in Support of a Sustainable
Commercial Solar Energy Industry
Energy
Sector
Management
Assistance
Programme                                                August 2000
rnES ^ QAAA D                                     FILE COPY
Papers in the ESMAP Technical Series are discussion
documents, not final project reports. They are subject to the
same copyrights as other ESMAP publications.



JOINT UNDP I WORLD BANK
ENERGY SECTOR MANAGEMENT ASSISTANCE PROGRAMME (ESMAP)
PURPOSE
The Joint UNDP/World Bank Energy Sector Management Assistance Programme
(ESMAP) is a special global technical assistance program run as part of the World Bank's
Energy, Mining and Telecommunications Department.  ESMAP provides advice to
governments on sustainable energy development. Established with the support of UNDP
and bilateral official donors in 1983, it focuses on the role of energy in the development
process with the objective of contributing to poverty alleviation, improving living conditions
and preserving the environment in developing countries and transition economies.
ESMAP centers its interventions on three priority areas: sector reform and restructuring;
access to modern energy for the poorest; and promotion of sustainable energy practices.
GOVERNANCE AND OPERATIONS
ESMAP is governed by a Consultative Group (ESMAP CG) composed of representatives
of the UNDP and World Bank, other donors, and development experts from regions
benefiting from ESMAP's assistance. The ESMAP CG is chaired by a World Bank Vice
President, and advised by a Technical Advisory Group (TAG) of four independent energy
experts that reviews the Programme's strategic agenda, its work plan, and its
achievements. ESMAP relies on a cadre of engineers, energy planners, and economists
from the World Bank to conduct its activities under the guidance of the Manager of
ESMAP, responsible for administering the Programme.
FUNDING
ESMAP is a cooperative effort supported over the years by the World Bank, the UNDP
and other United Nations agencies, the European Union, the Organization of American
States (OAS), the Latin American Energy Organization (OLADE), and public and private
donors from countries including Australia, Belgium, Canada, Denmark, Germany, Finland,
France, Iceland, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Portugal,
Sweden, Switzerland, the United Kingdom, and the United States of America.
FURTHER INFORMATION
An up-to-date listing of completed ESMAP projects is appended to this report. For further
information, a copy of the ESMAP Annual Report, or copies of project reports, contact:
ESMAP
c/o Energy, Mining and Telecommunications Department
The World Bank
1818 H Street, NW
Washington, DC 20433
U.S.A.



Field Performance Evaluation of
Amorphous Silicon (a-Si) Photovoltaic Systems in Kenya:
Methods and Measurements in Support of a
Sustainable Commercial Solar Energy Industry
A PROJECT OF
Energy Alternatives Africa (EAA)
&
Renewable Appropriate Energy Laboratory (RAEL) and the
Energy and Resources Group (ERG), University of California, Berkeley
Richard D. Duke+' 1, Shannon Graham+, Mark Hankins*, Arne Jacobson+, Daniel M.
Kammen+, Bernard Osawa*, Simone Pulver+, and Erika Walther+
*Energy Alternatives Africa
P. 0. Box 76406, Nairobi, Kenya
Tel: +254-2-254-714623 or 716287 * Fax: +254-2-720909 * Email: energyaf@,iconnect.co.ke
+ Renewable and Appropriate Energy Laboratory (RAEL) &
Energy and Resources Group (ERG)
310 Barrows Hall, University of California, Berkeley CA 94720-3050 USA
Tel: +1-510-642-1139 * Fax: +1-510-642-1085 * Email: dkammen@socrates.berkeley.edu
* Internet: ht p:H!socrates.berkelev.edu/-rael/aSikenya.htmnl
Science, Technology and Environmental Studies (STEP) Programn
Woodrow Wilson School, Princeton University
Address Correspondence to the Team Leaders:
Mark Hankins and Daniel M. Kammen
Joint UNDP/World Bank Energy Sector Management Assistance Programme
(ESMAP)



Copyright ( 1999
The International Bank for Reconstruction
and Development/THE WORLD BANK
1818 H Street, N.W.
Washington, D.C. 20433, U.S.A.
All rights reserved
Manufactured in the United States of America
First printing August 2000
ESMAP Reports are published to communicate the results of the
ESMAP's work to the development community with the least possible
delay. The typescript of the paper therefore has not been prepared in
accordance with the procedures appropriate to formal documents.
Some sources cited in this paper may be informal documents that are
not readily available.
The findings, interpretations, and conclusions expressed in this
paper are entirely those of the author(s) and should not be attributed in
any manner to the World Bank, or its affiliated organizations, or to
members of its Board of Executive Directors or the countries they
represent. The World Bank does not guarantee the accuracy of the data
included in this publication and accepts no responsibility whatsoever
for any consequence of their use. The Boundaries, colors,
denominations, other information shown on any map in this volume do
not imply on the part of the World Bank Group any judgement on the
legal status of any territory or the endorsement or acceptance of such
boundaries.
The material in this publication is copyrighted. Requests for
pernission to reproduce portions of it should be sent to the ESMAP
Manager at the address shown in the copyright notice above. ESMAP
encourages dissemination of its work and will normally give
permnission promptly and, when the reproduction is for noncommnercial
purposes, without asking a fee.



Contents
Acknowledgements .................................................................                                                         iii
Executive Summary .................................................................1
Market Background ................................................................3
Quality Concerns .................................................................4
Project and Methodology Description ................................................................6
Sample Characterization .................................................................7
Module Testing Results ..................................................................9
Customer Knowledge ................................................................ 12
Outdoor Testing Facility ................................................................                                              13
System  Design ................................................................ 14
Balance of System  Equipment ................................................................                                          16
Economic Analysis ................................................................                                                     16
Implications for Market Development .................................................................                                  19
Policy Recommendations ................................................................. 20
Conclusions ................................................................ 21
References ................................................................                                                               23
Figures
Figure 1. Sales of Amorphous Silicon and Crystalline Silicon PV Modules in Kenya .................3
Figure 2. I-V Curves Showing Staebler-Wronski Degradation for a "Brand B2" 12
Wp a-Si PV   Module .................................................................5
Figure 3. I-V  Curve Performance for a "Brand A" 12 Wp a-Si Module .................                                       ......................6
Figure 4. I-V  Curve Performance for a "Brand C2" 14 Wp a-Si Module ......................................7
Figure 5. Field Data Collection Sites In Kenya ............................................8.....................8
Figure 6. Average Measured Power Output for Five Brands of a-Si Modules in Kenya ............. 10
Figure 7. Linear Regression of Module Output vs. Age for "Brand A" and "Brand B 1 "
a-Si Modules ................................................................                                              12
Figure 8. Combined Effects of Non-Ideal Tilts and Shading on Actual Solar Energy
Collected ................................................................                                                 15
Figure 9. Cost of Off-Grid Electricity Options as a Function of Household Discount Rate ....... 19
Tables
Table 1. Summary of Module Performance for Working a-Si Modules ........................................9
Table 2. Failure Rates for a-Si Modules from  Field Tests in Kenya ....................                                 ....................... 11
Table 3. Retail Price per Rated Wp for Small Amorphous Silicon and Crystalline
Photovoltaic Modules ................................................................                                      13
i






Acknowledgements
The authors would like to thank Daniel Kithokoi, David Khisa, and Frederick Ochieng for their
expert assistance, as well as all of the technicians including Henry Watitwa, Maina Mumbi,
Steven Kanyingi, Simon Wachira Kimani, Dikson Kisoa, and James Muita who provided field
data collection support. We also appreciate the cooperation and assistance provided by the
relevant a-Si PV module manufacturers and a number of solar related businesses in Kenya.
Perhaps most of all, we are indebted to the many Kenyan families who graciously allowed us into
their homes to make our data collection possible.
We are also grateful to many members of the international photovoltaics community: most
notably Phil Covell, Frans Nieuwenhout, and Frank van der Vleuten for providing insightful
review comments on previous drafts of this report; NREL for solar simulator testing services;
Adam Payne for various expert consultations; and Rick Mayberry of Performance Data Systems
for developing our custom testing equipment. We also express our gratitude to the Dexter
Environmental Trust for their generous funding, without which this project would not have been
possible. The US Environmental Protection Agency also provided important support to one of
the authors during the project through the Science to Achieve Results (STAR) Fellowship.
Finally, we dedicate this work to our good friend and colleague, David Khisa. He contributed
greatly to this project as both a fine professional and friend. May he rest in peace.
iii






Executive Summary
Kenya has an active market for photovoltaic (PV) solar home systems (SHSs), with cumulative
sales in excess of 100,000 units, and current sales of approximately 20,000 modules per year.
Small, 10 to 14 Watt single junction amorphous silicon (a-Si) modules dominate this market,
largely due to their lower retail price relative to similar sizes of crystalline PV modules.
Amorphous silicon modules sell for approximately US$ 5.00 per rated peak Watt (Wp) in Kenya,
while most brands of similarly sized crystalline modules sell for approximately US$ 8.00 per
rated Wp.
Despite this commercial success, there is substantial concern about the performance of single
junction thin-film a-Si-both because of the technology's uneven quality record and the
uncertainty introduced by short-term degradation which occurs when this type of PV module is
initially exposed to the sun. To address these concerns, in 1999 an interdisciplinary research
team from the Energy and Resources Group at the University of California at Berkeley, Princeton
University, and Energy Alternatives Africa from Nairobi, Kenya, conducted a study of the long-
term field performance of single junction a-Si PV modules in Kenya.
This study confirms that modules made by two of the three companies that dominate the Kenyan
a-Si PV market offer long-term performance roughly comparable to crystalline PV. Thus, quality
brands of single junction a-Si modules provide a highly cost-effective alternative to crystalline
modules for SHSs-especially for households only willing or able to purchase a relatively small
system.
There are, however, concerns about the susceptibility of available single junction a-Si modules to
physical breakage. More importantly, one manufacturer has built a one-third share of the Kenyan
a-Si market while selling modules with substantially inferior performance. The success of this
brand despite its considerably higher price per measured Wp suggests that rural consumers are
ill-equipped to compare the relative performnance of different module brands. Some combination
of public information, certification/labeling, standards (e.g. equipment quality requirements for
accessing PV loan funds or government restrictions on uncertified modules), and new SHS
business models (e.g. fee-for-service) may prove helpful in educating and protecting consumers,
thereby ensuring that this market reaches its potential. There are, however, risks and limitations
associated with each of these mechanisms that necessitate a cautious approach.
1






Field Performance Evaluation of Amorphous Silicon (a-Si)
Photovoltaic Systems in Kenya: Methods and Measurements in
Support of a Sustainable Commercial Solar Energy Industry
Market Background
Kenya has an active solar home systems (SHSs) market, with cumulative sales in excess of
100,000 units and current sales of about 20,000 systems per year-all without any substantial
subsidy and with only minimal programmatic support. There are more than 40 independent
import and manufacturing companies, as well as hundreds of vendors, installers, and after-sales
providers to serve this demand.
When the SHSs market emerged in the mid-1980s, typical systems used crystalline (x-Si)
modules of 40 Wp (Acker and Kammen, 1996). In 1989, however, 10 Wp amorphous silicon (a-
Si) modules entered the market, capturing the majority of SHSs sales within five years (van der
Plas and Hankins, 1998). Since then, total a-Si sales in Kenya have increased dramatically, from
10 kWp in 1989 to 270 kWp in 1998 (Figure 1).
Figure 1. Sales of Amorphous Silicon and Crystalline Silicon PV Modules in Kenya
300- -                                  ,--.--j----,--- -
250-   _         amorphous silicon PV
.t        ~~~~~~-)   crystalliiie PVm| 
0 0  1 00       I -                      -
0     0
, 200- t    2    t                         ;<r~~~-.04 
o                                 -~~~ -         .--- --
-,     50-                                           I -  _ _
1988     1990     1992      1994     1996      1998
Year
Data Source: Energy Alternatives AFRICA, 1999
Essentially all a-Si modules sold in Kenya go into SHSs, and a substantial majority of all
household systems use a-Si modules. That said, most a-Si SHSs in Kenya hardly qualify as
systems at all. In one typical pattern, families simply tie a module to their roof and directly wire
it to a car battery that they already own and charge in town. Among other system design issues,
the lack of a charge controller raises concerns about battery longevity; however, the critical
advantage of this approach is that it reduces the incremental cost of "going solar" to as little as
3



US$ 55. This has made PV generated electricity accessible on a fully commercial basis to
thousands of Kenyan households that otherwise would not have been able to afford it.
On a global basis, the a-Si market was initially dominated by small single-junction cells used for
calculators and other consumer applications. Annual production of these cells by Japanese firms
has hovered in the vicinity of 5 MWp since 1987 (Nikkei Electronics, 1/25/99, No. 735). This
market is generally considered distinct because producing very low output consumer cells is far
easier than achieving uniform deposition of the active a-Si layers over substrate areas large
enough to build conventional PV modules.
Single-junction a-Si modules are produced from a number of factories located in France, Croatia,
China and the United Kingdom-all using production technology originally derived from the
Chronar Corporation of the United States. Precise annual production figures for module-scale
single junction a-Si are not available, but the total output of single-junction a-Si modules in 1999
probably did not exceed 5 MWp. In addition to Kenya, important markets for single-junction a-
Si modules used in SHSs include China, Indonesia and Zimbabwe, among others.
As of 1997, cumulative production of multi-junction a-Si for modules had only reached about 3
MWp (Peterson, 1997); however, in 1997 BP Solarex (10 MWp dual junction) and United Solar
Systems Corporation (5 MWp triple junction) both started production from a-Si plants. I Both of
these manufacturers have announced major expansion plans, and Kaneka Solar Tech started
production from a 20 MWp a-Si facility in late 1999, with plans to expand to 40 MWp of
production capacity in 2000.2
These production figures suggest total a-Si production for 2000 of approximately 30 to 45 MWp
(including consumer cells, single junction modules and multi-junction modules), or roughly one-
fifth of global PV sales. If the new multi-junction a-Si production capacity comes on line as
planned, a-Si may maintain or possibly increase its market share.
Quality Concerns
Despite their commercial success in Kenya there is substantial concern about the performance of
single junction thin-film amorphous silicon (a-Si) PV modules. New a-Si panels suffer
predictable light-induced efficiency degradation before stabilizing (Figure 2).  Although
manufacturers are supposed to rate panels at their post-degradation output levels, this "Staebler-
Wronski" degradation (Staebler and Wronski, 1977) introduces performance measurement
complexity that may add to the reputation problems of a-Si. Additionally, single-junction a-Si
manufacturers have generally not been willing to match the warranty terms offered for crystalline
panels, fueling doubts about the relative reliability of this PV technology.
Most of the a-Si modules sold in Kenya do include some sort of manufacturers' warranty:
* Free Energy Europe (FEE) offers a 10 year warranty;
• NAPS offered a 5 year warranty before their module factory was purchased by FEE;
* Koncar offers a 5 year warranty;
1  www.nrel.qov/ncpv/pdfs/tf united.pdf and www.nrel.ov/ncpv/pdfs/tf solarex.Ddf
2  w%w. nrel. sov/ncpv/pdfS/npcinc. pdf
4



•  Intersolar offers a 5 year warranty; and,
* Uni-solar offers a 20 year warranty.
APS has no stated warranty in their literature. For those manufacturers that do offer a warranty,
the terms include the usual protection against manufacturing defects and a drop in output below
85% or 90% of rated power depending upon the manufacturers specific termns.
The importing agents for all of the major a-Si brands sold in Kenya have historically been
cooperative about replacing defective modules. Replacements have run less than 1% for all but
the worst performing brand in our study (for which the importer reports returns of about 10% of
modules sold). There are, however, serious concerns about the practical value of these
warranties for rural Kenyan SHS owners.
First, rural households are unable to measure the output of modules, and the dealers are generally
not well-equipped to help them and may prefer to discourage such testing to avoid having to
process returns. Battery failure is the first sign of possible panel failure; however, batteries
regularly fail even when used with a module that is functioning well. It is therefore hard for a
rural household to know whether their module is performing within the warranty terms. As a
result, only absolutely "dead" modules tend to come back for warranty replacement.
In addition, if a company is sold or goes bankrupt the Kenya representatives for the old brand
may choose not to honor the defunct manufacturer's warranty. It is important to note, however,
that Free Energy Europe purchased the NAPS a-Si manufacturing facilities in 1998 and has
stated that they will honor warranties on modules sold by NAPS before the change in ownership.
Figure 2. I-V Curves Showing Staebler-Wronski Degradation for a "Brand B2" 12 Wp a-
Si Photovoltaic Module
*Pmax -I5. 
1.2                              o Pmax  11.2 W
:   {                ~~~~~~x Pmax= 10.6 W.
0.8 
0.4
c              *    Cum. Solar Energy = 0 kWh / rn
0.2     0    Cum. Solar Energy = 62 kWh I n    _
x    Cum. Solar Energy = 404 kWh  rd
0                      10-
o (         51                    1'5        20O        25Z
Voltage (volts)
5



Project and Methodology Description
During the Spring of 1999, graduate students and staff from Energy Alternatives Africa (Nairobi,
Kenya) the Renewable and Appropriate Energy Laboratory (RAEL) of the University of
California Berkeley and the Science, Technology and Environmental Policy (STEP) program at
Princeton University conducted a field and laboratory-based testing program with the primary
purpose of assessing the performance and longevity of single-junction a-Si modules typically
sold for SHSs in Kenya. In addition to module testing, we also interviewed the owners of each of
the 145 SHSs in our sample and conducted a range of tests to assess overall system performance.
Finally, we established a pair of outdoor test sites (in Nairobi and Berkeley) to closely track the
initial Staebler-Wronski degradation (as well as any subsequent degradation over the time frame
of our testing) in a set of new panels.
We developed a testing methodology able to estimate module output normalized to standard test
conditions (25 degrees Celsius and insolation of 1000 W/m2). This involves a customized low-
cost I-V tester that records current (I) and voltage (V) pairs as the resistance is varied using a
potentiometer. The resulting I-V curve (e.g. Figures 2, 3 & 4) can be used to estimate module
output at the maximum power point, or any other voltage, by multiplying voltage times the
associated current.
Figure 3. I-V Curve Performance for a "Brand A" 12 Wp a-Si Module
Isc = 0.92 amps
1- 8         ffV__             _  Voc= 21.9 volts
P   = 9.9W
_____  _____ I    l              Fill Factor = 0.49
0.8- .1! P  = 9.5W (power @12.5 volts)
-'   --Typical Operating Curren       Crtyp    4torer po2ynolts)
Curve Fit 4th order ponomial
0.6-
0.4                          Typical    
C;                     ~~~~~~~~~Operating  
0.2- __                     Voltage_                      _
0          5         10         15         20         25
Voltage (volts)
To account for module temperature variations, we measured the temperature on the back of the
module with a Type-K thermocouple. We used an inexpensive but rugged and accurate Licor
200-SA pyranometer to measure solar radiation on the module surface. Finally, to standardize
the test procedure, we designed a simple test rack that ensured that both the module and
pyranometer were pointed directly at the sun during the test.
6



The overall accuracy for this approach is approximately ±5%. This estimate of the accuracy is
based on a calibration of our test method against a solar simulator at the United States
Department of Energy's National Renewable Energy Laboratory (NREL). The repeatability of
the test for clear sky conditions is ±5%. This means that, under clear sky conditions, 95% of
repeated tests of the same module will yield maximum power estimates within ±5% of the
average maximum power for that module. This methodology is generally not sufficient for
testing individual modules for compliance with specified power ratings, though it can be used to
this end if the module is severely under performing. It is, however, well suited to our primary
objectives of assessing long-term degradation rates and comparing the average performance of
different brands, models and vintages of a-Si modules under field conditions. See Jacobson, et
al., (2000) for more information about the I-V test methodology.
Figure 4. I-V Curve Performance for a "Brand C2" 14 Wp a-Si Module
Isc =0.90 amps
1                      -I-----, .Voc = 22.9 volts
,    l        ~~~~P   = 7.2 W
0.8  0                  Fill Factor = 0.35
0.8                 _.-l_.__ P  = 7.0 W (power @ 12.5 volts)
XA                  '                    tiCurve Fit = 4th order polynomial
W   0.6  -- Typical Operating Current-,
0.4  -            _   _                     _
fi                   j ~~~~~~~~~Typlical       
rDg                             Operatingi    
0.2 -            oVotage --
0
0          5         1 0         1 5       20         25
Voltage (volts)
Sample Characterization
Figure 5 shows the geographic distribution of our field data collection sites. Seventeen of the
145 homes we visited had two modules, and two had three, yielding a total of 164 PV modules.
Of these solar panels, approximately 85% were 10 to 14 Wp a-Si modules and the other 15%
were crystalline panels ranging in size from 10 to 51 Wp. We intentionally sought out a-Si
systems-testing crystalline systems only as convenient or if we encountered particularly old
modules. Accordingly, the average age of the amorphous modules in our sample was 2.7 years
vs. 3.7 years for crystalline modules.
7



Figure 5. Field Data Collection Sites In Kenya
/                                                             /
K                   ~~~~~~~~~Eastern
Rift Valley
*  North-Eastern
Wesern.
I Western                       ,/MOUNT KENYA
Nyanza
Nyanza            Central
-  .4*bi Area
Nairobi                                  Coast
N                           !
N                               X               ,-Mombasa
* Field Sites  70    0    70 Kilometers
There are three principal manufacturers that have dominated the a-Si market in Kenya:
Nest& Advanced Power Systems (NAPS) of France marketed two models in Kenya (an 11
Wp panel with model #A13R, and a 12 Wp panel introduced in 1997 with model #11601),
and the factory was recently purchased by Free Energy Europe (www.free-energy.net);
Koncar of Croatia which has marketed their 12 Wp module (model # KT1-3A) in Kenya
since 1990 (www.koncar-solar.tel.hr); and,
8



*  Intersolar of Wales which has marketed two models, the 11 Wp Phoenix (model # B107W)
and the 14 Wp Phoenix Gold (B108D) (www.intersolar.com).
As noted above, all three of these are derivative of the single-junction a-Si technology originally
commercialized by Chronar Corporation of New Jersey, USA during the late 1970s and 1980s.
In reporting results we do not list PV module manufacturers by name. Rather, we refer to each
brand using a letter, and number different models in chronological order. The three module types
currently sold on a large scale in Kenya are, in no particular order, Brand A, Brand B2 and Brand
C2. Of the 143 amorphous panels in our sample, Brand B2 and Brand A each accounted for 24%
of the total. Brand B 1 accounted for another 22%, followed by Brand C1, Brand C2 and Brand
E, all with shares of 10% or lower.
Module Testing Results
We found substantial variation in the average quality of different module brands (Table 1; see
also Figure 6 for a graphic comparison of module performance) with Brand B2 panels
perforning best, Brand A panels performing a close second, and Brand Cl and Brand C2
modules trailing substantially. Including panels from our outdoor testing facility, but excluding
cracked or failed modules (i.e. those producing less than 10% of rated capacity), the average
Brand B2 module in our sample produced 89% of rated output. Moreover, we are 95% confident
that the average performance for Brand B2 modules in Kenya is between 86%  and 92%  of their
rated output. Brand Bl modules performed similarly with a mean output of 88% of rated power
and a 95% confidence interval ranging from 85% to 91% of rated output.
Table 1. Summary of Module Performance for Working a-Si Modules3
Rated       Average                         95%          Average   #Modules
Module Model    Max.          Measured      Percentage    Confidence        Age of       Tested
Power    Max. Power        of Rated        Interval      Modules
(Watts)      (Watts)        Output    (L % points)4    (years)
BrandA               12           10.0           83%           ±3 %            2.8          31
BrandBI              11           9.7            88%           ±3%             3.1          31
Brand B2             12           10.6           89%           ±3%             0.9          32
Brand Cl             I 1          6.8           61%            ±14%            2.4          5
Brand C2             14           7.7           55%            ±9%             1.5          12
Brand D              25          22.5           90%             n/a            5.0          1
Brand E              10           7.2           72%      -      11%            5.9          4 4
3  In addition to the 130 modules tested in the field, the information in Table 1 includes the results from modules tested at the
University of Califomia, Berkeley and at Energy Altemative Africa's offices in Nairobi. The additional modules tested
include 3 brand A modules, 2 brand B 1 modules, 3 brand B2 modules, and 6 brand C2 modules. These statistics all exclude
failed modules, defined as those producing less than 10% of rated capacity. Cracked modules and modules performing at
pre-stabilized power output levels are also excluded.
4   The 95% confidence interval around the percentage of rated output is given in percentage points. The interval spans the
range that is plus or minus two standard errors from the average power times the "Student's t" statistic value for the number
of tests in the sample. The value is divided by the average power to get a range in percentage points. This informnation tells
us, for example, that we can be 95% confident that the mean output for brand B2 modules is between 86% and 92% of their
rated output.
9



The Brand Bl and B2 modules compare favorably with the 17 crystalline (x-Si) modules of
various vintages and brands that we tested in the field (see Figure 6). On average, none of the
modules in our sample performed at their rated output levels, however, this is not unexpected. A
number of researchers have reported results from field performance tests indicating that
crystalline and amorphous PV modules often perform 5-15% below their rated power output
(Hester and Hoff, 1985; Jennings, 1987, Lehman and Chamberlin, 1987, Chamberlin, et al.,
1995).
Brand A modules produced, on average, 83% of their 12 Wp rating, and we are 95% confident
that the average performance of Brand A modules in Kenya is between 80% and 86% of their
rated power. While the quantitative difference in performance between Brand B2 and Brand A is
modest, we have large enough samples of each to be 95% confident that there is a real difference
in the average performance of these two module brands. That is, it is unlikely that the reported
performance differential is just the result of an unusual sample of modules.
Figure 6. Average Measured Power Output for Five Brands of a-Si Modules in Kenya;
Performance of Crystalline (x-Si) Modules is Included for Comparison
100 -1 -,   I  I   I-T    I     I'T1' I   I 'I-  I  I-   I T T I  II
60
~ 0
Wt~~ ~~~~~~~~~ 1S
0
Brand  Brand  Brand  Brand  Brand   x-Si
A     Bi    B2    C1        C2  (various)
Photovoltaic Module Brand
Brand Cl and Brand C2 products performed notably worse than the others. The older 11 Wp
Brand Cl modules averaged 61% of rated output with a 95% confidence interval ranging from
47% to 75% with the relatively wide range resulting from the relatively small number of Brand
Cl modules in our sample. Their current 14 Wp Brand C2 modules produce only 55% of rated
output on average, with a 95% confidence interval ranging from 46% to 64%.
At least some brands of a-Si modules also appear to suffer from high levels of failure due to
encapsulation problems. We also found some modules that had failed due to cracked glass
plates. Defining module failure as producing less than 10% of rated power, 54% of Brand Cl
and 40% of Brand C2 modules in our sample had failed vs. only 6% of Brand A modules and 0%
10



of Brand BI or B2 modules (Table 2). These modules were excluded from our mean
performance estimates because our sample may have missed some failed modules that had been
removed from customers' rooftops. Nonetheless, accounting for failed modules would widen the
performance gap substantially if the true failure rates for each brand are similar to what we
encountered.
Table 2. Failure Rates for a-Si Modules from Field Tests in Kenya5
Module Model           Failed Modules (%)6      Cracked Modules (%)7            # Modules
Encountered
Brand A                        6%                         3%                       31
Brand B 1                      0%                         6%                       32
Brand B2                       0%                         6%                       32
Brand C1                       46%                       29%                        13
Brand C2                       40%                        0%                        10l
Brand D                        0%                        50%                        2
Brand E                        38%                       20%                        8
Other (unknown)                50%                       0%                         2
These failure and cracking rates are for our data set only. They may underestimate failure and cracking
rates for a-Si modules in Kenya, as people may discard failed units.
It should be noted that over the past decade all three of the companies have made modifications
aimed at improving module quality. Moreover, a senior company representative from the
manufacturer for Brand Cl and Brand C2 has reported that he is aware of quality problems and
expects improvements to the manufacturing process instituted during 1999 and 2000 to
substantially ameliorate the problem. Preliminary testing of the newly released brand C2
modules by RAEL does suggest improved performance; however, more testing is needed in order
to confirm these results.
In addition to comparing performance across different brands we also considered module
degradation as a function of age. Figure 7 is consistent with degradation of 1% per year for
Brand A and Brand BI modules.8 It is possible, however, that incremental improvements in
manufacturing techniques explain this difference in performance for the differently aged
modules.9
Testing over an eight year period by PVUSA indicated a 1-5% per year degradation rate for
arrays of both a-Si and crystalline modules.10 Our analysis is therefore broadly consistent with
5   This table includes only those modules (130) that we encountered in the field.
6   Failed modules are defined as modules that have an output that is less than 10% of the rated output.
7   This category includes only those cracked modules that were still operational. Cracked modules that had failed are listed as
failed modules. Note that the percentage listed is the fraction of the totalfunctioning modules that are cracked.
8   Brand B2 modules were not included because they have not been available long enough to show a clear long-tern
degradation trend. Other brands were not included due to small sample sizes.
9   Such an improvement would be consistent with learning curve theory for manufacturing processes as discussed in Argote
and Epple (1990), Neij (1997), and Duke and Kammen (1999).
10 The Photovoltaics for Utility Scale Applications (PVUSA) research facility is located in Davis, Califomia, USA. Note,
however, that array degradation may be due in part to increased voltage losses in the interconnections and other array level
effects rather than the degradation of individual PV modules. PVUSA (1998) and Townsend, et al. (1998).
11



PVUSA's data, tending to confirm their result that a-Si and crystalline panels have similar long-
term degradation rates.
Figure 7. Linear Regression of Module Output vs. Age for "Brand A" and "Brand BI"
a-Si Modules (consistent with a 1% per year decrease in performance)
i 1.2
0
4)       ______E                  ___ __
¢;8-    0.8                  -------
-u--  Brand "A" (12 Wp PV modules)
0.4      --*--Brand "B1" (11 Wp PV modules) -I_   __
2  0.2   _-       power = 0.87 - 0.013*Age
- -----power = 0.94 - 0.013*Age
gi  0      2          4           6          8          10
PV Module Age (years)
Customer Knowledge
Despite the stark quality variations described above, our survey indicates that only about 9% of
respondents thought they knew the brand of the module they owned, and 15% of these
respondents answered incorrectly. In addition, over 40% of respondents would not even hazard a
guess about how long their panel would last, less than 3% of respondents knew whether they had
an amorphous or crystalline solar module, and only about 6% of respondents had an opinion
about whether amorphous or crystalline panels last longer.
Customers who do not even know the brand of their panel, let alone having any specific
knowledge of that model or manufacturer's track record, cannot distinguish high quality modules
from inferior goods. However, there may be other proxies for brand names such as country of
origin or distinctive visual aspects (Hong and Wyer, 1990). Our survey would not have detected
these possible quality signals, and our results may therefore overstate the level of ignorance
among rural Kenyan purchasers of a-Si modules. Additionally, the highest performing a-Si
module in our study, brand B2, enjoys a market share that is approximately twice that of its
nearest competitor. This may suggest that at least some rural customers are differentiating
between brands.
12



Nonetheless, C2 modules appear to have made modest gains in market share during 1999 and the
first half of 2000 despite their substantially higher average cost per effective Wp.11 Table 3
shows that, at about $US 5.00 per rated Wp, C2 modules are among the lowest priced modules
sold in Kenya; however, adjusting for their lower average performance, C2 modules are easily
the most expensive a-Si module in the market (and this holds even without accounting for the
much higher breakage and failure rate observed in our sample for this type of module). This
suggests that many customers are ignorant of the relative quality of the different module types-a
conjecture that is supported by our survey data indicating minimal knowledge of brand and type
even for users' own modules.
Table 3. Retail Price per Rated Wp for Small Amorphous Silicon and
Crystalline Photovoltaic Modules
i Panel                      Panel Typ             $/rated Wp     $/measured Wp
Brand B2                         a-Si       12        4.50             5.00
Brand A                          a-Si       12        5.00             6.00
Brand C2                         a-Si       14        5.00             9.00
Polycrystalline                  x-Si       20        8.00             9.00
Finally, with regard to security concerns, it is interesting to note that only 15 percent of
respondents indicated that they personally knew someone whose solar panel had been stolen.
This suggests that module theft may not be a major problem in the regions from which we drew
our sample-perhaps in part because of the considerable care taken by many owners to protect
their modules.
Outdoor Testing Facility
As noted above, we also tested 9 new a-Si modules at Energy Alternatives Africa's outdoor
testing facility in Nairobi, Kenya and 5 new a-Si modules at the Renewable and Appropriate
Energy Laboratory (RAEL) of the University of California, Berkeley. Figure 2 shows the
Staebler-Wronski degradation for a Brand B2 module. Initially, the I-V curve indicates a fairly
high fill factor but the curve quickly flattens, with the majority of the degradation occurring in
just 14 days (the 62 kWh/mi of cumulative solar exposure corresponds to 14 days for this
measurement)'2.
As expected, this testing has confirmed average Staebler-Wronski degradation of 25%-30%, with
most of the losses occurring during the first month. These results vary considerably by
manufacturer, however, with Brand C2 modules suffering losses as high as 42%. The final
stabilized power output for the three brands tested at our outdoor facility are similar to the results
collected during our field tests.
Sales figures based on personal communications of the authors with the relevant distributors in July and December of 1999
as well as February of 2000.
12 Fill factor is defined as the ratio of the actual power output to the product of the short circuit current and the open circuit
voltage. Numerically, it is calculated by Pmax + (Isc * Voc).
13



System Design
In addition to testing module performance we also gathered considerable data pertinent to system
configuration. The factors considered include module tilt, orientation, and shading; battery size;
wire size; use of battery terminals; and the role of charge controllers.
Regarding module tilt, an analysis using seasonal insolation data and standard solar incidence
equations indicates that, throughout Kenya, mounting modules horizontally generally maximizes
both annual energy production as well as output in the worst month of the year.'3 Based on the
observed tilts for the SHSs in our sample (and assuming effectively random orientation
determined by rooflines) over 95% of the panels in our sample are positioned such that they lose
no more than 10% of total possible annual solar energy. Similarly, shading accounts for a less
than 10% loss in annual solar energy for over 85% of the panels in our study. Finally, we
estimate that 65% of the panels in our sample lose less than 10% of available solar energy due to
the combined effects of non-ideal tilt, orientation, and shading (Figure 8). Note that this analysis
assumes that there are no seasonal patterns in daily cloud cover (e.g. always cloudy in the
morning, but sunny in the afternoon). This assumption may result in small errors in our analysis.
With regard to balance of system components, our data indicate that batteries are generally
oversized. Most 12 Wp modules in our sample used 50 Ah batteries. Substituting a 20 Ah
battery would save the owners about $US 20, or 12% of initial system cost, and the systems
would likely perform better. At an 80% depth of discharge, a 50 Ah battery represents over 10
days of storage for a 12 Wp module. Given that most users' demand chronically outstrips the
supply from these small modules, they are likely to drain a 50 Ah battery to an extremely low
state of charge and then keep it there indefinitely.
In contrast, with a 20 Ah battery, if the owners use only half of their daily energy production for a
little over a week they could fully recharge from an 80% discharge state. In practice, 20 Ah
batteries are likely to remain in a chronically low state of charge as well, but there is more
likelihood that they will periodically reach higher charge states-and, most importantly, they are
less expensive to purchase and replace.'4
Fewer than one in ten of the systems in our sample included a charge controller, though most of
those with a controller did have a low voltage disconnect (75%) and a battery voltage indicator
(92%). A low voltage disconnect, if properly set and not by-passed, protects the battery from
overly deep discharge, which is almost certainly the most common cause of shortened battery life
for typical Kenyan SHSs. It is, however, essential to consider the parasitic load from charge
regulators (which can consume as much as one third of the output from a 12 Wp module) as well
as the relative benefits of foregoing a regulator and applying the savings to the cost of a second
panel.15
13 It is, however, advisable to mount modules with a slight tilt (e.g. 5 to 10 degrees) to ensure that water efficiently drips off
the panel.
14 It is also important to point out that a 12 Wp module provides a maximum output of only about one amp. The risk of
overcharging is therefore minimal even for a 20 Ah battery in a system that lacks a charge regulator.
15 van der Plas and Hankins, 1998.
14



Figure 8. Combined Effects of Non-Ideal Tilts and Shading on Actual Solar Energy
Collected
70 5 4-! -1
80                        ,2L IVr!i  Tj-l 
co  60    - Combined losses due to
50    . non-ideal tilts and shading   j  I    T  IIIIi
o (north facing orientations)  i7                     11111
ot 40          jT           - -        i-1---   "               I
eC 30                       I         Z              vy .
10                           .   .LA  ......L 
0-50  55   60   65   70   75   80   85   90   95  100
actual/ideal (%)
annual average daily solar energy
80   r.,tj... 
70                                                  + 
t 60          Combined losses due to           I
a)    l  non-ideal tilts and shading              Ii'
0  50   >    (east-west facing orientations)   | +-        t
4- 40I 
(D30  ---iL
0-50  55   60   65   70   75   80   85   90   95  100
actual/ideal (%)
annual average daily solar energy
Actual annual average daily solar energy collected at each panel is expressed as a percentage of the total possible
annual solar energy available at the site. North facing orientations present the worst case scenario and east-west
facing orientations are the best case.
15



Voltage losses due to wire sizing never exceeded 2.5% for any of the systems in our sample;
however, only 34% had battery terminals, with the remainder using alligator style clips or simply
wrapping exposed wires around their battery terminals. We also found that all of the systems
that used battery terminals had tight connections vs. only 20% of the batteries without terminals,
and 36% of the batteries had corrosion on or near the wiring connections. These are important
correctable design and maintenance issues because significant voltage losses can occur from
loose or corroded connections.
Balance of System Equipment
Batteries made by Associated Battery Manufacturers (ABM) made up the largest share of our
survey sample with 71% of the batteries identified. ABM batteries trade under various names
including Chloride Exide, Thomas White, and Dagenite. Automotive and Industrial Battery
Manufacturers (AIBM) made up the next largest share of our sample, with 18% of the batteries
identified, all from their Jua Tosha and Voltmaster lines.
Low electrolyte levels were generally not a major problem for batteries tested in our survey,
probably because the low charging currents from the 10 to 14 Wp a-Si solar modules relative to
the average size of the batteries will rarely result in out-gassing of the batteries' electrolyte.
It is possible to make a rough estimate of state of charge (SOC) by measuring the specific gravity
of the electrolyte in the cells. These data indicate that 39% of the batteries in the sample had a
very low state of charge, 21% had a medium state of charge, and 35% had a high state of
charge.'6
Our data on battery ages and battery replacement intervals suggest that many families that use 10-
14 Wp solar home systems will replace their batteries every one to two years. This replacement
interval may be longer than the useful battery life since some families may not be able to afford
to replace their battery even though it is no longer performing well.
There also appears to be a high level of familiarity with basic lead-acid battery maintenance
among users.  There is, however, no clear correlation between average battery life and
respondents' answers to various questions about their maintenance habits.
Finally, about two-thirds of respondents indicated that they had not yet replaced their original
fluorescent tubes and, on average, respondents who had replaced their bulbs did so only once
every 3.5 years.
Economic Analysis
Based in part on our measurements of actual module output, this section compares the lifecycle
cost per kWh for the three principal off-grid electricity options available to rural Kenyan
16 These specific gravity data may overestimate the state of charge since, in some of the cases, high specific gravity readings
were likely due to the addition of acid by users unaware that they should only add distilled water to battery cells. On the
other hand, the batteries in our sample rarely receive equalization charges so electrolyte stratification is likely to be common
and specific gravity readings taken near the top of battery cells will tend to yield incorrectly low state of charge estimates. It
is not clear which of these potential sources of errors is likely to dominate, so it is unclear whether our results overstate or
understate the average battery state of charge.
16



households: battery charging without any panel, battery charging using a crystalline panel, and
battery charging using an a-Si panel.
For the base case analysis we make the following assumptions:
*   20 year system lifetime;
*   2 year average battery life;
*   20 year crystalline module life and 10 year a-Si module life;
*   US$ 1.00 cost per battery charge;17
*   panel owners do not bring their batteries to charging stations for supplemental charges;
*   system efficiency losses of about 20%  for panel based systems and 0%  for battery-only
systems;
*   the labor cost associated with installing panel based systems is US$ 12.00 on the assumption
that it requires one day of a skilled technician's time while battery-only systems are self-
installed at no cost;
*   wiring costs US$ 5.00 for panel based systems, half that for battery-only systems, and it
must be replaced after 10 years; and,
users spend US$ 0.60 annually on distilled water for their batteries.
For the battery-only case we make the further assumptions that users take their batteries for 40
charges per year and are able to obtain a charge worth 80% of the batteries capacity each time.
Prices for all system components are based on informal surveys of distributors conducted by the
authors during July of 2000, and delivered module amperage is calculated based on our
measurements for each module type, assuming an average charging voltage of 12.5 volts. For
panel-based systems we use a 20 Ah "Jua Tosha" battery while the battery charging systems use
50 Ah "Chloride Exide" batteries. Both are locally manufactured batteries partially optimized for
deep discharge cycles in rural household applications. The former is more than twice as
expensive on a per Ah basis, but nonetheless better suited for 12 Wp SHSs because of its 25%
lower unit cost and more appropriate size.
We further assume a 12 Wp NAPS module for the a-Si case, but estimate the amperage and cost
of a 12 Wp crystalline module by simply scaling down the price and output of a 20 Wp
crystalline Solarex module, which is one of the most commonly available small crystalline
panels.
Estimating discount rates for rural Kenyan households is an inexact science and we therefore
consider the cost per kWh of each possible system option as a function of a wide range of
discount rates. Assuming a discount rate of 25% for our base case, the per kWh life-cycle cost of
delivered electricity from an a-Si SHS is about US$ 0.73 versus US$ 0.91 for a crystalline SHS
17 This breaks down as US$ 0.60 in charging fees; $US 0.30 in transport and $US 0.14 for the cost of time and inconvenience.
The latter component is equivalent to one hour of foregone wages using 2000 annual work hours and an average rural
income of SUS 289 in 1995 from Nyang (1999).
17



and US$ 0.92 for the battery charging only case.18   These results are sensitive to parameter
choices.
Figure 9 illustrates the importance of the assumed discount rate. Similarly, if we assume that
SHS owners use 50 Ah batteries with their 12 Wp panels, as many in fact do, then the associated
per kWh costs at the 25% discount rate increase to $US 0.88 for a-Si and $US 1.06 for
crystalline, and battery charging emerges as the cheapest option for discount rates any higher than
30%.
The assumed frequency of battery charging is also an important parameter choice. A family that
uses a 50 Ahr battery to provide electricity for three hours of television and lighting will
completely drain their battery approximately once per week. For our base case, however, we
assume only 40 charges per year because financial constraints may induce some families to
conserve on battery usage and capital constraints and inconvenience often prevent families from
maintaining their batteries in a useable state at all times. If we increase the charging frequency to
52 charges per year, the cost of the no-panel battery charging option drops below the crystalline
cost for discount rates exceeding 19%, and it beats the a-Si option for discount rates over 34%.
Similarly, if we eliminate the estimated per-charge inconvenience and time factor, then the cost
of the battery-only option beats crystalline for discount rates exceeding 22%, and a-Si for
discount rates in excess of 40%. Changing both parameters to favor the battery case allows the
no-panel case to beat a-Si for discount rates exceeding 28%.
Note, however, that our base case is generous to battery charging in assuming that customers
drain their battery to the 20% charge level and obtain a full charge every cycle.19 In fact,
consistent deep discharges reduce a battery's ability to accept a full charge and battery stations
may not always give their customers the fullest possible charge. There is, moreover, a substantial
hassle factor involved with lugging a battery to and from the charging station on a regular basis
and our estimates only imperfectly capture this cost.
On the other hand, the base case results are quite robust to some parameter choices. For
example, we assume that a-Si single junction modules last only 10 years in our base case in order
to account for the fact that they are typically more vulnerable to breakage given their more fragile
framing and encapsulation design. This is consistent with an annual breakage rate of about 8%.
If annual a-Si module breakage rates are assumed to be 0% then the cost per kWh for an a-Si
system drops by only about four cents for a discount rate of 10% and only one cent for discounts
rates in excess of 25%.
Similarly, changing the average battery lifetime does not notably affect the relative costs of the
different systems-though assuming a shorter battery life raises the cost of all three options
considerably. For example, a one year battery life implies that the a-Si option costs US$ 1.10 per
1 8 The crystalline case is more expensive, particularly for lower discount rates, because of the higher per Wp cost of these
modules. Moreover, our analysis is generous to crystalline in that we assume that 12 Wp crystalline modules would sell for
only 60% of the cost of a 20 Wp module. In fact, the per Wp cost of crystalline modules typically increases significantly as
they are scaled down below 20 Wp due to higher circuitry and framing costs as well as the need to cut individual crystalline
cells in half.
19 This concern also negatively affects the performance of low wattage SHSs since our data suggest that many SHS batteries
are chronically in a low state of charge; however, we partially account for this by assuming a 20% charging inefficiency for
the SHS case.
18



kWh for a discount rate of 25% (as opposed to US$ 0.78 in the base case). It is important to note
that battery life is a function of system type and design; however, there are insufficient data
available to assess the relative average battery lifetimes for the different cases we consider.
Figure 9. Cost of Off-Grid Electricity Options as a Function of Household Discount Rate
$4.00
$3.50                     battery charging with an a-Si panel
$3.00              -8-  battery charging with a crystalline pane
$2 50    8,    |  P  ~battery charging without a panel
$2.50..............
C $2.00                  ___ _
$1.50             ++                      - 1
$1.00
$0.50                 -
$0.00
0        10        20        30        40        50
Household Discount Rate (%)
Implications for Market Development
This study presents evidence that the best quality a-Si modules available in Kenya provide
excellent long-term service of comparable reliability to crystalline modules. There is, however,
evidence of dramatic and largely undetected quality variation across brands.
This has two critical implications for development of the Kenyan market for SHSs. First,
thousands of rural Kenyan households unfortunate enough to select the seriously under-
performing a-Si brand have lost most or all of their investment in what is often the most
expensive durable good they own. This is a substantial economic and social cost in its own right.
Second, widespread consumer ignorance about the relative quality of different brands, may
discourage potential SHS customers. This occurs both because consumers face uncertainty about
the performance of the particular module brand they buy and because they may base their
performance expectations on a pooled quality estimate roughly derived from the average
performance of all the installed modules (including both high- and low-performing brands) in
their region (Duke et al, 2000).
This is also one plausible factor helping to explain why so many rural Kenyan households
continue to rely on battery-only systems despite the relatively high life-cycle cost of this
approach. Approximately 300,000 rural Kenyan households have lead-acid batteries that they
regularly bring to charging stations, while more than 100,000 additional Kenyan families have
19



opted to use a solar panel to charge their batteries (Hankins et al, 1997). As shown in Figure 9,
battery charging is not competitive with SHS on a life-cycle basis except for very high real
discount rates.
The prevalence of battery-only systems despite the relatively high cost of this approach is also
consistent with the fact that rural Kenyans households typically have very little savings and little
opportunity to borrow at non-usurious rates (i.e. high household discount rates). Despite this,
even at a monthly fee of just US$ 3.00, only about 38 percent of the SHS owners we interviewed
responded that they would have preferred to rent rather than buy the system they currently own.
If these figures reflect true demand, then fee-for-service is unlikely to emerge as a major SHS
delivery mode in Kenya unless substantial subsidies become available.
Policy Recommendations
As detailed in a forthcoming manuscript by the authors (Duke et al, 2000), there are a range of
potential mechanisms available to address the quality variation and user ignorance problems
outlined above. These include testing and certification, associated quality seals of approval or
importation restrictions, user education and more centralized SHS delivery modalities.
The Iternational Electrotechnical Commission (IEC) issued an international standard (EEC 1646)
for thin-film PV modules in 1996; however, at present none of the a-Si modules with a
substantial market share in Kenya comply with this standard.20 The PV Global Approval
Program (PVGAP, www.pvgap.org), an organization supported by various multilaterals and PV
industry groups, has a stated priority of focusing on quality problems in developing countries. In
particular, they are working to:
•  ensure that available IEC standards are universally recognized;
* develop "recommended standards" in cases where the IEC has yet to act; and,
* develop a PVGAP quality "mark" for PV components and a PVGAP quality "seal" for
whole systems.
PVGAP expects a quality mark for modules to come into use shortly; however, it is unclear how
quickly the a-Si module manufacturers that sell into the Kenyan market will seek or receive
approval under this program. Moreover, absent considerable user education, a "seal of approval"
may have little impact on the Kenyan market. This sort of public education campaign may prove
difficult and expensive given the dispersed and remote nature of the target audience.
Another possibility is that the Kenyan government could ban the sale of modules that do not
meet quality standards. This approach would, however, risk excluding relatively high quality and
uniquely affordable small a-Si modules from the Kenyan market place until and unless their
manufacturers are able to obtain certification. It could also create opportunities for excessive or
corrupt government interference in the SHS market place.
20 Uni-solar modules are IEC1646 compliant, however, they are considerably more expensive than other a-Si modules and
have a trivial market share. Free Energy Europe expects to achieve IEC1646 compliance for its modules by the end of
2000.
20



One alternative to government mandated standards would be to encourage more centralized SHS
delivery mechanisms that ensure quality components. Revolving credit mechanisms for SHSs,
for example, often restrict loans to systems using certified components.
Similarly, a fee-for-service approach may have the potential to provide high quality systems to a
broad range of rural Kenyan households. Fee-for-service companies can only be profitable if the
systems they rent function well, so they have strong incentives to select high quality components.
Moreover, this mechanism may be embedded in a stable regulatory framework that allows for the
provision of subsidies without destructive variability, thereby increasing SHS penetration levels
(Greene et al, 1999). More careful survey work to assess the potential demand would provide
useful background for policy makers and companies considering fee-for-service options for
Kenya.
Conclusions
Single junction amorphous-silicon PV modules have suffered from a reputation for poor quality
since they entered the marketplace. This study presents evidence that this reputation is partially
deserved. One of the top selling a-Si modules in the Kenyan market falls far short of acceptable
quality standards and thousands of rural households have suffered substantial economic hardship
from investing in these products. Despite these sobering facts, it appears that this manufacturer
is taking substantial steps to improve the quality of its modules.
Inportantly, this study also clearly demonstrates that the other single junction a-Si modules
commonly sold in Kenya deliver long-term performance that is comparable to that available from
crystalline modules-but at a substantially lower retail price. There are legitimate concerns that
even the best quality single junction a-Si modules may be subject to significant breakage risk;
however, in the context of high household discount rates and falling module prices, the practical
economic difference between a 20 and 10 year expected module lifetime due to breakage may be
trivial.
In conclusion, quality amorphous silicon PV modules can provide a highly cost-effective
altemative to crystalline modules for SHSs-especially for households only willing or able to
purchase a 12-20 Wp system. There are, however, severe quality variations among the existing
manufacturers, and many consumers appear to be unable to reliably discern these differences.
Some combination of public information, certification/labeling, standards (e.g. restrictions on
uncertified modules by the government or PV loan funds) or new SHS business models (e.g. fee-
for-service) may prove helpful in educating and protecting consumers, thereby ensuring that this
market reaches its potential. There are, however, risks and limitations associated with each of
these mechanisms that necessitate a cautious approach.
21






References
Acker, R. and Kammen, D. M. (1996) "The quiet (Energy) revolution: the diffusion of
photovoltaic power systems in Kenya," Energy Policy, 24, 8 1-111.
Akerlof, G. A. (1970) "The Market for 'Lemons': Quality uncertainty and the market
mechanism" Quarterly Journal of Economics, 84, 488-500.
Argote, L. and D. Epple (1990) "Learning curves in manufacturing," Science 247, 920 - 924.
Chamberlin, C. E., Lehman, P., Zoellick, J., and Pauletto, G. (1995) "Effects of Mismatch Losses
in Photovoltaic Arrays", Solar Energy, 54(3), 165-17 1.
Chamberlin, C. E. (1998) Personal communication (Humboldt State University: Arcata,
California, USA).
Duffie, J. A., and Beckman, W. A., (1991). Solar Engineering of Thermal Processes, 2nd Edition,
(John Wiley & Sons, New York, USA).
Duke, R. D., Jacobson, A., Kammen D. M., and Hankins, M. (2000). "Mechanisms to Ensure
Product Quality in the Markets for Photovoltaics" draft to be submitted to The Energy
Journal.
Duke, R. and Kammnen, D. M. (1999) "The economics of energy market transformation
programs," The Energy Journal, 20(4), 15 - 64.
Greene, N., Duke, R. and Bryk, D. (1999) "Regulating for Renewable Rural Electrification,"
Natural Resources Defense Council Report.
Hankins, M., Ochieng F. Omondi, and Scherpenzeel J. (1997) "PV Electrification in Kenya: A
Survey of 410 Solar Home Systems in 12 Districts", submitted to The World Bank,
ESMAP, Washington, DC, USA.
Hankins, M. and van der Plas, R. J. (1998) "Solar energy in Africa: a reality", Energy Policy, 26,
295 - 305.
Hester, S. and Hoff, T. (1985) "Long-Term PV Module Perforrnance", 18th IEEE Photovoltaic
Specialists Conference, Las Vegas, USA.
Hong, S. and Wyer, R. (1990) "Determinants of Product Evaluation: Effects of the Time Interval
between Knowledge of a Product's Country of Origin and Information about its Specific
Attributes," Journal of Consumer Research, 17, 277-288.
23



Jacobson, A., Duke, R., Kammen, D. M., and Hankins, M. (2000) "Field Performance
Measurements of Amorphous Silicon Photovoltaic Modules in Kenya", Conference
Proceedings, American Solar Energy Society (ASES), Madison, Wisconsin, USA, June
16-21, 2000.
Jennings, C. (1987) "Outdoor Versus Rated Photovoltaic Module Performance", 19th IEEE
Photovoltaic Specialists Conference, New Orleans, USA.
Lehman, P. A. and Chamberlin, C. E. (1987) "Field Measurements of Flat Plate Module
Performance in Humboldt County, California," 19th IEEE Photovoltaic Specialists
Conference, New Orleans, USA.
Li-Cor (1994) "Radiation Measurement Instruments" (product literature), Li-Cor, Inc., Lincoln,
Nebraska, USA.
Neij L. (1997) "Use of experience curves to analyze the prospects for diffusion and adoption of
renewable energy technology," Energy Policy 25(13), 1099-1107.
Nyang, Frederick Oloo (1999). Household Energy Demand and Environmental Management in
Kenya, (Doctoral Dissertation, University of Amsterdam, Netherlands).
Peterson, T. (1997) "Making photovoltaics economical for electric utilities," Electric Power
Research Institute Strategic R&D Report, Fall.
PVUSA (1998) "1997 PVUSA Progress Report", submitted to the California Energy
Commission (CEC) and the Sacramento Municipal Utility District (SMUD).
Staebler, D.L. and C.R. Wronski (1977) "Reversible conductivity changes in discharge-produced
amorphous Si," Applied Physics Letters, 31(4): 292-294.
Townsend Timothy U., Reyes Antonio B., Newmiller Jeffrey D., Hutchinson Paul A., Seki Art,
Whitney Daniel D. (1998) "PVUSA's Maui Host Site PV System: A Revamping Case
Study", ASES Solar 98 Conference, Albuquerque, NM, USA.
24






The World Bank
181 8 H Street, NW
Washington, DC 20433 USA
Tel.: 1.202.458.2321  Farx  1.202-522.3018
Internet: www.worldbank.org/esmap
Email: esmap@worldbank.org
A joint UNDP/Wor,d Bank Programme