ArticlePDF Available

Abstract and Figures

Today's demanding project financing climate requires developers to hone annual photovoltaic (PV) energy estimates with unprecedented accuracy - and to back the estimates with meaningful long-term performance guarantees. With some snowy locales in the U.S. and southern Canada becoming increasingly popular for MW-scale PV systems, lenders are now requiring that snow losses be estimated as part of their energy simulations. The literature is exceptionally thin on this subject - we have been unable to find even a single side by side study that directly quantifies the difference between an always-clean array versus an identical one left to naturally accumulate and shed snow. This paper describes the design and reports results for a side by side PV test bed installed in California near Lake Tahoe in December 2009. It has been designed to gauge the energy loss due to snow for three common tilt angles. Results from the first winter are presented, with insights for future model development and ongoing measurements.
Content may be subject to copyright.
MEASURING AND MODELING THE EFFECT OF SNOW ON PHOTOVOLTAIC SYSTEM PERFORMANCE
Loren Powers, Jeff Newmiller and Tim Townsend
BEW Engineering, San Ramon, CA, U.S.A.
ABSTRACT
Today’s demanding project financing climate requires
developers to hone annual photovoltaic (PV) energy estimates
with unprecedented accuracy -- and to back the estimates with
meaningful long-term performance guarantees. With some
snowy locales in the U.S. and southern Canada becoming
increasingly popular for MW-scale PV systems, lenders are now
requiring that snow losses be estimated as part of their energy
simulations. The literature is exceptionally thin on this subject -
we have been unable to find even a single side by side study
that directly quantifies the difference between an always-clean
array versus an identical one left to naturally accumulate and
shed snow [1],[2].
This paper describes the design and reports results for a side by
side PV test bed installed in California near Lake Tahoe in
December 2009. It has been designed to gauge the energy loss
due to snow for three common tilt angles. Results from the first
winter are presented, with insights for future model
development and ongoing measurements.
BACKGROUND
Historically, PV modules installed in snowy climates have been
part of small, off-grid arrays mounted at very steep tilt angles.
This is done both to shed snow quickly and to maximize winter
output. Unfortunately, this concedes too much annual energy
to be a good design strategy for larger contemporary systems.
Today’s snowy climate PV systems tend to be installed at angles
shallow enough to make them prone to snow loss.
Both weather and array design factors influence the amount of
snow loss. Weather factors include the quantity and quality
(moisture content) of the snow, the recurrence pattern of
storms, and the post-storm pattern of temperature, irradiation,
wind speed, wind direction, and relative humidity. Array design
factors essentially boil down to orientation (fixed or tracking,
tilt, azimuth, and tracker rotation limits) and the surrounding
geometry, that is, open rack or building integrated. Building
features can either help (e.g., melt) or hinder (e.g. dam up or
drift) natural snow shedding.
Figure 1 illustrates how roof tilt and features can influence
whether a series of snowfalls will shed or accumulate. This roof
is in Truckee, CA, with the photo taken three days after a
December snowfall. The steeper pitch of 35-40° is clear, while
the shallower 18-23° section has retained several separate
layers of snow and has effectively become a shallower
“winterized” slope with each successive event. On the left side,
the extra drifted snow from the valley and higher roof segment
illustrates what could happen if a PV array were present.
Figure 1. Influence of tilt angle on snow retention
The National Renewable Energy Laboratory’s (NREL) 30-year
TMY-2[3] database is widely used by solar researchers, and for
many sites, includes two snow-related data columns. Unlike the
hourly data it lists for solar and weather variables, daily data
are listed for:
1. Snow depth (cm)
2. Number of days since last snowfall
Lacking field measurements of snow loss, BEW developed an
analytical model in 2008 to make use of NREL’s data. The
estimates we have prepared have fallen in the 2-5% annual loss
range. These results are not huge, but are not negligible. Our
experience includes a mix of ground-mounted tracking arrays
that are subject to heavy snows but which shed them rapidly by
virtue of the tracking mechanism, along with others in less
snowy areas that may experience comparable percentage
losses because they are oriented at shallow fixed tilts. Our
coarse snow loss estimates contrast sharply with anecdotal
reports of larger snow losses for some fixed tilt arrays. For
example, NREL’s PVWATTS program advises that 70% winter
month reductions were noted in Minnesota for a fixed 23° tilt
array and 40% losses were noted even for a 40° tilt array[4].
One compensating aspect of snowy climates is they do tend to
have minimal soiling losses in the summer months due to
regular year-round precipitation. In the southwest U.S.,
summertime dust losses can reach 20%, causing annual energy
losses of 5% or more in the absence of manual washing. For
snowy climates, operators are faced with parallel questions:
what is the value and cost of manual snow removal? While the
cost of snow removal is outside the scope of this paper, the
value of snow removal is a key objective of our ongoing work.
HYPOTHESIS
The state of the art in predicting annual energy losses can be
improved by applying results from a side by side PV module
test bed to generalized simulations. It should be possible to
estimate the annual energy impact of snow to within ±1%. This
will require carefully controlled field measurements, with the
understanding that the permissible uncertainty for predicting
monthly snow losses may still be more than an order of
magnitude higher than the desired annual uncertainty. The
reason this comparatively large error is tolerable is because the
sensitivity of annual energy forecasts to a single winter month’s
production is low – often, less than 5% of annual production
occurs in a winter month for a low-slope array. Therefore,
relatively large errors in estimating snow losses will not
necessarily spoil the accuracy sought in annual revenue
forecasts.
Monthly and annual snowfall varies widely from year to year,
even in consistently snowy locations. This large natural
variability makes precise determinations for a specific year
unnecessary and probably misleading. BEW feels an
appropriately realistic goal in developing snow loss estimates is
to produce monthly snow loss reference tables with as much as
a ±20% uncertainty for a given climate, as that level of monthly
uncertainty still translates to a much smaller annual uncertainty
of about ±1% after seasonal radiation weighting is accounted
for.
SITE AND TEST BED CONSIDERATIONS
Our goal with the first of what we hope to be a network of
snow test beds has been to measure the most significant
variables that can be readily correlated with reduced PV
output. Candidate variables in approximately decreasing level
of significance include:
1. Snowfall/snow depth
2. Structure orientation (fixed or tracking with tilt, azimuth,
and rotation angles as applicable, and open-rack or
building integrated mounting)
3. Visual record of snow buildup
4. Air and module temperatures
5. Plane of array irradiation
6. Wind speed and direction
7. Snow moisture content
8. Relative humidity
It was felt the first five items in the list could be addressed
within our first-time private budget and our ability to make
good use of the data; we did not feel ready to commit the
funds to perform a comprehensive test covering a multitude of
tracking, building integrated, and off-azimuth systems, nor to
collect a wealth of micro-climate specific data such as wind,
snow moisture, or humidity.
Instead, we found a host operator in Truckee, CA (near Lake
Tahoe) who was willing to both place the 12’x35’ test bed on a
flat, unobstructed parcel and to keep one set of modules and
irradiance sensors clean. The Truckee test bed’s results will be
used to calibrate (or refine) the analytical model and to develop
simpler empirical models, if feasible.
A custom Campbell-based datalogger and camera with a cell
modem and web display interface was designed and fabricated.
The data acquisition system (DAS) is equipped with six sensors
to record short-circuit currents (as a proxy for power) for three
pairs of cleaned and snowy PV modules, as well as several
temperature and irradiance sensors. The three tilt angles
include 0° (flat, as a worst-case), 24° (latitude minus 15°), and
39° (latitude). While one steeper pair at 54° (latitude plus 15°)
was also considered, we reasoned it would be unnecessary to
simply find that, as we neared the 55° critical angle for point
release avalanches[5], the amount of lost energy would be
negligibly small. Plus, there are virtually no commercial arrays
placed at such steep tilt angles and comparatively little value to
obtain such data.
Truckee, California, is a high-altitude (≈6,000’/1,800 m) location
with an average of 200 in./500 cm. of snow per year. However,
a 50
+
year database[6] shows the standard deviation is ±37%,
with extremes ranging from 50-200% of normal. Monthly data
vary even more than the annual totals. January, the snowiest
month, accounts for one-fourth of the annual total. However,
January’s 48 in. average also exhibits a standard deviation twice
as large as the annual deviation (±73%) and has exhibited a
range of near zero to over 300% of normal. Given this high
degree of variability, simple monthly snow loss estimates to the
nearest whole percentage point seem more than adequate.
The Lake Tahoe area is not a prominent solar market, though
the Truckee Sanitary District installed a 125 kW
P
35° fixed tilt
array in 2009 and there are several other commercial PV
installations in the Lake Tahoe region. Compared to Truckee’s
200”/yr, the following well-established commercial solar
markets and their average annual snowfall include[7]:
1. Denver, 60 inches
2. Milwaukee, 47 inches
3. Boston/New England, 43
+
inches
4. Detroit (and Ontario Canada), 42 inches (and much more
going eastward toward Buffalo)
5. Chicago, 38 inches
6. Mid-Atlantic region, 20-30 inches
Figure 2 shows the most recent year’s snowfall trend as a solid
bold line, with dashed lines showing the monthly averages and
their normal 1-std. deviation envelope. From June 2009
through May 2010, the snowfall of about 190 inches was 96%
of average, so our first year’s results should be very
representative for this location. February 2010 was notably
dry, receiving just 48% of its average snow, yet this was still
well within the bounds of the normal year to year variation for
that month. In the past year, only May 2010 fell outside the
normal range, with 12 inches received instead of the average 4
inches.
0
10
20
30
40
50
60
70
80
90
100
Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct
Snowfall, in.
Truckee 2009-10 monthly snowfall vs. long-term typical ranges
Snow, in.
LTAvg Snow, in.
low norm.
hi norm.
2009-2010 Snowfall of ≈190
inches was 96% of average. The
annual snowfall is highly variable:
1 std. dev. is ±37%. Monthly
variability is even higher, as
shown by the error bands.
This year, only May was outside
the normal snow range.
Snow data from U.S. SNOTEL Stn.
#834 , 4 mi. south and 700' higher
than BEW test site.
Figure 2. Truckee monthly snowfall trends
ARRAY DESIGN
The Truckee test bed consists of three pairs of south-facing
Mitsubishi 175 W
P
poly-Si modules portrait-mounted at 0°, 24°,
and 39° tilt angles. One of each pair of modules is manually
kept clean of snow and frost. Each module has an inactive 18”
border of similarly textured and colored material to minimize
edge effects. The module pairs are spaced far enough apart to
prevent row shading, even on the winter solstice. Air and
module temperatures are recorded, along with module current.
The modules are short-circuited, producing up to 8 amps of dc
current at 1,000 W/m
2
.
Three Licor pyranometers are also used, one on each plane,
with a fourth pyranometer mounted facing downward from the
rear side of the 39° tilt plane. This sensor was put in to help
characterize the radiation received from ground-reflected and
north sky sources
1
, but from its snow-protected position, later
proved invaluable in helping identify and rehabilitate the
roughly 5% of records that were compromised by snow
accumulation on the “clean” side modules, since it was never
obscured by snow. Our hosts made over two dozen service log
entries over the winter, yet there were still some occasions
when otherwise clear conditions were not captured by the test
bed. We relied on a combination of camera evidence and data
screening to identify and adjust anomalous records using
quality-checked data to scale and replace errant records. For
future studies, we intend to outfit the clean side modules with
thermostatically controlled, insulated electric heat tape on the
module backsides. This will ensure rapid melting of snow and
1
An interesting side result: the monthly irradiation on the back
side of the module was 25% of the front side irradiation when
the ground was mostly snow-covered in Dec-Apr, but just 10%
when dry ground prevailed for most of May.
ice and minimize both the cleaning labor and the likelihood of
anomalous data points.
Figure 3 illustrates the test bed design. Concrete ballasts are
used to anchor the Unistrut, aluminum stock, and 4x4 lumber
framing. Pro Solar racking is used to mount the modules. A
plexiglass on blue-painted OSB laminate is used to border the
modules. The data logger and current shunt sensor enclosure is
mounted on the north side of the assembly, and poles to show
snow depth and to mount the camera are located nearby. This
view is to the NW.
Figure 3. Test bed design
INSTALLATION AND OPERATION
The test bed was installed over a dry three-day period at the
beginning of December 2009 and was up and running just a few
hours before the season’s first major storm hit. Figures 4 and 5
show the installation’s progress.
Figure 4. Installation begins
Figure 5. Installation complete
Figures 6 and 7 show a typical cleaning episode and hourly
snapshot, respectively. The left, or west, side of the array is the
manually cleaned side.
Figure 6. Cleaning the array
Figure 7. Webcam snapshot
As these photos show, the snow is deep enough to pile up
higher than the 18” low edge of the array and dam up and
impede the natural shedding of snow. In retrospect, we should
have raised the structure perhaps 2’ higher and spaced the flat
array section a foot or so farther from the middle section, and
would probably do so in subsequent installations to avoid
confounding the results and making the impact worse than it
need be. On the other hand, almost all fixed tilt commercial
systems are mounted very close to the roof membrane and are
spaced closely enough to create exactly this kind of damming,
so this geometry may be more realistic for some types of
arrays.
RESULTS
The first winter was statistically very normal. The lost energy
due to snow buildup in the 7-month winter season ranged from
as little as 25% for the 39° tilt = latitude orientation to as much
as 42% for the flat orientation. The seasonal results project to
losses in annual output of 12%, 15%, and 18% for the 39°, 24°,
and 0° tilts, respectively. While these results are hugely
significant for this location, it is not clear how well such results
should translate to other locations. Some inferences can be
made, though.
Table 1. 2009-10 Truckee snow season results
Month Snow % loss in generation for
each tilt (Nov loss was
estimated using similar May
data)
inches % avg. 0 deg. 24
deg.
39
deg.
Nov 14 88 10 8 6
Dec 44 126 81 80 79
Jan 36 75 100 90 80
Feb 20 48 93 69 29
Mar 37 100 51 28 17
Apr 29 193 37 20 16
May 12 300 9 7 5
Nov-May 192 96 42 33 25
Annual
(estimated)
192
+
96
+
18 15 12
For example, Denver has a latitude and elevation similar to
Truckee, though its average annual temperature is 3.7 °C
warmer and it receives only 30% of the snowfall that Truckee
gets. However, like Truckee, it:
receives snow in all but the mid-summer months
its snowiest month, March, gets about the same amount
of snow as Truckee saw this May (12”), and
has the same average temperature in March as what
Truckee recorded this year in May (4 °C)
Based on this, one might infer that the snow losses in Denver in
March may be comparable to the 5-9% losses seen in Truckee
in May for similarly tilted fixed arrays.
Figure 8 illustrates the table data. The observed energy loss for
each month and tilt angle are plotted on the left scale, and the
% of average snowfall is shown as a dashed line referenced to
the right scale.
0%
100%
200%
300%
400%
500%
600%
700%
800%
900%
1000%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Nov
Dec
Jan
Feb
Mar
Apr
May
% Average Snowfall
% Energy Loss
Truckee monthly snow loss for three tilt angles
% loss 0 deg. % loss 24 deg. % loss 39 deg. % avg snow
Figure 8. Monthly snow losses for three tilt angles
Figure 9 removes the snowfall data and expresses the losses as
a function of tilt angle. Losses are assumed to approach zero in
any month as tilt angles approach 60 degrees. This plot offers
some thin clues as to what the losses might be for tracking
arrays. Actual measurements will need to be made, of course,
but casual observations by BEW’s staff and discussions with PV
operators suggest tracking systems will shed snow similarly to
fixed tilt arrays tilted 20-30 degrees higher. For example, a
single-axis tracking system inclined at 20 degrees in Truckee
might experience snow losses of 8% per year, or about half of
the snow loss a fixed 20 degree tilt array may experience.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0 10 20 30 40 50 60
% Energy Loss
Array Tilt Angle, degrees
Truckee 2009-10 snow season:
Energy loss vs. Tilt Angle for measured months
Losses at 60 degree tilt and greater are set to virtually zero
Annual
Dec
Jan
Feb
Mar
Apr
May
Figure 9. Monthly losses as a function of tilt angle
Figure 10 illustrates the annual snow-related energy loss as a
function of tilt angle. Some liberties in extrapolation were
taken here. For one, we assume that snow losses become
negligibly small at some critical angle, perhaps as low as 45
degrees but conservatively shown here as being reduced to
negligibly small at 60 degrees. We also applied long-term
average solar radiation data for the Truckee-Tahoe TMY-3 NREL
station for the purpose of estimating full-year energy loss since
our test bed has only been active for six months.
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
0 10 20 30 40 50 60 70 80 90
% Energy Loss
Array Tilt Angle, degrees
Truckee 2009-10 snow season energy loss
Losses at 60 degree tilt and greater are set to virtually zero
Calculated loss based on measured data
Predicted based on cos(tilt)^2
Possible empirical model assumes
maximum flat array loss is 0.1%
annual loss per inch of snow (20%
for 200 inches), multiplied by the
cosine(tilt)^2.
% Loss = 0.1*(Snow, in.)*cos
2
(tilt).
The RMS error for this correlation
is ±2%.
Figure 10. Measured and fitted trends for energy loss as a
function of array tilt
Here, we offer the simplified observation that the generalized
relationship between losses, snowfall, and tilt angle can
perhaps be adequately represented by the equation:
Annual % loss = 0.1 x [Snow, in.] x cos
2
(tilt) Eqn. 1
The first coefficient, 0.1, was not regression-fitted. It carries the
implied units of % per inch. It was selected based on the
observation that a near-20% annual loss corresponded to a
near-200 inch annual snowfall, or 0.1%/yr/inch of snow. This
correlation suggests a typical error of ±2% for predicting annual
energy loss, with the overall correlation looking pretty good up
to about 45 degree tilt angles, and fairly poor for commercially
invisible steeply tilted arrays. This is not good enough to call
the job done, but, subject to additional data collection at other
locations, potentially represents a considerable improvement
over the current lack of any simple empirical estimating tools.
We arrived at the cos
2
relationship partly by inspection and
partly, as shown below in Figure 11, by optimizing the exponent
to minimize the RMS error. While there is a clear relationship
between tilt angle and the natural gravity-driven tendency to
shed snow, a first-order cosine relationship alone is not
aggressive enough to explain the observed loss trend. The
upper blue trend line below shows this. The cos
2
relationship
works much better, as does the cos
3
form, which actually
exhibits the best overall RMS error. We suggest the cos
2
form
only because of its closer fit to the observed data for the more
common sub-40 degree tilt angles. Higher-order exponents in
the proposed cos
N
relationship clearly tend to under-predict
relative to the observed bold red trend line.
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
0 10 20 30 40 50 60 70 80 90
% Energy Loss
Array Tilt Angle, degrees
Truckee 2009-10 snow season energy loss
Losses at 60 degree tilt and greater are set to virtually zero
Calculated loss based on measured data
Predicted based on cos(tilt)
Predicted based on cos(tilt)^2 Predicted based on cos(tilt)^3
Predicted based on cos(tilt)^4
Predicted based on cos(tilt)^5
RMS error , by exponent:
cos(tilt) => ±4.5%
cos
2
(tilt) => ±2.2%
cos
3
(tilt) => ±1.9%
cos
4(
tilt) => ±2.6%
cos
5
(tilt) => ±3.4%
Figure 11. Alternate curve fits to annual snow loss model
CONCLUSIONS AND FUTURE WORK
The first winter of operation has yielded a wealth of significant
data on snow-related impact on PV output. For one of the
U.S.’s snowiest urban areas, it was observed that annual losses
of 12-18% may be expected in a typical year for fixed tilt arrays
mounted at tilt angles ranging from 39 degrees to 0 degrees
(flat). Monthly losses were substantially higher, with an entire
month’s output lost for a shallow tilt angle when several feet of
snow fell.
A promising simple annual snow loss relationship was posed,
which suggests annual energy loss may be estimated as the
product of a 0.1% per inch snow loss, multiplied by a cos
2
(tilt
angle) adjustment factor.
We would like to extend some of our observations for snow
loss in the milder winter months at Truckee and apply them to
represent more severe months at less snowy locations, and
presented such an example for Denver. However, we will await
better site-specific data before suggesting the Truckee data can
be responsibly applied to other climates.
There is a clear relationship between tilt angle and energy loss,
though the relationship will be influenced by factors we did not
eliminate in our installation, namely, the damming of snow
caused by too low of an array height and too little spacing
between rows. BEW plans to modify the Truckee test bed to
address these factors for the next winter season, and hopes to
find partners to install similar test beds in other snowy areas
where photovoltaic systems are being deployed.
REFERENCES
1 Becker, Gerd, An Approach To The Impact Of Snow On The
Yield Of Grid Connected PV Systems, Bavarian Association for
the Promotion of Solar Energy, Munich, 2007.
2 Marion, Bill, Instrumentation For Evaluating PV System
Performance Losses From Snow, NREL, ASES 2009 conference,
Buffalo, NY, May 2009.
3 The National Renewable Energy Laboratory (NREL) published
the National Solar Radiation Database, also commonly referred
to as the Redbook, in 1994 (Bill Marion, principal author), with
web address: http://rredc.nrel.gov/solar/old_data/nsrdb/1961-
1990/tmy2/
4 NREL’s PVWATTS website address is
http://rredc.nrel.gov/solar/calculators/PVWATTS/system.html .
This PVWATTS citation discusses snow losses in Minnesota.
5 Powers, Phil, Wilderness Mountaineering, Stackpole Books,
1993, page 35.
6 Western Regional Climate Center, Truckee Ranger Station
#049043, 1903-2009 data, 52 valid years for annual snow
records, web address: http://www.wrcc.dri.edu/index.html
7 One source for snowfall data has this web address:
http://www.currentresults.com/Weather-
Extremes/US/snowiest-cities.php
... The output showed a decrease in the annual energy production of 12%, 15% and 18% for 39°, 24°and 0°tilt respectively. The significant finding from this research work was that the reduction in power loss is a function of snow coverage on the PV panel surface and tilt angle of the PV array [45]. Marion et al. investigated the energy loss from the PV system due to snow coverage for six PV systems installed in Colorado and Wisconsin for the winters of 2010-2011 and 2011-2012. ...
... The study found a reduction in energy loss of 90% and a decrease in annual energy production from 1% to 12% [44]. Table 5 addresses the impact of snowfall on PV productivity [44][45][46][47][48]. ...
... Poly-crystalline Powers et al. [45] Annual power losses of 12-18% United States and Canada. Poly-crystalline Table 6 Effects of sandstorm on PV performance Reference Sandstorm impact Location Type of cells ...
Article
The energy yield from the solar photovoltaic plant mainly depends on available solar flux, quality of the related power-conditioning equipment incorporated in the system, technical specifications of the panel, the geographic location and also on the environmental parameters. This study provides a comprehensive review of the effect of environmental factors on the various components of photovoltaic systems. It emphasises the environmental factors such as dust, ambient temperature, wind velocity, humidity, snowfall, hailstorms and sandstorms, which deteriorates the energy efficiency of solar plants and the various failure modes of the panels caused by these factors. Finally, the focus is on the methods to find out different failure modes of photovoltaic panels and the various mitigation techniques to improve the energy yield. These mitigation techniques are essential for positioning of photovoltaic arrays in remote, desert, dusty and areas proximate to higher wind speed. This review provides an outlook for the developers to take precautionary measures before locating the site and designing the solar plants.
... This could be explained with snow deposit in the PV system. According to the report by Powers et al. [28], It was found that the losses due to snowfall are dependent on the angle and technology of the PV system that occur with naturally accumulated snow. These samples should not be considered while building the model and should be treated as outliers. ...
... The predicted power generation for the whole day on August 20, 2016 was 17.3 kWh and the actual generation was 18.8 kWh. Figure 6. 28 shows the predicted vs actual graph. RF performed better than both DT and SVM model for the sunny day. ...
Thesis
The main aim of smart grid research field is to have increased integration of large-scale renewable energy systems. However, the integration is challenging as the Photovoltaic (PV) power generation is prone to fluctuations and it is affected by different weather conditions. Therefore, the accurate forecasting becomes vital for grid operators to manage the grid operations. Solar energy forecasting is still in the exploration phase. In this study, multiple machine learning models are used, and different weather parameters are analyzed for PV forecasting. The performance of Random forest (RF) model is compared with Support vector machine (SVM) and Decision tree (DT) models. The proposed forecasting models are tested with 4 kW PV device installed at fortiss GmbH and comparison of the models is done through error analysis. The accuracy is evaluated using historical weather forecasts. The results shows that RF based model performed better than SVM and DT based models. This study focuses on presenting a forecasting model to produce forecasts for thin-film PV cells which is not investigated in the related work done in this field. Therefore, the role of diffuse radiation parameter in determining the prediction accuracy is examined. The focus of the thesis is to determine the best possible PV forecasting method while having two different datasets: one with solar insolation readings and one without them by just using the readily available weather information. The study proposes a novel approach of PV forecasting where a hybrid model is built based on principle of pipe-lining clustering, classification and regression algorithms. Clustering and classification algorithms group the data-points with similar weather conditions. After that, Segmented regression is applied on these groups. Weather forecast of the next day is utilized to determine the models used for forecasting. The results show that proposed forecasting model for PV systems is effective and promising.
... Annual production losses from a snow-covered PVs are directly proportional to the amount of snow 775 received and proportional to the squared cosine of the tilt angle of the panels (Powers et al., 2010). 776 ...
Article
Full-text available
The inclusion of photovoltaic (PV) technologies add extra functionalities in a building by replacing the conventional structural material and harnessing benign electricity aesthetically from PV. Building integration (BI) and building attached/applied (BA) are the two techniques to include PV in a building. Currently, first, and second-generation PV technologies are already included for BIPV and BAPV application in the form of wall, roof, and window whereas third generation PVs are under rigours exploration to find their potential suitability. To alleviate enhanced temperature from both BIPV and BAPV, active and passive cooling can be introduced, however passive techniques are influential in trimming down the temperature for retrofit building. Shading from snow, dust cover and nearby building can be an obstacle for BIPV/BAPV application. The hydrophobic (icephobic) self-cleaning coating is suited for snow covering PV while hydrophobic and hydrophilic are both applicable for anti-soiling. Electric vehicles, autonomous switchable glazing, low heat loss glazing and lightweight BIPV are the different future application for PV in BI and BA integration.
... Snowfall in these climates can be a significant source of energy loss [2] and consequent reduction in levelized cost of energy. Prior studies have found that annual energy losses due to snow can exceed 30% [3], but they are typically less than 10% [4][5][6][7][8]. Accounting for even relatively small snow-related losses in PV system performance is important for large scale systems [9]. ...
Conference Paper
Full-text available
Energy losses due to snow coverage can be high in climates with large annual snowfall. These losses may be reduced with region-specific system design guidelines. One possible factor in snow retention on PV systems could be frame presence and/or shape. Sandia is studying the effect of module frame presence on photovoltaic module snow shedding for a pair of otherwise-identical PV systems in Vermont. The results of this study provide a summary of the findings after the 2018-2019 winter period. The results clearly show that the presence of a frame inhibits PV performance in mild winter conditions.
... As shown by Aarseth et al. [119], it is indeed possible to increase the annual production sufficiently to offset some costs, though how much may be a matter of local weather conditions, as pointed out by Pawluk et al. [126]. In some regions one may only add 1-3% to the annual production [2], while more snow rich regions, be it due to altitude or latitude, may yield >15% more energy produced per year [127]. ...
Article
As building integrated photovoltaics (BIPV) are becoming increasingly popular, the demand for optimized utilization will be increasing with respect to efficiency, aesthetics and reliability. In cold climate regions, we predict that there will also be a growing focus on how to avoid snow and ice formation on the exterior surfaces of BIPV. During the winter period there is substantially less incident solar radiation. This is also the period when the solar radiation is most needed for heating, lighting and power production purposes. The task to avoid accretion of snow and ice is challenging due to the fact that snow, ice and ambient weather conditions exist in countless variations and combinations. Snowfall, freezing of rain water and condensation of air moisture with subsequent freezing, are examples of aspects that have to be addressed in a satisfactory way. The present study aims to review the cold weather challenges facing BIPV, the strategies for overcoming them and the opportunities that follow from successfully overcoming them.
... A six-year-data acquisition of a 28degree roof mount PV system in Germany showed that the values of proportional annual yield reduction by snowfall losses range from 0.3 to 2.7% [11]. The authors of [12,13] introduced a generalised monthly snow loss model taking into account the effect of insolation, humidity, temperature, ground interference, and tilt angle for two PV sites in Truckee, CA, USA. Average annual losses of 6-26% for different tilt angles of 35°-0° were obtained. ...
Article
Full-text available
In this study, the snow melting behavior of several PV technologies, all installed at the same location under identical conditions, is analyzed based on the time-dependent changes of the snow cover, which is extracted from images of a monitoring webcam, for various temperature and irradiation conditions. From this study, conclusions can be drawn for the optimum module technology for a given location with respect to snow-dependent yield losses. In particular, the melting behavior is analyzed regarding its dependence on the ambient temperature and the irradiation level. Finally, the relevance of snow cover-related losses is discussed. The study shows that comparably large frameless modules exhibit the highest snow shedding rates. Hence, they are snow-free for longer periods thereby increasing their potential for electricity generation in snowy regions. In summary, this paper reveals the beneficial snow removal properties of large frameless modules for snowy areas by applying a novel image processing technique for the determination of the snow-covered area fraction of the modules.
Article
Full-text available
Background: Due to the lack of reliable diagnostic tools, clinical data on the significance of most invasive fungal infections are difficult to assess and information on frequency, disease pattern and prognostic impact still largely relies on autopsy data. Methods and results: To determine temporal trends in invasive fungal infections, we analyzed data from 8124 autopsies performed between 1978 and 1992 on patients who died at the University Hospital of Frankfurt/Main. During that period, a total of 278 invasive fungal infections were found. The prevalence rose from 2.2% (1978-82) and 3.2% (1983-87) to 5.1% in the most recent years (P < 0.001). Besides the emergence of mixed and unclassified infections, this was mainly due to a significant increase in Aspergillus infections (P < 0.001), whereas the prevalence of Candida infections was stable and even showed a declining trend within the last years. The highest infection rates were found in aplastic syndromes (68%), followed by AML (25%) and AIDS (19%). In the majority of cases (76%), invasive fungal infection was related to the immediate cause of death. However, the proportion of patients with endstage underlying conditions increased significantly over time from 53% to 80% (P < 0.001). Accordingly, the number of patients who were not considered terminally ill but had died from fungal infection dropped from 35% to 17% within the last years (P < 0.01). Conclusions: These observations document significant changes in frequency, aetiology and underlying disease processes in invasive fungal infections at autopsy and underscore the continuing need for more effective prevention, diagnosis, and treatment.
Conference Paper
Full-text available
When designing a photovoltaic (PV) system for northern climates, the prospective installation should be evaluated with respect to the potentially detrimental effects of snow preventing solar radiation from reaching the PV cells. The extent to which snow impacts performance is difficult to determine because snow events also increase the uncertainty of the solar radiation measurement, and the presence of snow needs to be distinguished from other events that can affect performance. This paper describes two instruments useful for evaluating PV system performance losses from the presence of snow: (1) a pyranometer with a heater to prevent buildup of ice and snow, and (2) a digital camera for remote retrieval of images to determine the presence of snow on the PV array.
Article
Full-text available
Invasive fungal infections cause significant morbidity and mortality in immunocompromised children. The prevalence of invasive aspergillosis (IA) is increasing as a reflection of the rising numbers of immunocompromised patients and the increasing use of aggressive immunosuppressive treatment regimes for hematologic malignancies and transplantation. IA is almost exclusively seen in severely immunocompromised or critically ill children, including those with the classic risk factors (particularly neutropenia, hematopoietic stem cell transplant or solid-organ transplantation, hematological malignancies, use of systemic immunosuppressive agents or cytotoxic therapies). Early treatment improves survival rates, but the diagnosis of aspergillosis remains difficult and, while IA has been relatively well-characterized in adults, far fewer studies have described optimal treatment for the pediatric population. This article reviews and compares the newer, less-invasive diagnostic techniques that are becoming available and focuses on the data specifically from pediatric trials regarding efficacy, safety and pharmacokinetics of the antifungals used for IA.
Article
Full-text available
Invasive aspergillosis (IA) is a serious complication in patients undergoing allogeneic haematopoietic stem cell transplantation (HSCT), particularly from donors other than HLA-identical sibling. All 306 patients who underwent alternative donor HSCT between 01 January 1999 and 31 December 2006 were studied. Late IA was defined as occurring >or=40 days after HSCT. The median follow-up was 284 days (range, 1-2709). Donors were matched unrelated (n=185), mismatched related (n=69), mismatched unrelated (n=35) and unrelated cord blood (n=17). According to European Organization for Research and Treatment of Cancer/Mycoses Study Group criteria, 2 patients already had IA at HSCT, 23 had early IA and 20 had late IA (IA incidence 15%). Eight patients had proven and 37 probable IA. Multivariate analyses showed that significant predictors of IA were delayed neutrophil engraftment, extensive chronic GVHD (cGVHD), secondary neutropenia and relapse after transplant. Early IA was associated with active malignancy at HSCT, CMV reactivation and delayed lymphocyte engraftment. Late IA was predicted by cGVHD, steroid therapy, secondary neutropenia and relapse after HSCT. IA-related mortality among IA patients was 67% and was influenced by use of anti-thymocyte globulin, steroids, higher levels of creatinine, and lower levels of IgA and platelets. The outcome of IA depends on the severity of immunodeficiency and the status of the underlying disease.
Article
A retrospective review of medical records, microbiology and pathology laboratory records, and nosocomial infection surveillance data was undertaken to describe the experience with culture-documented aspergillus infection in pediatric cancer patients at our facility. Sixty-six patients were identified from a 34-year period. The most common underlying diagnosis was leukemia. Risk factors included neutropenia, immunosuppression, and prior antibiotic therapy. On the basis of clinical presentation, 23 patients were believed to have disseminated disease and 43 to have localized disease. The lung was the most frequently affected organ. Despite aggressive medical and surgical management, overall mortality was 85% within the first year after diagnosis. Patients who presented with disease in sites other than the lungs fared better than patients with initial pulmonary involvement (P = .0014). Aspergillosis continues to be associated with poor outcome. Development of improved medical and adjuvant therapies, including surgery, is warranted.
Article
Invasive fungal infections (IFIs) continue to cause considerable morbidity and mortality in haematopoietic stem cell transplant recipients. The epidemiology of IFI has changed since the late 1980s, with a trend towards a reduction in invasive infection due to opportunistic yeasts and an increase in invasive mould infections, particularly by Aspergillus spp. Since the introduction of fluconazole for prophylaxis, the incidence rate of invasive candidiasis is close to 5% and the risk factors related to invasive candidiasis are gastrointestinal tract colonisation, cytomegalovirus disease and a prior episode of bacteraemia. The highest risk for invasive aspergillosis was observed in older patients and patients with graft-versus-host disease and immunosuppressive therapy, steroid use (>1-2 mg/kg/day), persistent neutropenia and certain types of transplantation (cord blood transplant, allogeneic mismatched or T-cell depletion). In those cases, rational preventive measures must be implemented and vigilance is necessary in order to diagnose infection as soon as possible.
Article
Invasive mould infections remain major causes of infection-related mortality following hematopoietic stem cell transplantation (HSCT). In this review, we summarize the recent advances in the diagnosis, prevention, and management of invasive mould infections in HSCT recipients. The evolving epidemiologic characteristics of post-HSCT invasive mould infections, specifically the rising incidence of Aspergillus and non-Aspergillus mould infections in the postengraftment period, necessitate the development of preventive strategies. The efficacy of prophylactic broad-spectrum triazoles against invasive mould infections in HSCT recipients has now been demonstrated in two large prospective studies. However, concerns over drug absorption, interactions, and costs may shift attention from universal prophylaxis to risk stratification and preemptive strategies. In this regard, recent studies have highlighted the potential of genetic polymorphism analysis to identify HSCT recipients at risk for invasive aspergillosis, and efforts are underway to improve the predictive values of antigen and nucleic acid detection assays. Emerging data on risk factors for invasive aspergillosis relapse after HSCT, antifungal drug monitoring, and the use of galactomannan testing to monitor treatment response may help inform therapeutic decisions for HSCT recipients. Evidence-driven management of invasive mould infections in HSCT recipients is becoming increasingly individualized, integrating host factors and pharmacologic and epidemiologic considerations. However, the optimal approach to invasive mould infection prevention in HSCT recipients remains to be resolved by prospective clinical studies.
Article
Detection of galactomannan antigen in serum or bronchoalveolar lavage fluid is useful for diagnosis of invasive aspergillosis, but the current serologic tests are not useful. (1-->3)-beta-D-glucan can be detected in the serum of patients who have aspergillosis, candidiasis, and a few other mold infections, but false-positive results limit its reliability for diagnosis. None of these methods is useful for diagnosis of zygomycosis and certain other less common mold infections. This article focuses on the detection of antigen and beta-D-glucan for diagnosis of invasive aspergillosis and candidiasis.
Article
Invasive fungal infections (IFIs) represent a major complication in recipients of hematopoietic stem cell transplantation and solid-organ transplantation. The incidence of IFIs in transplant recipients has increased over the past 20 years, and these infections continue to be associated with high morbidity and mortality. This article reviews the important concepts guiding the management of IFIs in transplant recipients, including epidemiologic trends, new risk factors, and a timetable of infections, pathogens, therapy, and prevention of these infections. An emphasis is given to invasive aspergillosis.
Article
Acute graft-versus-host disease (GVHD) significantly limits the application and the success of allogeneic hematopoietic stem cell transplantation (HSCT). Novel therapies that target the aberrant immune response underlying GVHD are reviewed with particular emphasis on immunomodulatory agents currently incorporated into clinical trials. In addition, regenerative stromal cellular therapy (RSCT) is discussed as an emerging form of novel GVHD therapy. Knowledge for transplant immunology, particularly as it relates to underlying pathophysiology of GVHD, has dramatically increased over the last decade. As a result, new immunomodulatory therapies have been used to treat steroid-refractory GVHD. However, their success has been limited by their lack of clinical experience during HSCT as well as by their associated toxicity profiles. RSCT uniquely offers the potential to enhance donor-derived hematopoiesis and immunity and to ameliorate adverse sequelae associated with GVHD. An exciting era incorporating the use of cellular therapeutics during HSCT has arrived. As the experience and understanding for cellular therapies, in general, and RSCT, in particular, increases, so too will their success in benefiting the HSCT recipient beyond limitations of current pharmaceutical agents.