This note reports measurements of the output of bifacial photovoltaic
panels, both flush-mounted and reverse-tilt-mounted, on a sloped roof in Los Angeles.
Light striking the back side of photovoltaic solar panels is usually wasted.
Panels which are "bifacial", or sensitive to light on both sides,
are commercially available. Reports of the benefit from back illumination
by reflection of bifacial panels tilt-mounted over moderate albedo backgrounds
vary between roughly 2% and 10% (see [1-10]).
The goals of this note are a) to verify that the front-side output from the
bifacial panels is competitive with monofacial panels when flush mounted, and
b) to measure the back-side output as a percentage of the front-side output when reverse-tilt-mounted.
The panels are connected to a SolarEdge 7600 inverter that also provides per-panel
data on panel output and voltage. Readings
are taken at random times about once every ten minutes, which makes accurate comparisons difficult.
Accuracy is not specified, but is said
to be roughly 5%.
shows the output for each panel during April 23rd, 2016, from dawn to dusk.
You can see that the west-facing panels light up later in the day; you can also see the morning and evening shading at work.
A thin shadow flits across the array at 8:15am,
a tree starts to shade the lower row of the south-facing array at 5pm,
and the shadow of a ladder (!) shades panel 1.2.13 after 2pm.
Tilt mounted bifacials
Bifacial panels 1.1.1 and 1.1.2 are reverse tilt mounted, and receive significant indirect light reflecting.
Flush mounted bifacials
The remaining panels are flush mounted 15 cm above the roof.
Bifacial panel 1.1.3 is mounted at the corner of the array, and may have significant back illumination.
Bifacial panel 1.1.4 is mounted next to it, and has little back illumination.
Originally, the installer improvised using a flat roof tilt mount kit,
and did not follow the GxB300's data sheet's guidance on placement of mounting clamps,
leading to a shaky, saggy mount.
On the advice of a mechanical engineer, the installer switched to an Iron Ridge adjustable tilt mount
plus a crossbrace, and positioned the clamps properly.
The installer, not realizing these were bifacial panels, mounted the SolarEdge optimizer modules
such that they covered much of one or two cells on the back side of each panel.
Also, the minimum height above the roof was only about 10cm,
and there is very little space between the panels,
so very little direct sun reaches the roof beneath the tilted panels at midday.
It is not known how much this affects output. After the current measurements are complete
it would be good to correct some of these and re-measure. Possible corrections:
Shift panels up by about 15 cm to let more sun in below the panels
Shift bottom mount clamps to new mounting hole to maintain relative position on panel
Flip top clamp feet around to partly maintain relative position on panel
Replace center double clamp with single clamps, and spread panels apart as far as rail will allow, or
better, get slightly longer rails to allow a 30 cm gap
Move inverters to lowest rail to increase distance to panels and reduce shading
Data will be gathered each Saturday during the test period.
total energy per day for each panel on string 1, from SolarEdge's monitoring portal
current power and module voltage for each panel on string 1, from SolarEdge's monitoring portal
the temperature of the center of the front side of each bifacial panel at 1pm, measured with an IR thermometer
weather information from the National Weather Service report for the nearest station (KQCT)
Temperature readings will be taken with an Extech
IR250 IR thermometer pressed directly against the front of
the panel; the median of the readings at the center of four cells
nearest the desired point will be used.
For panels 1.1.1-1.1.2, the desired point is the center of the panel.
For panels 1.1.3-1.1.5, the desired point is one cell above the center of the panel (as the center is not reachable).
Each Saturday evening after sunset, one bifacial panel's back side will be covered with
Duvetyne (a very dark cloth),
and the other panel's back side will be uncovered, such that the cover alternates between the panels each week.
Each time the cover is moved, a timestamped photo will be taken after the move (and possibly before, to
document the condition of the cover).
Flush mounted bifacials
Prism Solar defines the term "Bifacial Gain in Energy" (BGE) as follows: BGE = (Etotal / Efront) - 100%, or
(Eback / Efront).
BGE for bifacial panel 1.1.3 will be estimated as median(E(1.1.3)/E(1.1.4)) - 100%, where
E(x) is daily energy for panel x over the test period.
Given that the accuracy of SolarEdge's monitoring is unknown, this won't
be terribly accurate.
Swapping the cables to the two panels and repeating might be a way to
average out some systematic measurement error, but I haven't checked to
see if they'd reach yet.
Tilt mounted bifacials
A key problem in comparing the output from covered vs. uncovered panels is power loss due to heating.
Covering the back side of the panel heats it up by about 7 degrees centigrade at midday.
The temperature coefficient PMPP for the GxB300 is -0.28%/C, so even with no back light,
covered panels should produce 1.7% less energy at midday than uncovered ones.
One approach to compensating for this is to measure only calm, sunny days, and only at local noon,
and apply Tc(PMPP) to correct for the temperature difference. This will underestimate
BGE, which is largest when the sun is lowest in the sky, by an unknown amount.
To exclude cloudy days, only data for days with a total energy generation of 51kWh or greater and a smooth output curve will be used.
To reduce cooling from wind, data will be discarded for days with a wind speed any time during the day of over 10 km/h in the local NWS report.
Raw BGE will be estimated each week as median(P(uncovered)/P(covered)) - 100%, where
P(uncovered) is the power measured at 1pm for the uncovered panel, and
P(covered) is the power measured at 1pm for the covered panel, corrected for the temperature
difference between the two panels.
After an even number of weeks, the raw BGEs for the weeks will be averaged.
Averaging alternating readings like this should compensate for both module power
differences and measurement equipment miscalibration.
According to Prism Solar's Design Guide for Bifacial Solar Modules v4,
for a single row of tilt mounted panels, the expected annual BGE is
(0.30/deg)*(tilt in degrees) + (11.5/m)*(h in meters) + (0.134/%)*(albedo in percent)
where h is the mininum height in meters (here, 0.1), and tilt is the tilt relative to the surface (here, 40
degrees). The tilt parameter of 40 degrees is out of range; their model only handles up to 35 degrees, so
let's substitute that.
i.e. BGE = 0.3*35 + 11.5*0.1 + 26 * 0.134 = 10.5 + 1.1 + 3.5 = 15.1%
If observed BGE is underestimated by 50% (due to only being measured when it's at its lowest),
and that formula is accurate, we'd expect to observe a BGE of about 7%.