Professional Documents
Culture Documents
Tracking Pelaez2019
Tracking Pelaez2019
Abstract—Single-axis tracking is a cost-effective deployment the increased output power from the extra-collected irradiance
strategy for large-scale ground-mount photovoltaic systems in offsets the higher cost of manufacturing bifacial cells and mod-
regions with high direct-normal irradiance. Bifacial modules in ules compared with monofacial PV modules. Further cost reduc-
1-axis-tracking systems boost energy yield by 4%–15% depending
on module type and ground albedo, with a global average of 9%. tion is expected through large-scale manufacturing techniques.
This benefit is in addition to the 15%–25% energy gain already Glass–glass modules have also been predicted to have longer op-
afforded by single-axis tracking relative to fixed-tilt deployments. erational lifetimes due to better matching of the thermal prop-
Here, we compare model results against field performance data for erties of the package materials [3]. This may also reduce the
two side-by-side bifacial/monofacial tracked systems—one in Albu- levelized cost of energy from bifacial modules. Bifacial mod-
querque, NM, USA, and one in eastern Oregon. The Albuquerque
system shows monthly rear irradiance gain of 10%–14.9%, and ules can also be used in 1-axis-tracking systems to further in-
the Oregon bifacial system has an average performance ratio 9.4% crease energy yield and offset system cost. Bizarri [4] recently
higher than the monofacial system. Both results match bifacial ir- presented results from the La Silla PV plant in Chile, where
radiance model results within uncertainty. Simulations show that a 550-kWp single-axis bifacial module array demonstrated a
smart tracking algorithms can offer more than 1% improvement 12% increase in performance with respect to standard single-
on annual energy yield by adjusting tilt angle under cloudy condi-
tions. Finally, ray-tracing simulations investigated edge brighten- axis monofacial technology. Stein et al. report daily potential
ing, suggesting a 15%–25% increase in rear irradiance at the ends bifacial gains between 8% and 14% for two single-axis trackers
of tracker rows, but up to 20% loss from center-mounted torque at Albuquerque, New Mexico, USA [5]. This promising result
tubes, creating multiple shadows. suggests that further evaluation and optimization could lead to
Index Terms—Bifacial photovoltaic (PV) module, configuration significant increases in energy yield for bifacial PV systems in
factor (CF), irradiance, model, performance, ray tracing, single- 1-axis-tracking configurations.
axis tracking. In this paper, we compare measured field performance of
several single-axis-tracked bifacial systems with neighboring
I. INTRODUCTION
monofacial systems, and with modeled expectation based on
HE solar market has seen a renewed interest in bifacial
T photovoltaic (PV) technology, which promises significant
levelized cost of energy savings in comparison with conven-
two bifacial irradiance models. In prior work, we described a
RADIANCE [6] based ray-trace model and a configuration fac-
tor (CF) model [7] for rear-side irradiance Grear calculation
tional monofacial PV modules [1], [2]. Bifacial solar cells and of fixed-tilt systems, and we verified them for fixed-tilt condi-
modules can collect light from both sides, including light re- tions with field data [7]–[9]. Here, we extend these models for
flected from the surrounding ground surface. To some degree, single-axis-tracking applications.
The Radiance model offers the possibility of reproducing
Manuscript received September 25, 2018; revised November 24, 2018; ac- complex scenes, including tracker-element shading. Run times
cepted January 8, 2019. This work was authored in part by the National Renew- of several seconds or minutes are possible with simple geome-
able Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, try and when using a cumulative-sky approach, which calculates
for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-
08GO28308. in part by the U.S. Department of Energy’s Office of Energy front-side and rear-side bifacial irradiance based on a single an-
Efficiency and Renewable Energy (EERE) under Solar Energy Technologies nual cumulative sky source for the year [10]. This composite
Office (SETO) Agreement Number 30286. (Corresponding author: Silvana Ay- source comprises all hourly Perez diffuse-sky and direct-solar
ala Pelaez.)
S. A. Pelaez is with the University of Arizona, Tucson, AZ 85705 USA, and contributions for the year. The CF model also incorporates the
also with the National Renewable Energy Laboratory, Golden, CO 80401 USA Perez tilted-surface model [11] and assumes 2-D geometry to
(e-mail:,silvanaa@email.arizona.edu). calculate the fractional irradiance entering and leaving the sur-
C. Deline is with the National Renewable Energy Laboratory, Golden, CO
80401 USA (e-mail:,chris.deline@nrel.gov). faces of an infinite-length PV array. This approach is also fast
P. Greenberg is with the NRGWise Lighting, Albany, OR 97321 USA (∼seconds), but does not include complex shading or finite-
(e-mail:,nrgwise.lighting@gmail.com). system edge effects.
J. S. Stein is with the Sandia National Laboratories, Albuquerque, NM 87123
USA (e-mail:,jsstein@sandia.gov). Both models are freely downloadable [12], [13] and can be
R. K. Kostuk is with the University of Arizona, Tucson, AZ 85705 USA used to evaluate single-axis-tracking scenarios. With single-axis
(e-mail:,kostuk@email.arizona.edu). tracking, the modules are no longer at a fixed tilt, and the clear-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. ances to the ground and with neighboring rows in the array
Digital Object Identifier 10.1109/JPHOTOV.2019.2892872 are constantly changing. Tracking algorithms can be used to
2156-3381 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Fig. 1. On a tracking system, the fixed parameters are the distance between
the centers of rotation and the axis height. Module ground clearance, tilt, and
separation between arrays vary with the solar position. The trackers presented
in this paper rotate around an N–S axis.
Fig. 2. Sandia bifacial tracking array with front and rear irradiance sensor
shown (center). Two rows with GCR 0.28.
calculate these parameters based on the ground coverage ratio
(GCR) 3) determining the irradiance received by the ground by
CW accounting for shading and restricted view of the sky due
GCR = (1)
rtr to the tracker’s position;
4) determining the irradiance for the backside of the PV
where CW is the PV collector width (overall width of the mod-
module.
ules in a row) and rtr is the distance between the rotation axis
of the panels as shown in Fig. 1. GCR is used in tracking algo-
rithms to implement backtracking corrections to the tilt of the B. Tracking Radiance Ray-Trace Model
trackers, based on minimizing shading from neighboring arrays. The Radiance bifacial PV model [11] has also been updated to
This correction becomes particularly important for arrays with allow for tracking systems. Additional steps include calculating
higher GCRs. We can also define a normalized axis height H: the array tilt, ground clearance, and row-to-row spacing for each
axis height time step. The conventional fixed-tilt simulation workflow uses
H= . (2) a single annual sky source to calculate annual average bifacial
CW
gain [9]. For a tracked system, multiple scene geometries are
required, along with the solar resource corresponding to each
These normalized parameters allow comparisons between
tracker tilt angle. Here, we create separate simulation and scene
tracker designs of different dimension (e.g., 2-up landscape ver-
geometry for each tracker tilt in 5° increments, along with the
sus 2-up portrait) because the self-shading geometry and bifacial
cumulative hourly solar resource corresponding to this solar
rear irradiance depends on these normalized parameters, not on
zenith angle.
absolute dimensions.
Once the position of the tracker is known, the Perez tilted-
surface model can be recalculated for each configuration. For III. FIELD COMPARISON SYSTEMS AND METHOD
the ray-trace model, a modified cumulative sky is used to deter- Several bifacial tracking systems were investigated for this
mine the diffuse-sky irradiance received by the array at different study. The first is a small-scale research array at Sandia National
angles throughout the year. Laboratories, and the second is a set of 100-kW commercial
systems in eastern Oregon.
II. BIFACIAL MODELS FOR 1-AXIS TRACKING
The tracking algorithm from PVLib [14] is used to compute A. Sandia National Laboratories PV System
the array tilt for both the RADIANCE and CF models. Back- This deployment consists of two rows of 1-axis-tracked bi-
tracking corrections have been employed to reduce self-shading facial modules (Fig. 2). Reference cell detectors are mounted
of the panels throughout the day based on the GCR of the system. on the front and back of some of the modules to measure solar
illumination. The tracker axis height is 0.5 m, and the track-
A. Tracking CF Model ers are spaced from each other with GCR = 0.28. The ground
albedo for the site was unmeasured, but assumed to be 0.25
The CF model described in detail elsewhere [7] has been up-
for aged concrete. The field data are used to validate modeled
dated to allow for single-axis-tracked bifacial systems. Given
Grear /Gfront using the CF model based on site-measured me-
an array(s) with a specific axis height and orientation, the ad-
teorological data, also available on the PVDAQ database [15].
ditional tracking-specific steps of the updated CF model are as
follows:
1) determining the array’s tilt, ground clearance, and row-to- B. Eastern Oregon Demo Site
row spacing with other arrays based on the specific time, Two commercial tracked systems east of Klamath Falls, Ore-
location, and backtracking (optional); gon, are within 20 km of each other (Fig. 3). One is composed
2) identifying the ground region that is shaded by the PV of 100 kW of Silfab 285 bifacial modules in 2-up landscape,
arrays under this configuration; adjacent to 100 kW of Trina 300-W monofacial modules in
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
1-up portrait. The second system is composed of 200 kW of coefficients including temperature coefficient are included, then
Silfab 285 bifacial modules, also in 2-up landscape orientation. Yf ,m o deled will reflect the nonbifacial aspects of performance
Each system uses six Chint 36-kW inverters, with ac production that differ between two module types. Equation (3) is rewrit-
monitoring. Hourly site-irradiance satellite data are provided by ten with a correction factor based on the front-only difference
The Weather Company’s Cleaned Historical API to calculate expected for the two module types, which will help isolate the
performance ratio (PR). energy gain due to bifaciality
Inspection of the site geometry showed module collector
Yf ,bifi Yf ,m o deled,m ono
width of 2 m and row spacing of 5.65 m, and a measured GCR of BGE ,bifacial = 100% × −1 .
0.35. Similarly, a 1.5-m tracker hub-height measurement indi- Yf ,m ono Yf ,m o deled,bifi
cates a normalized axis height H = 0.75. Field current–voltage (4)
(I–V) curves were taken for one bifacial module in the center The BGE ,bifacial field-measured energy yield value is used
of a row under three different rear-albedo conditions: Rear ir- to validate the bifacial optical models described above that
radiance completely blocked with black cloth; natural ground model Grear . Here, measured BGE ,bifacial is compared with
cover; and high-reflectance (∼80%) white ground cloth. The BGE ,M o del where
cloths were placed covering the space between three of the Grear
west-most trackers. Relative to the zero rear-irradiance condi- BGE ,M o del = ϕP m p × (1 − ηloss ) . (5)
Gfront
tion, natural ground cover increased module power by 10%, and
reflective ground cover increased power by 20%, under mostly For bifacial system performance modeling, Grear and Gfront
sunny conditions. The field of view for all cells in the module are front and rear modeled irradiance, respectively; ϕP m p is the
was ∼150°. PV module 1-sun bifaciality; the rear vs front power ratio is
defined in [19]; and ηloss accounts for additional bifacial loss
C. Field Data Comparison Method terms, such as shading loss and irradiance mismatch. Here, ηloss
is assumed to be zero, and Grear is averaged across the back of
Bifacial system performance can be evaluated by comparing the bifacial module. Previous work [9] has shown good agree-
measured and modeled energy yield (Yf ) and PR [16] for bifa- ment between modeled and measured spatial distribution of rear
cial and reference monofacial systems. Measured ac electrical irradiance. However, the effect of nonuniform illumination on
energy is aggregated for each site, excluding times when the performance and BGE has not been considered here.
site experienced inverter or tracker issues. Overall energy gain
for a bifacial system is determined by comparing energy yield IV. RESULTS
Yf [kWh/kW] for both monofacial and bifacial systems [5]
A. Model Comparison and System Optimization
Yf ,bifi
BGE = 100% × −1 . (3) The advantage of 1-axis tracking for high-DNI climates is
Yf ,m ono
well established [20], [21]. Fig. 4 shows annual Grear /Gfront
Although BGE gives a value of overall system performance bifacial improvement for two modeled TMY3 climate condi-
advantage, not all of this gain is due to a system’s bifaciality. tions: Albuquerque, NM (high DNI), USA, and Seattle, WA
Improved low-light efficiency and better temperature coefficient (low DNI), USA. Both bifacial irradiance gain and front-side
can improve performance as well and will be captured in (3). irradiance are reduced as tracker spacing is reduced, due to in-
Therefore, isolating the bifacial response requires normaliza- creased self-shading on the front and reduced ground-reflected
tion of Yf by modeled front-side performance for both module irradiance on the rear surface. For a modeled albedo = 0.25
types. Here, Yf ,m o deled is calculated with the PVLib single- and H = 0.75, in Albuquerque, the rear-irradiance ratio ranges
diode model [17], [18] using inputs of hourly site irradiance from Grear /Gfront = 8%–9%. In Seattle, due to increased diffuse
data, and standard test conditions (STC) (front side only) mod- irradiance, Grear /Gfront = 9.5%–11%.
ule parameters. The inverter’s efficiency as a function of power The CF model was applied to evaluate tracker bifacial gain
is also considered in calculating Yf ,m o deled . If proper model for many locations around the globe using satellite-based TMY
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Fig. 5. Modeled Gre a r /Gfro nt (%) for 1-axis-tracked systems over natural
ground cover. Assumed geometry: GCR = 0.35, H = 0.75.
Fig. 6. Measured and modeled rear-irradiance gain Gre a r /Gfro nt for March
30, 2017.
Fig. 8. Side-by-side production for two 100-kW tracked systems. Bifacial
PR ranged from 0.7 to 0.9, and averaged 9.4% higher than the monofacial
irradiance data and satellite-measured albedo values from system’s PR.
NASA [22] (see Fig. 5). The system configuration assumed was
0.35 GCR, H = 0.75. For some high-albedo equatorial loca- values in May show anomalously large BGE , potentially due
tions, the gain from the 1-axis bifacial tracking was found to be to tracker misalignment reducing Gfront during this month.
as high as 20%, but a more typical global average value was 9%.
Note that ϕP m p = 100% and ηloss = 0 was assumed, which C. Eastern Oregon Data
may be too optimistic for real fielded bifacial systems.
AC energy yield data were collected for the side-by-side bi-
facial and monofacial installations. PR values range from 0.63
B. Sandia National Laboratories Results
to 0.90 for the two systems, with a cumulative average of 0.738
Fig. 6 shows 5-min measured and modeled Grear /Gfront for for the monofacial, and 0.807 for the bifacial systems. Although
a single sunny day. It is clear that Row 1 is positioned to the field I–V curve measurements indicate comparable front-side
east of Row 2 and therefore has a reduced Grear in the morning capacity for the two systems, the measured PR was on aver-
when trackers are pointing east, and the backside has a view age 9.4% higher for the bifacial system than for the monofacial
of the row behind. Likewise, Row 2 has a reduced Grear in system.
the afternoon when trackers are pointing west and the back As mentioned above, system performance can be influenced
of the modules have an obstructed view due to the position of by effects other than bifaciality, including temperature coef-
Row 1. A large increase in measured BGE is also visible for both ficient differences. To isolate bifacial gain, (4) is used, with
rows during afternoon cloudy conditions. This further illustrates expected system PR based on site temperature, irradiance, and
the increased bifacial gain that can be expected under cloudy module nameplate (front side) parameters. These modeled PR
conditions, or in climates with high diffuse-irradiance fraction values are 0.777 for the monofacial, and 0.795 for the bifa-
[1], [23]. cial systems resulting in a corrected BGE,Bifacial = 7.0%. This
Cumulative rear versus front production is analyzed for 7 means that about 2.4% of the measured performance advan-
mo and compared with BGE ,M o deled using the CF model tage of the Silfab HIT modules is due to improved front-side
and onsite measure irradiance data. The cumulative measured performance, rather than bifacial response.
BGE shown in Fig. 7 is 11%, with monthly variation between Monthly BGE,bifacial values are shown in Fig. 8 based on
10%–14.9%. The monthly BGE ,M o deled varies little from measured ac energy production, as well as bifacial performance
month to month, with cumulative average =12.3%. Measured model results. Here, CF model hourly results are calculated
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Fig. 9. Bifacial gain BGE , b ifa c ia l per (4) measured from ac production data,
compared with CF bifacial energy-gain estimates, monthly results.
Fig. 12. Yearly cumulative distribution of BGE across the center row of a
20-module × 7-row single-axis-tracking array for (a) H = 0.4 and (b) H =
0.75. BGE is significantly higher (13.9%) at the edges of the row. Radiance
simulation. GCR = 0.35, Albuquerque climate.