Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

An Overview of Spread Spectrum Time Domain

Reflectometry Responses to Photovoltaic Faults


Mashad Uddin Saleh, Student Member, IEEE, Chris Deline, Senior Member, IEEE, Evan Benoit, Student
Member, IEEE, Samuel Kingston, Student Member, IEEE, Ayobami S. Edun, Student Member, IEEE,
Naveen Kumar Tumkur Jayakumar, Student Member, IEEE, Joel B. Harley, Member, IEEE, Cynthia
Furse, Fellow, IEEE, Michael Scarpulla, Senior Member, IEEE

Abstract—Spread spectrum time domain reflectometry circuit interruption (AFCI) [2], detect the faults and protect the
(SSTDR) is a broadband electrical reflectometry technique system but do not locate the faults to allow fixing the root cause.
that has been used to detect and locate faults on live Time-domain reflectometry (TDR) techniques [5] are well-
electrical systems, including photovoltaic (PV) systems. In established methods for both locating and detecting faults in
this paper, we evaluate the detectability and localizability electrical systems. An electromagnetic signal is sent through the
from both existing literature and our own measurements system, and reflections from impedance discontinuities return
using SSTDR of open circuit faults, connection faults, short to the test end where they are analyzed. The time delays
circuit faults, ground faults, arc faults, shading faults, between the incident signal and the reflections indicate the
bypass diode faults, and accelerated degradation faults in distances to the discontinuities, and magnitudes and phases of
PV cells and mini-modules. Reflection magnitudes for these the reflected signals are related to the types of impedance
faults are compared. Preliminary data on buried and discontinuities.
grounded PV cable along with arc fault detection are
Each type of reflectometry uses a different type of signal and
presented.
method of analyzing the reflections [5]. Traditional TDR uses
Index Terms—Degradation, reflectometry, arc faults, ground stepped or pulsed voltages. Frequency domain reflectometry
faults, photovoltaic cells, spread spectrum time domain (FDR) [6] and mixed-signal reflectometry (MSR) [7] use
reflectometry (SSTDR), open circuit faults, connection faults, sinusoidal signals. Chaos or noise domain reflectometry [8], [9]
short circuit faults, ground faults, arc faults, shading faults, bypass use existing or added noise in the system. Spectral [10] or
diode faults, accelerated degradation faults. binary [11] time domain reflectometry use pseudo-noise (PN)
codes, and spread spectrum time domain reflectometry
I. INTRODUCTION (SSTDR) [12] uses modulated PN codes. The magnitudes of the

F aults (any problem or imperfection) in photovoltaic (PV)


systems can occur due to degradation, weather,
mishandling, etc. Common electrical problems include ground
reflections will be essentially the same for each reflectometry
approach, if the bandwidth is the same [5]. In this paper, we
evaluate SSTDR for detecting and location faults in PV
faults (GF), arc faults (AF), open circuits faults (OC), systems. A major advantage of this method over other
accelerated degradation faults (ADF), shading faults (SF), reflectometry approaches is its ability to detect and locate faults
bypass diode fault (BDF), connection faults (CF), and broken within operating (energized) electrical systems.
cells or modules (BRK) [1]. Recent surveys of advanced fault Originally developed for the location of faults on live aircraft
detection approaches include model-based difference wires [13], SSTDR has been demonstrated for detection and
measurements [2], real-time difference measurements [3], and location of open and short faults [12], arc faults [14], conductor
output signal analysis [1]. These studies include initial damage [15], monitoring power electronic devices [16]. In PV
assessments of SSTDR for detection and location of mainly systems, it has been used to detect, and in some cases locate
ground and arc faults, but the range of faults had not yet been ground faults [17], arc faults [18], disconnection faults [12], and
thoroughly tested. We seek to extend that assessment in this broken panels [19]. The magnitude of the reflected signal
paper. Commercially available protection schemes including indicates the magnitude of the impedance of the fault, and the
ground fault detection and interruption (GFDI) and arc fault delay indicates its location. In this paper, we will review the

This material is based upon work supported by the U.S. Department of UT 84112, USA (email: evan.benoit@utah.edu, u0763921@utah.edu,
Energy’s Office of Energy Efficiency and Renewable Energy (EERE) under naveen.tumkurjayakumar@utah.edu).
Solar Energy Technologies Office (SETO) Agreement Number DE- A. S. Edun and J. B. Harley is with the Department of ECE, University of
EE0008169. Also, we want to express gratitude to Gardner Energy (West Florida, Gainesville, FL 32611-6200, USA (email: aedun@ufl.edu,
Haven, UT) for their support with testing on their PV modules. Also, we want joel.harley@ufl.edu).
to thank Kent Terwilliger and Byron McDanold for their help with C. M. Furse is with the Electrical and Computer Engineering Department,
measurements at NREL. University of Utah, Salt Lake City, UT 84112 USA, and LiveWire Innovation.
M. U. Saleh is with the Department of ECE, University of Utah, Salt Lake (e-mail: cfurse@ece.utah.edu).
City, UT 84112, USA (email: mashad.saleh@utah.edu). C. Deline is with M. Scarpulla is with the Department of Materials Science and Engineering
National Renewable Energy Laboratory, 1617 Cole Blvd, Lakewood, CO and the Department of ECE, University of Utah, Salt Lake City, UT 84112,
80401 (email: chris.deline@nrel.gov). E. Benoit, S. Kingston and N. K. T. USA (email: mike.scarpulla@utah.edu).
Jayakumar are with the Department of ECE, University of Utah, Salt Lake City,

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
relative magnitude of reflections from common PV faults to W50A0093 SSTDR instrument [23]. Its characteristic
evaluate the potential capability of SSTDR technique to detect impedance is ZSSTDR = 68 Ω. We used 24 MHz continuous
and locate faults in PV power plants. mode, which uses a 24 MHz PN code modulated by a 24 MHz
In section II, we will explain the SSTDR measurement square wave. Table II shows the single cell mini-module
techniques and parameters. In section III, we will review parameters from National Renewable Energy Laboratory
common types of faults seen in PV systems, and provide a (NREL) cell team. Illuminated current-voltage (I-V)
quantitative comparison of the magnitudes of SSTDR characteristics were measured in two ways: using Kelvin probes
reflections from these faults. This is used to determine the at both ends of each bus bar; and with multipoint probes
detectability and localizability of the faults using SSTDR. The distributed along the bus bars. The Kelvin probe measurements
conclusions are summarized in section IV. get the proper short-circuit current density, Isc, but have a
reduced fill factor due to added series resistance. The multipoint
II. SSTDR RESPONSES probe measurements get the proper fill factor but have reduced
Spread spectrum time domain reflectometry (SSTDR) [20] Isc due to shading.
sends a square-wave modulated pseudo-noise (PN) code signal Figure 1(b) shows the SSTDR signatures for open circuit
into an electrical system. Reflections from impedance (OC), short circuit (SC) loads, and from a PV cell 50 ft from the
discontinuities return to the SSTDR instrument, where the SSTDR. The SSTDR signature is normalized to the magnitude
returned signal is cross-correlated with a delayed version of the of the open circuit response at this location, which the effects
incident signal to create the reflectometry signature. In Fig. 1 of attenuation in the cable. A similar but inverted signature is
(a), the setup schematic for an open circuit (OC), short circuit seen for the short circuit load. The distance (from the SSTDR)
(SC), and a PV single cell mini-module at the end of a twin lead shown in Fig. 1 was calculated by taking the measured time
leader cable is shown. delay and dividing by the velocity of propagation, vop. We
smoothed and spline interpolated the peaks as described in [22]
for improved accuracy. In addition to the reflection signatures
shown at the load, there is always a similar pulse at the
impedance discontinuity from the SSTDR instrument to
transmission line (distance =0), which we have not shown.
(a) Table I
PV cable [21] and SSTDR parameters
1 Open circuit
Parameters Values
Short circuit
0.8
PV single cell mini-module Conductor diameter (dc) 2.94 mm
0.6
Conductivity of copper (𝜎𝜎𝑐𝑐 ) 5.98 × 107 S/m
0.4

Loss tangent of XLPE 4 × 10−4


0.2

𝜇𝜇𝑟𝑟 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋 0.999994


Normalized Reflection

XLPE insulation thickness (𝑡𝑡𝑖𝑖𝑖𝑖 ) 3.375 mm


-0.2

-0.4
Distance between conductors (Dc) 11 mm

-0.6
𝜀𝜀𝑟𝑟𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋 2.3

-0.8
𝑍𝑍𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 68 Ω
𝑓𝑓𝑚𝑚 24 MHz
-1
vop 0.722 ± 0.008
35 40 45 50 55 60 65 70 75 80
Zo 178 Ω
Distance [feet]
ZSSTDR 68 Ω
(b)
Fig 1. Measured SSTDR reflection responses for a simple test system (a). The Table II
smoothed and normalized SSTDR responses (b) are shown for various loads Single cell mini-module parameters
(Short circuit, open circuit, and a PV single cell mini-module).
Parameters Values

The twin-lead cable is made up of 10 (AWG) gauge multi- Cell area 244.2 cm2 ± 0.1%
conductor cross-linked polyethylene (XLPE) coated 2 kV Open circuit voltage (VOC) 0.6267 ± 0.0011 V
photovoltaic wires [21] taped together to form a twin-lead cable,
Short circuit current (ISC) 9.035 ± 0.055 A
with parameters given in Table I. To find its characteristic
impedance, we attached a potentiometer as the load and tuned Maximum Current (Imax) 8.542 ± 0.052 A
it until reflections were not visible at the SSTDR. This gave Zo Maximum voltage (Vmax) 0.527854 ± 0.000074 V
= 178 Ω. The velocity of propagation, vop = 0.722 ± 0.008 times Maximum power (Pmax) 4.509 ± 0.028 W
the speed of light in vacuum (co), was found by measuring the
SSTDR response of an open circuit on a known length of cable Efficiency 18.46 ± 0.11 %

[22]. For our tests, we used a Livewire Innovation ARNOLD

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
Table III
Overview of the SSTDR Technique for fault detection and location
Detected Fault Conditions Advantages and Limitations
OC CF GF AF SF BDF BRK ADF
 Fault detection independent of fault
current levels [13]
Detection: √ √ √ √ √ * √ √  Can be used in live systems [10]
Location: √ √ * * * * * *  Can both detect and localize faults [10]
 May require baseline for assessment
[12]
 More complex, less accurate on
parallel branched circuits [24]
[12] [12] [14] [17] [18] [19]

√ Demonstrated * Potentially Feasible (Work in progress)

Note: GF: Ground Fault, AF: Arc fault, OC/SC: Open Circuit / Short Circuit Fault, SF: Shading Fault, BDF: Bypass Diode Fault, CF: Connection Fault,
ADF: Accelerated Degradation Fault, BRK: Broken Cell/Module Fault.

We created mini-modules with one PV cell which shading faults (SF), bypass diode fault (BDF), connection faults
wasencapsulated in glass and ethylene vinyl acetate (EVA), (CF), and broken cells or modules (BRK) [1]. In this section,
similar to a typical PV module [25]. The cells (6-inch by 6-inch we will review these types of faults, and provide quantitative
custom-made polycrystalline silicon solar cells similar to [26]) comparisons of their SSTDR reflections. We also compare the
were soldered to busbars covered by EVA and then covered use of SSTDR to detect and locate faults with contemporary
with a back sheet (made of Tedlar composite). MC4 connectors detection techniques and summarize in Table III.
were soldered to the lead tabs.
In reflectometry (TDR, SSTDR, etc.), the reflection A. Open Circuit (OC), Connection Faults (CF), Short Circuit
coefficient (Γ) tells how much signal will be reflected, and is a (SC) Faults
function of the impedances of the transmission line Zline and
fault Zfault: Open (OC, with impedance Z = ∞) and short circuit (SC, with
Z=0) faults can be caused by damaged or corroded connectors,
𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑍𝑍𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 −𝑍𝑍𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
𝛤𝛤 = = (1) physical damage to the wiring or PV elements, or as an after
𝑉𝑉𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑍𝑍𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 +𝑍𝑍𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
effect of arc faults of various kinds. Connection faults (CF),
which are open circuits or high impedance, nearly open circuits
in which 𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 and 𝑉𝑉𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑒𝑒𝑛𝑛𝑛𝑛 are the reflected and incident
can occur between modules because of improper installation,
voltages. The reflection coefficient is frequency-dependent,
damage to connectors and insulators from weather,
however, when the fault is not frequency-dependent (such as
rough handling during installation or maintenance, corrosion,
OC, SC, CF), Γ is equal at all frequencies and can be applied
etc. These faults create the largest reflections of any of the faults
across the time domain reflection signature [27]. For OC, Γ= 1.
and are therefore the easiest to detect and locate.
For SC, Γ= -1. These can be seen in Fig. 1 (b), where the OC
Using today’s non-SSTDR techniques, OC and CF faults can
magnitude is normalized to 1, the SC has the same shape and
be detected but not located using output voltage and current at
magnitude as for OC but is inverted. For OC and SC, all of the
SSTDR signal is reflected either in phase (OC) or out of phase maximum power points [2] [28] and output DC power [29].
(SC). This is the largest reflection we can expect in any Location is often done by visual inspection and general
electrical system, and we will consider our ability to locate all electrical debugging (such as sequentially disconnecting
other types of faults relative to this maximum reflection portions of the system to determine which section has the fault).
coefficient (Γ= 1). The location of the reflection (the x-axis in Using SSTDR, connection faults can be detected and located
Fig. 1 (b)) is found by multiplying the time delay by the velocity between PV cells and modules [12]. The maximum change in
of propagation on the cable. SSTDR can detect and locate OC SSTDR normalized reflection magnitude was reported as 0.5
and SC faults in realistic wiring systems without disconnecting for disconnections between seven mini-modules connected in
them and while energized [12]. In the following section, we will series and 0.7 for disconnections between eight full-sized
evaluate the magnitude of the reflection for various types of series-connected modules [12]. However, for larger strings, the
faults in PV systems, to evaluate the relative magnitude of the change in magnitude might be smaller due to attenuation and
reflections from these faults, which is the key parameter to increased multiple reflections in the string.
determine if reflectometry can detect and locate them.
B. Ground Faults (GF)
III. DETECTING FAULTS IN PV STRINGS
There are several types of faults that can disrupt power from Ground faults (GF) occur if there is a connection between a
PV systems. These include ground faults (GF), arc faults (AF), current-carrying conductor and a grounding path [1]. This is a
open circuits faults (OC), accelerated degradation faults (ADF), very common fault in PV systems. Generally, GF are detected

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
by analyzing the output voltage and current [30] and output insulated, undamaged) with a short circuit on the end was run
power [31]. Ground faults are more complex than simple short through a 5-foot section of metal conduit, as shown in Fig. 3.
circuit (SC) faults because the return path for the reflection is Tap water was poured into the metal conduit to submerge the
no longer a single, low-loss wire. It is now typically a larger, portion of the cable in the conduit, simulating the filling of a
distributed conductor (such as earth/dirt, chassis, or the buried conduit by water. This changed both the magnitude of
framework holding the PV modules). This reduces the effective the reflection and velocity of propagation (seen from the delay
reflection coefficient through both multipath reflections and in the reflection), as shown in Fig. 4. Future experiments to
attenuation. SSTDR has been used for detecting and locating detect and locate ground faults in a full PV system are
ground faults (to the airframe) in aircraft wiring [14] and ground recommended.
faults in PV systems [17], indicating that this type of fault can
be detected in PV systems.
A simple experiment imitates a ground fault and gives an
example of the magnitude of the reflection from GF. One lead
of a twin lead wire (18 AWG lamp cord 50 feet long with a
short on the end) was manually shorted to a grounded metal
plate (8 feet by 6 feet) (shown in Fig. 3). The fault was
introduced at 43.5 feet by removing a 0.5-inch section of
insulator from one of the wires making up the twin lead and
manually touching it to the metal plate. The SSTDR signals Fig 3. A portion of a 50-feet long lamp cord is fed through a 5-feet long metal
shown in Fig. 2 clearly indicate that the fault is detectable, both electrical conduit where the wires are held about 1.5 inches from the steel plate.
in the initial reflection at 43.5 feet and in subsequent multiple the twin lead lamp cord is fully insulated (undamaged). The conduit is either
dry or filled with water. For the ground fault experiment, the portion of wire
reflections. The maximum change in normalized SSTDR was stripped of insulation and touched to the metal plate.
reflection magnitude is 0.15 at the location shown by asterisks
in Fig. 2.
1

Dry
0.4
No fault 0.8
Wet
0.3 Ground fault
0.6

0.2
0.4

0.1
0.2
Normalized Reflection

0 0
Normalized Reflection

-0.1 -0.2

-0.2 -0.4 delay

-0.6
*
-0.3

-0.8
-0.4 0.15

* -1
-0.5
40 50 60 70 80 90 100
30 40 50 60 70 80
Distance [feet]
Distance [feet]

Fig. 4. SSTDR measurements for fully insulated, undamaged lamp cords in


Fig 2. SSTDR measurements of a ground fault on a 50-foot twin-lead lamp cord
with a short circuit on the end. Insulation was removed from one wire of the metal electrical conduit, dry or full of water. The delay indicates the reduction
twin lead at 43.5-feet from the SSTDR. It was manually touched to the metal in the velocity of propagation.
plate shown in Fig. 3, simulating a ground fault.
C. Arc Faults (AF)
Another qualitative measure, “area” was introduced in [17]. Arc faults are short circuits that carry a large current that arcs
This area is found from the sum of the absolute values of the between two conductors. They can be caused by damage or
element-wise differences between the average autocorrelation insulation breakdown [4]. Parallel arc faults occur between
plot (similar to Fig. 1 (b)) with a ground fault in a specific conductors, while series arcs occur between two segments of a
location and the baseline measurement of the system with no- single conductor. Currently, AF are detected on PV systems by
fault. Using this analysis, ground faults were found to be analyzing the output currents [32] [33] or signal to power ratio
detectable. A preliminary method was also presented to locate (arc power to inverter noise) [34]. A parallel arc fault is
the fault position. The detectable fault area is about 3 times the effectively a short circuit (SC), but for a very transient duration
system noise area, which is similar to our observations. (often only fractions of a millisecond) [14]. Series arc faults
Another type of ground fault can be imitated by submerging produce partial disconnections (OC) such as those seen in [19],
a cable in water. A 50-foot long twin-lead lamp cord (fully and these are often intermittent, as well. In many cases, this arc

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
is so fast that it does not produce enough energy and heat to trip the test setup shown in Fig. 5 (a) [35]. The test string had an
a traditional circuit breaker, so arc fault breakers that evaluate open circuit voltage (VOC) of 650 V and short circuit current
the output current are used to detect (but not locate) them. The (ISC) of 18 A. The goal was to simulate the arc produced by a
impedances of both parallel and series AFs are similar to their faulting MC-4 connector. Two copper rods aligned axially were
SC and OC counterparts, but they occur for only very short used to simulate the connector contacts. An arc was created
times and therefore require a very fast system that is actively across a very small separation between the tips of the rods. The
testing while they occur. rods were then slowly moved apart until the arc dissipated. The
The SSTDR technique was used to detect series and parallel SSTDR signatures (initial, final and several tests during the
AF in PV systems [18] and to detect and locate parallel AF in period the arc was present) are shown in Fig. 5(b). Similar tests
aircraft [14]. In [18], the magnitude of the normalized reflection were also reported in [18]. The maximum change in normalized
from an arc fault was about 0.24 times that of the baseline SSTDR reflection magnitude is 0.3 at the location shown by
SSTDR signature. We also see similar results. The system asterisks in Fig. 5 (b).
design tradeoffs between test speed and sensitivity limit the
distance from the SSTDR to which test systems can be sensitive D. Shading Fault (SF)
to AF. It is important to note that for SSTDR, the current Shading faults can be temporary (e.g. clouds, flying birds,
induced by the arc (typically a very noisy signal) does not dust, or shadows) or permanent (soiling, long term dust
interfere with the test system. In fact, unlike arc fault breakers accumulation, debris). SFs can cascade to produce short circuit
and other arc fault detection schemes, it is the impedance faults or hot spots [36]. Hot spot heating occurs when a number
change (the short/ open circuit of the arc), not the arc current of series-connected PV cells cause a large reverse bias across
that is detected and analyzed. the shaded cell, leading to large dissipation of power in that cell.
Shading faults are detected by analyzing output power, voltage
and current [29] [37] by existing power electronics. These test
methods cannot locate which cell or panel has the fault, but
visual inspection may be used for this purpose.

(a)
(a)
0.4

0.3

0.2

0.1

0
Normalized Reflection

-0.1

-0.2

-0.3

-0.4 Shading Module A


0.12
Shading Module B
-0.5
Day data without shading

180 200 220 240 260 280

Distance [feet]

(b)
(b) Fig. 6. (a) Schematic for the test setup to take shading fault measurements, and
Fig. 5. (a) PV arc fault test setup is made up of two copper rods that are slowly (b) SSTDR measurements from data taken in full sun with and without shading.
separated to create an arc similar to those produced by faulty MC-4 connectors.
(b) SSTDR can locate this arc.
Shading faults create impedance changes in the PV module,
similar to those from day/night variation, albeit over a smaller
We tested SSTDR for series arc faults in small PV systems
section of the module. A preliminary experiment to detect SF
at the National Renewable Energy Laboratory (NREL) using
was done at our test station at the University of Florida. A 194-

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
ft leader cable was used to connect two PV modules (36 cell F. Broken Cell/Module Fault (BRK)
modules from Renogy) in series. SSTDR measurements were
taken in full sun. Fig 6 (a) shows the setup schematic. Fig. 6 Cells or modules can be broken due to mishandling during
(b) shows the normalized reflection with no modules shaded, transport or installation, weather and extreme environmental
and with either module A or B shaded. The maximum change conditions or bad packaging. Infrared (IR) imaging,
in normalized SSTDR reflection magnitude is 0.12 at the electroluminescence imaging, and visual/ manual inspections
location shown by asterisks in Fig. 6 (b). are normally used to detect and locate broken cells/modules in
Moreover, day/night changes are also detectable, as can be a string [39]. Not all breaks cause impedance changes
seen in Fig. 7. The experiment was done in our test station at (especially when the busbars remain intact), but when they do,
the University of Florida. Fig 7 (a) shows the schematic of the we would expect SSTDR to be able to detect and locate BRK
setup, and Fig. 7 (b) shows the SSTDR responses during the faults. As an example, SSTDR responses before and after
day (in full sun) and at night. breaking a Jinko JKM285W module are shown in Fig. 8 [19].
The maximum change in normalized SSTDR magnitude is 0.1
which is shown by asterisks in Fig. 8. In addition to changes in
the first reflection, additional changes are seen in later
(multiple) reflections.
0.4
No broken cell in the module

(a) 0.3 Broken cells in the module

0.2

0.1
0.5 Night

Day 0
0.4
Normalized Reflection

-0.1
0.3

-0.2
0.2

-0.3
0.1
Normalized reflection

0 -0.4

*
-0.1 -0.5
0.1
*
-0.2 -0.6
40 60 80 100 120 140 160

-0.3
Distance [feet]

-0.4
Fig. 8. SSTDR responses before and after breaking the Jinko PV module.
100 150 200 250
G. Accelerated Degradation Fault (ADF)
Distance [feet]

(b) Degradation faults occur in PV modules due to aging and poor


Fig. 7. (a) Schematic for the test setup to take day/ night measurements, and (b) material or climatic stress. Degradation can take many forms
SSTDR measurements from day and night. including deterioration and changes in the dielectric materials
that make up the PV cells, corrosion or increased resistance of
E. Bypass Diode Fault (BDF) the busbars or their connections, yellowing or deterioration of
the glass or polymer materials, etc. [40] [41]. The degradation
Bypass diodes are present in almost all commercially that impacts the module’s performance may be detected
available PV modules to protect against reverse bias and through J-V monitoring, but pinpointing the degraded
resultant hot spots and damage to cells from partial shading. module(s) is not possible with this technique [42].
Bypass diodes are connected in reverse polarity between strings To test the effect of accelerated degradation on SSTDR
of PV cells (or modules) and have no effect on the forward bias responses, single cell mini-modules similar to those described
output. Bypass diodes fail as either open (OC) or short (SC) in section II were degraded using damp heat (DH), damp heat
circuits [38]. Analyzing output power can help detect BDF [29], along with potential induced degradation (DH+PID) or
but these are notoriously difficult to detect and locate. There are humidity freeze (HF).
some configurations where we could expect SSTDR to be able A single cell PV mini-module was kept inside a damp heat
to detect and locate BDF (such as when they fail as series OC), (DH) chamber at 85°C and 85% relative humidity for 67 hours.
and others (failing as series SC) where it could be extremely This is only a fraction of the 1000 hours of degradation
difficult if not impossible. This is an area that warrants further specified by the IEC 61215 standard [43]. It is difficult to
investigation. estimate what outdoor exposure the IEC 61215 test represents,
and different module designs will experience very different

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
stress from this exposure [44]. the particular magnitudes we observed are specific to the PV
SSTDR measurements were taken before and after DH systems we measured and the location of the faults. Faults that
degradation and are shown in Fig. 9. The maximum change in are further from the SSTDR will experience additional
normalized reflection is 0.02, which is shown by asterisks in attenuation and a reduced reflection magnitude. The ability of
Fig. 9. Even this small amount of DH degradation introduces a SSTDR to detect and locate faults in PV systems is summarized
measurable change in normalized reflection. In addition to the in Table III.
change in magnitude, a time delay is also seen, which indicates 1

a change in the velocity of propagation (2.32%). This would be


consistent with an aging-induced change in electrical
0.9

permittivity of the materials in the mini-module, which would 0.8

in turn cause a change in the velocity of propagation. 0.7

0.6

Change in Normalized Reflection


Before DH
0.5
After DH
0.2
0.4

0.3
0
0.2
*
Normalized Reflection

0.1
-0.2 *
0
OC CF GF AF SF BRK ADF

-0.4
Type of faults

Fig. 10. A comparative analysis of SSTDR magnitude for different fault


conditions. * Submerging the cable in water and damp heat accelerated aging
* also change the velocity of propagation.
-0.6 * 0.02

40 60 80 100 120 To determine if these faults are detectable in practice, we


Distance [feet] would need to know the accuracy of the SSTDR test system.
Fig. 9. SSTDR responses from mini-modules before and after damp heat For our SSTDR instrument, the maximum variance was found
degradation. as <1% at the end of the leader cable, as well as the magnitude
of normal variations in the PV system (such as the day/night
To check the damp heat and PID effect together, a single-cell variation). In summary, most of the common faults in PV plants
PV mini-module was aged in a DH+PID chamber. Again, this (open circuit faults, connection faults, ground faults, arc faults,
represents only a small amount of accelerated aging, as full shading faults, broken cell/module faults, and accelerated
degradation tests are normally run for at least 1000 hours. degradation faults) could be detected and located using the
Similar to DH aging, a time delay is seen in the SSTDR SSTDR technique.
response, indicative of a change in velocity of propagation
(1.57%) and 2.88% of attenuation. Detailed measurements are ACKNOWLEDGMENT
shown in [19] for this case. This material is based upon work supported by the U.S.
Department of Energy’s Office of Energy Efficiency and
IV. CONCLUSIONS AND DISCUSSIONS Renewable Energy (EERE) under Solar Energy Technologies
Office (SETO) Agreement Number DE-EE0008169. Also, we
The changes in the normalized SSTDR reflection for each of
want to thank Kent Terwilliger and Byron McDanold for their
the tests described above are summarized in Fig. 10. this gives
help with measurements at NREL.
a qualitative measurement of the relative detectability of
various faults in PV power plants using SSTDR. This would
DISCLOSURE
apply to other types of reflectometry with similar bandwidth, as
well. Dr. C.M. Furse is a co-founder of LiveWire Innovation, Inc.
Open circuit and short circuit faults produce the largest which is commercializing SSTDR technology, and therefore
change in reflection magnitude (1.0), while for our tests, has a financial conflict of interest with this company.
accelerated degradation fault causes the least amount of change
(0.02). Other types of faults are in between the open circuit and
REFERENCES
accelerated degradation faults, broken cell/ module faults (0.1),
arc faults (0.3), ground faults (0.15), shading faults (0.12) and [1] D. S. Pillai, F. Blaabjerg, and N. Rajasekar, “A Comparative Evaluation
of Advanced Fault Detection Approaches for PV Systems,” IEEE J.
connection faults (0.7). The larger the magnitude of the Photovolt., vol. 9, no. 2, pp. 513–527, 2019.
reflected signal, the easier the fault is to detect. Detection is [2] E. Garoudja, F. Harrou, Y. Sun, K. Kara, A. Chouder, and S. Silvestre,
needed before location can be done, and faults with larger “Statistical fault detection in photovoltaic systems,” Sol. Energy, vol.
magnitudes are also easier to locate. We should emphasize that 150, pp. 485–499, 2017.

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
[3] R. Hariharan, M. Chakkarapani, G. S. Ilango, and C. Nagamani, “A [25] J. Pern, “Module encapsulation materials, processing and testing
method to detect photovoltaic array faults and partial shading in PV (presentation),” National Renewable Energy Lab.(NREL), Golden, CO
systems,” IEEE J. Photovolt., vol. 6, no. 5, pp. 1278–1285, 2016. (United States), 2008.
[4] M. K. Alam, F. Khan, J. Johnson, and J. Flicker, “A Comprehensive [26] “Aliexpress.com | Alibaba Group,” aliexpress.com. [Online].
Review of Catastrophic Faults in PV Arrays: Types, Detection, and Available: https://www.aliexpress.com/item/Energia-Solar-Direct-
Mitigation Techniques,” IEEE J. Photovolt., vol. 5, no. 3, pp. 982–997, 2016-Promotion-10pcs-High-Efficiency-4-33w-Poly-Solar-Cell-6x6-
May 2015. for-
[5] C. Furse, Y. C. Chung, C. Lo, and P. Pendayala, “A critical comparison Diy/32626393408.html?src=ibdm_d03p0558e02r02&sk=&aff_platfor
of reflectometry methods for location of wiring faults,” Smart Struct. m=&aff_trace_key=&af=&cv=&cn=&dp=. [Accessed: 30-Dec-2018].
Syst., vol. 2, no. 1, pp. 25–46, 2006. [27] L. A. Griffiths, R. Parakh, C. Furse, and B. Baker, “The invisible fray:
[6] C. Furse, Y. C. Chung, R. Dangol, M. Nielsen, G. Mabey, and R. A critical analysis of the use of reflectometry for fray location,” IEEE
Woodward, “Frequency-domain reflectometry for on-board testing of Sens. J., vol. 6, no. 3, pp. 697–706, 2006.
aging aircraft wiring,” IEEE Trans. Electromagn. Compat., vol. 45, no. [28] F. Harrou, Y. Sun, B. Taghezouit, A. Saidi, and M.-E. Hamlati,
2, pp. 306–315, 2003. “Reliable fault detection and diagnosis of photovoltaic systems based
[7] P. Tsai, C. Lo, Y. C. Chung, and C. Furse, “Mixed-signal reflectometer on statistical monitoring approaches,” Renew. Energy, vol. 116, pp. 22–
for location of faults on aging wiring,” IEEE Sens. J., vol. 5, no. 6, pp. 37, 2018.
1479–1482, 2005. [29] N. Gokmen, E. Karatepe, B. Celik, and S. Silvestre, “Simple diagnostic
[8] V. Velayudhan, I. Bzikha, and A. Reineix, “Experimental Detection of approach for determining of faulted PV modules in string based PV
Soft Faults on Cables Using Chaos Time-Domain Reflectometry,” arrays,” Sol. Energy, vol. 86, no. 11, pp. 3364–3377, 2012.
IEEE Trans. Electromagn. Compat., no. 99, pp. 1–7, 2018. [30] A. Umana and A. S. Meliopoulos, “Detection of cell-level fault
[9] C. Lo and C. Furse, “Noise-domain reflectometry for locating wiring conditions within a photovoltaic array system,” in 2016 IEEE/PES
faults,” IEEE Trans. Electromagn. Compat., vol. 47, no. 1, pp. 97–104, Transmission and Distribution Conference and Exposition (T&D),
2005. 2016, pp. 1–5.
[10] P. Smith, C. Furse, and J. Gunther, “Analysis of spread spectrum time [31] L. Chen, S. Li, and X. Wang, “Quickest fault detection in photovoltaic
domain reflectometry for wire fault location,” IEEE Sens. J., vol. 5, no. systems,” IEEE Trans. Smart Grid, vol. 9, no. 3, pp. 1835–1847, 2018.
6, pp. 1469–1478, 2005. [32] N. L. Georgijevic, M. V. Jankovic, S. Srdic, and Z. Radakovic, “The
[11] F. Auzanneau, “Binary time domain reflectometry: a simpler and more detection of series arc fault in photovoltaic systems based on the arc
efficient way of diagnosing defects in wired networks,” in 2018 IEEE current entropy,” IEEE Trans. Power Electron., vol. 31, no. 8, pp.
AUTOTESTCON, 2018, pp. 1–8. 5917–5930, 2016.
[12] M. U. Saleh et al., “Detection and Localization of Disconnections in [33] S. Chen and X. Li, “PV series arc fault recognition under different
PV Strings Using Spread-Spectrum Time-Domain Reflectometry,” working conditions with joint detection method,” in 2016 IEEE 62nd
IEEE J. Photovolt., pp. 1–7, 2019. Holm Conference on Electrical Contacts (Holm), 2016, pp. 25–32.
[13] C. Furse, Y. C. Chung, C. Lo, and P. Pendayala, “A critical comparison [34] H. Zhu, Z. Wang, and R. S. Balog, “Real time arc fault detection in PV
of reflectometry methods for location of wiring faults,” Smart Struct. systems using wavelet decomposition,” in 2016 IEEE 43rd
Syst., vol. 2, no. 1, pp. 25–46, 2006. Photovoltaic Specialists Conference (PVSC), 2016, pp. 1761–1766.
[14] C. Furse, P. Smith, M. Safavi, and C. Lo, “Feasibility of spread [35] W. R. Sekulic and P. F. McNutt, “Evaluating the Incident Energy of
spectrum sensors for location of arcs on live wires,” IEEE Sens. J., vol. Arcs in Photovoltaic DC Systems: Comparison Between Calculated and
5, no. 6, pp. 1445–1450, 2005. Experimental Data,” National Renewable Energy Lab.(NREL),
[15] C. Furse, P. Smith, and M. Diamond, “Feasibility of reflectometry for Golden, CO (United States), 2019.
nondestructive evaluation of prestressed concrete anchors,” IEEE Sens. [36] K.-H. Chao, S.-H. Ho, and M.-H. Wang, “Modeling and fault diagnosis
J., vol. 9, no. 11, pp. 1322–1329, 2009. of a photovoltaic system,” Electr. Power Syst. Res., vol. 78, no. 1, pp.
[16] A. Hanif, Y. Yu, D. Devoto, and F. H. Khan, “A Comprehensive 97–105, 2008.
Review toward the State-of-the-Art in Failure and Lifetime Predictions [37] H. Mekki, A. Mellit, and H. Salhi, “Artificial neural network-based
of Power Electronic Devices,” IEEE Trans. Power Electron., 2018. modelling and fault detection of partial shaded photovoltaic modules,”
[17] S. Roy, M. K. Alam, F. Khan, J. Johnson, and J. Flicker, “An irradiance- Simul. Model. Pract. Theory, vol. 67, pp. 1–13, 2016.
independent, robust ground-fault detection scheme for PV arrays based [38] D. S. Pillai and N. Rajasekar, “A comprehensive review on protection
on spread spectrum time-domain reflectometry (SSTDR),” IEEE Trans. challenges and fault diagnosis in PV systems,” Renew. Sustain. Energy
Power Electron., vol. 33, no. 8, pp. 7046–7057, 2018. Rev., vol. 91, pp. 18–40, 2018.
[18] M. K. Alam, F. H. Khan, J. Johnson, and J. Flicker, “PV arc-fault [39] M. A. Munoz, M. C. Alonso-García, N. Vela, and F. Chenlo, “Early
detection using spread spectrum time domain reflectometry (SSTDR),” degradation of silicon PV modules and guaranty conditions,” Sol.
in Energy Conversion Congress and Exposition (ECCE), 2014 IEEE, Energy, vol. 85, no. 9, pp. 2264–2274, 2011.
2014, pp. 3294–3300. [40] Z. Xiong, T. M. Walsh, and A. G. Aberle, “PV module durability testing
[19] Mashad Uddin Saleh, “Photovoltaic String Monitoring Using Spread under high voltage biased damp heat conditions,” Energy Procedia, vol.
Spectrum Time Domain Reflectometry (Expected Publication Date: 8, pp. 384–389, 2011.
May 2020),” University of Utah, Salt Lake City, 2020. [41] C. Ferrara and D. Philipp, “Why do PV modules fail?,” Energy
[20] P. Smith, C. Furse, and J. Gunther, “Analysis of spread spectrum time Procedia, vol. 15, pp. 379–387, 2012.
domain reflectometry for wire fault location,” IEEE Sens. J., vol. 5, no. [42] Y. Zhao, B. Lehman, R. Ball, J. Mosesian, and J.-F. de Palma, “Outlier
6, pp. 1469–1478, 2005. detection rules for fault detection in solar photovoltaic arrays,” in 2013
[21] “‘Spec 5851 SunGen XLPE, Photovoltaic Wire, 2000 V, UL Type PV Twenty-Eighth Annual IEEE Applied Power Electronics Conference
or 1000 V, CSA RPVU90, Single Conductor, Copper.’ [Online]. and Exposition (APEC), 2013, pp. 2913–2920.
Available: http://dtsheet.com/doc/1724396/spec-5851-sungen-xlpe-- [43] I. E. C. Standard, “61215," Crystalline silicon terrestrial photovoltaic
photovoltaic-wire--2000-v--ul-type... [Accessed: 23-Aug-2018].” . (PV) modules,” Des. Qualif. Type Approv., vol. 2.
[22] N. K. T. Jayakumar et al., “Post-Processing for Improved Accuracy and [44] J. H. Wohlgemuth and S. Kurtz, “Using accelerated testing to predict
Resolution of Spread Spectrum Time Domain Reflectometry module reliability,” in Photovoltaic Specialists Conference (PVSC),
(SSTDR),” IEEE Sens. Lett., 2019. 2011 37th IEEE, 2011, pp. 003601–003605.
[23] “‘Live Cable Fault Detection by LiveWire Innovation,’ LiveWire
Innovation. [Online]. Available: http://www.livewireinnovation.com/.
[Accessed: 07-May-2018].” .
[24] C. Furse, P. Smith, C. Lo, Y. C. Chung, P. Pendayala, and K. Nagoti,
“Spread spectrum sensors for critical fault location on live wire
networks,” Struct. Control Health Monit. Off. J. Int. Assoc. Struct.
Control Monit. Eur. Assoc. Control Struct., vol. 12, no. 3–4, pp. 257–
267, 2005.

Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.

You might also like