06745063 ground fault and insulation degradation detection and localization in
TRANSCRIPT
-
8/11/2019 06745063 Ground Fault and Insulation Degradation Detection and Localization in
1/5
Ground Fault and Insulation Degradation Detection and Localization in
PV Plants
Mohammed S. Agamy1, Maja Harfman-Todorovic
2, and Ahmed Elasser
2
1School of Engineering, University of British Columbia, Kelowna BC, Canada
2
General Electric Global Research Center, Niskayuna NY, USAAbstract
GroundfaultsareamajorhazardinPVnetworks.Early
detection
as
well
as
location
of
these
ground
faults
can
provide higher plant availability as well as protection forpersonnel
and
equipment.
Since
the
occurrence
of
a
ground
fault
or
even
insulation
degradation
changes
the
equivalent
ac
impedance(RLC)ofthedistributionnetwork,transientanalysisof
the
currents
at
string
or
array
combiners
can
be
used
to
detect
and
locate
the
fault
or
the
beginning
of
insulation
degradation.
In
this
paper,
the
analysis
of
ground
faults
and
insulation
degradation
and
their
effect
on
the
transient
waveforms
is
presented
for
grounded
and
floating
PV
networks.
The
current
spectral
analysis
can
clearly
distinguish,
identify
and
locate
faults
intheplant.
Index
Terms
PV
arrays,
Ground
faults,
Insulation
degradation,
Transient
analysis,
Fast
Fourier
Transform.
I. INTRODUCTION
Ground faults in the dc collection network of a photovoltaic
(PV) array can present significant hazards to both operators
and equipment, and if left undetected can lead to severe fire
damage [1]. Therefore, ground fault protection circuitry is
added on the dc side of the PV inverters to disconnect the
system when ground faults are detected. However most
classical ground fault detection methods rely on directly
measuring ground currents at the central inverter [2, 3], which
can be sufficient to detect the occurrence of the fault, but
cannot determine the location of the fault. Locating the faultposes a significant challenge in large plants that include
thousands of strings, which leads to longer down time of the
plant. With the increasing trend of employing string
combiners that have more monitoring capabilities, current
sensors can be placed to monitor each string current. In this
paper, methods of detecting and locating ground faults as well
as insulation degradation by analyzing the string current are
investigated. The fault/degradation detection is based on
spectral analysis of the string currents, which can be used to
detect network irregularities, due to the change of the
equivalent circuit seen by the string. This approach has been
used for non-intrusive diagnostics of electrical machines [4-6]
and is now being extended to PV arrays. Successfulidentification of fault location allows the isolation of the faulty
line and quickly re-connect the rest of the system, which
increases the overall plant availability and the energy yield.
II. CURRENT RESPONSE DUE TO GROUND FAULTS AND
INSULATION DEGRADATION
Solar power plants are laid out with large numbers of strings
of PV modules connected in parallel to feed a grid tied
inverter through long stretches of cables. Based on equivalent
circuits of PV modules, cables and input of the inverter, the dc
network can be represented by an equivalent resistive,
inductive/capacitive (RLC) network. Ground faults cause
transients in the different current paths based on the equivalent
RLC network seen by each path. Therefore, analyzing current
measurements in the different paths of the dc-network can
indicate the occurrence of a fault as well as determine the
location of the fault based on the measured transient current
signature. In the following sub-sections, the analysis of current
signatures of an example PV plant is presented. Different
combinations of grounding of the inverter and PV array are
studied to investigate the effect on the current signature during
faults and whether it can be used to locate the fault in the dc
network. Furthermore, the current transient at startup is also
studied to detect and locate the degradation in the dc cables.
The cables connecting the PV strings to the inverter are
modeled using their pi- equivalent circuit, with each cable
represented by 10 cascaded pi sections. Cables with 600V
insulation are used for the analysis in this paper. PV module
output capacitance and parasitic capacitances to ground are
also included in the analysis as shown in fig. 1.
The analysis presented in this paper relies on detecting
variations in the differential current to detect ground faults or
insulation degradation rather than directly measuring groundpath current, which simplifies the sensor requirements, as the
string current sensors in smart combiners or current feedback
sensors in systems with distributed power converters can be
used to detect ground fault conditions. Since faults lead to a
change in the equivalent RLC circuit, the equivalent resonant
frequencies in the network change and consequently the
r l
c/2 c/2
-
8/11/2019 06745063 Ground Fault and Insulation Degradation Detection and Localization in
2/5
frequency spectrum of the measured differential current leads
to both detecting the occurrence of a fault as well as
determining its location in the solar farm.
Current sensing can be made at either the PV array side or
the inverter side, i.e. either before or after the fault. In both
cases the change in current pattern indicates whether or not a
fault has occurred.
A.
Grounded PV Arrays and Grounded Inverter
In grounded systems, any ground fault will result in creating
two paths for fault current, one from the inverter side of the
cable and the other from the PV string side. Fig. 2 shows the
system layout as well as the harmonic spectrum of the fault
current at different locations of the dc cable (at 10% of the
cable length away from the PV array side and at 50% of the
cable length from the array).
The current spectrum shows differences in amplitude and
frequencies for the two different fault locations, which can be
used to detect and isolate the faulty line
B.
Floated PV Array and Floated Inverter
In case of a floated array and a floated inverter as shown inFig. 3, locating a ground fault becomes more complicated as a
first fault does not provide a dc-current path and classical
methods rely on the occurrence of a second fault to close the
loop and indicate the ground fault problem and shutdown the
plant. The measurement of transients can indicate the
occurrence of the first ground fault, as shown in Fig. 4, which
can be resolved and thus avoiding the occurrence of any
damage as a cause of the second fault.
Furthermore, the occurrence of a second fault can also be
identifiable and its location determined as shown in Fig. 5,
where the current signature is different if the second fault
occurs on the same line or different lines
C.
Floated PV Array and Grounded Inverter
System layout and spectral analysis comparisons for
currents in both normal and faulty lines are shown in fig. 6 for
the architecture with floated strings. The frequency spectra for
faults occurring at different locations along the line are
distinct, which facilitates the isolation of the compromised
zone of the solar farm.
D.
Detection of Insulation Degradation
Another detectable state is the insulation degradation. Since
degradation is more of a steady state condition rather than a
sharp transient, it can be detected by comparison of currents
from healthy and degraded lines during startup transient. Even
if the starting currents in different strings are significantly
different, degradation in a certain zone/cable of the dc-
Fig. 2 (a) Example PV plant with grounded strings and grounded
inverter, (b) Current spectrum of faulted line with the fault at 10%
(Green) and 50% (Blue) of the line length away from the PV array
-
8/11/2019 06745063 Ground Fault and Insulation Degradation Detection and Localization in
3/5
network can be located by calculating the differences in
currents from different string combiners. Fig. 7 and fig. 8
show examples of the start-up current transient with different
degrees of dc cable insulation degradation and the associated
frequency response. In fig. 7, the degradation is severe and the
equivalent impedance to ground is modeled as 1k, while in
fig. 8, the degradation is less significant and modeled through
100k impedance to ground. The frequency spectra are still
distinct, which indicates the capability of providing early
warning signals to maintain degrading elements.
-
8/11/2019 06745063 Ground Fault and Insulation Degradation Detection and Localization in
4/5
Fault and degradation detection on the dc network depends
only on the transient response of current and/or voltage since
the dc steady state decays to zero as seen in all the time
response waveforms shown in the previous figures. Therefore,
data acquisition and analysis should be performed on a
moving time window. The width of the window depends on
the system layout and components, which determine the time
constants and resonant frequencies of the system. Spectral
analysis is performed for the waveforms in each individual
window.
III.SYSTEM ARCHITECTURE AND FAULT DETECTION
ALGORITHM
In order to locate faults on the dc-network, string current
sensors are required at each string combiner box. The
resolution of the current sensors will determine the types and
severity of faults and/or degradation that can be detected. The
current measurements are then communicated to a central
control unit that analyzes the signals from all combiner boxes
to determine if a fault exists and where it is located based on
the current transients as described in prior sections. Fig. 9
shows an example of such a system, where the communication
can be either hard wired or wireless. The central controller
then performs waveform analysis to determine the type and
location of the fault/degradation. The analysis can be a
spectral analysis as shown in the previous examples or any
other identification and classification method using neural
networks or wavelet analyses.
Since this analysis is based on comparisons between normal
and fault conditions the system should have a saved record of
the expected normal waveforms as a reference. Furthermore,
offline simulations should be made to determine the expected
dominant frequencies that should be monitored since these
frequencies are dependent on type of modules, cables, theircross-section and length and the overall plant layout.
The moving window during which the current waveform is
analyzed is also dependent on the system dominant
frequencies as well as measurement and control bandwidth
and should be specified during the commissioning of the plant.
Fig. 10 shows a basic summary of the proposed fault location
algorithm.
IV.CONCLUSION
A method to locate ground faults in the dc network in large
PV plants is presented. Only the measurement of differential
currents at the string combiners is used. These measurements
are then communicated to a central processor to analyze and
compare the different waveforms to the expected operating
conditions. Since there is an increased interest in smart string
combiners, this method can easily utilize them to provide
better protection, maintenance and diagnostic tools in the PV
plant. The current signatures are distinct for different fault
locations and for different types of faults and from other
transients, which prevents the occurrence of false trips due to
other system transients. The method can be extended to any
dc-distribution system. It should be noted that the type of faultor degradation detected depends on the measurement
resolution and bandwidth.
Fig. 10 A conceptual flowchart of the basic monitoring and fault identification
-
8/11/2019 06745063 Ground Fault and Insulation Degradation Detection and Localization in
5/5
REFERENCES
[1] B. Brooks, The Bakersfield Fire, SolarPro, February/March 2011, pp.
62-70.
[2] W. Bower & J. Wiles, Investigation of Ground-Fault Protection
Devices for Photovotaic Power Systems Applications, Proceedings of
the IEEE Photovoltaic Specialists Conference (PVSC) 2000, pp. 1378-
1383.
[3] D. Stellbogen, Use of PV Circuit Simulation for Fault Detection in PV
Array Fields, Proceedings of the IEEE Photovoltaics Specialists
Conference (PVSC), 1993, pp. 1302-1307.
[4] U. Orji, Z. Remscrim, C. Laughman, S. Leeb, W. Wichakool, C.
Schantz & Robert Cox, Fault Detection and Diagnostics for Non-
Intrusive Monitoring using Motor Harmonics, Proceedings of the IEEE
Applied Power Electronics Conference and Expo (APEC) 2010, pp.
1547-1554.
[5] J. Cusido, L. Romeral, J. Ortega, J. Rosero & A. Espinosa, Fault
Detection in Induction Machines Using Power Spectral Density in
Wavelet Decomposition, IEEE Trans. on Industrial Electronics, Vol.
55, No. 2, February 2008, pp. 633-643.
[6] P. Karlsson & J. Svensson, Fault Detection and Clearance in DC
Distributed Power Systems, Proceedings of Nordic Workshop on
Power and Industrial Electronics, August 2002.
Disclaimer: