You are on page 1of 4

Fault finding in a 1 MW photovoltaic plant by reflectometry

L. Schirone, F.P. Califano


Terza Universita di Roma, Via Segre, 2 - 00146 Roma - Italy - FAX+39 6 5561373

U. Moschella, U. Rocca,
ANIT, Via Pacinotli, 20 - 16151 Genova - Italy- F A x i 3 9 10413012

ABSTRACT

Time Domain Reflectometry has been used for the input step negative in our system, in the case of an
diagnostic monitoring of a large size photovottaic (PV) open/short circuit a decreasinghncreasing ramp adds
plant in operating conditions. By analysing the to the voltage measured in normal conditions (see Fig.
waveforms obtained when a step-voltage excitation is 1 ).
propagated down the electrical line connecting the PV During the first microsecond of the waveforms in
generators to the inverter, we were able to detect, Fig. 1 voltage evolution depends on the connection
identify and localise the most common fault conditions, details between PV field and the measurement point:
such as breaks of the circuit, insulation defects, wiring bending of the wires, proximity to other conductors,
anomalies. Best compromise between analysis time etc.. After this time, a small slope's modification is
and precision was achieved by testing groups of three caused by crossing sectioning switches. The sign of
paralleled strings (panels): in this way we were able to waveform's slope changes corresponding to the
detect in a short time whether a fault was present, while reflected signal produced when the pulse enters
its position was separately determined with a good photovoltaic modules and meets with an abrupt
precision by repeating the test on the single strings. modification of propagation conditions, due to the lower
The whole 1 MW PV plant was tested in a couple of characteristic impedance of the PV modules [ l ] . In
days. Finally, we discuss some economical aspects of normal operating conditions, the next change in slope
this fault finding technique. is caused by discontinuity of propagation
characteristics at the end of PV string and occurs after
a delay Atmax depending on the length of electrical
I NTR OD U CTl ON path inside the modules. In the presence of a fault, a
reflected wave is produced before Atmax , after a delay
Fault finding is a major problem in the assembling
and the maintenance of a large scale photovoltaic (PV) depending on fault position.
field, due to the time requested for manual searching.
Time Domain Reflectometry (TDR) has proven to be an
effective approach to determine both the type and the
position of a fault in many kinds of electrical
transmission lines and we already demonstrated its
application in a single photovoltaic string [l].
We report the results of the tests performed in
operating conditions on the 1 MW PV plant ELIO1,
-70

I
located in Vasto, Italy: typical fault conditions are
discussed, by illustrating the corresponding registered h

waveforms. Some economical aspects of this fault L


finding technique are discussed in conclusion. 5 -80 wted at 15th module
-0
4-

5
FUNDAMENTALS OF TIME DOMAIN -90
REFLECTOMETRY
Opened at 15th module
Reflectometric systems can be regarded as
closed loop radar's: a step-voltage excitation
propagates down the electrical line under test, while t h e
voltage is monitored by an oscilloscope. The signal
propagates down the transmission line consisting of
the PV modules and the series-connecting wires and is
Fig. 1. Typical waveforms recorded when TDR is
partially reflected when a faulty module is reached,
producing in the waveform a delayed replica of input performed on a string in the presence of a fauk on 15th
pulse. Measured delay allows us to calculate the module: in the case of an open/short circuit a
distance travelled by the signal and so provides an decreasinghncreasing ramp adds to the voltage
indication of the fault position [I]. The type of a fault is measured in normal condition. The string was about
revealed by the sign of the reflected signal [2]: being 150 meters far from observation point.

846

CH3365-4/94/0000-0846 $4.00 0 1994 IEEE First WCPEC; Dec. 5-9, 1994; Hawaii
EXPERl MENTAL S ET-UP The recorded waveforms could also give
information on other wiring anomalies, such as the
The PV plant wiring scheme is based on strings connection of a module in inverted polarity, as shown in
consisting of 28 series-connected 47 Wpk modules. Fig. 3. The effect is rather small, and only could be
The strings are paralleled in 3-elements groups (panels) observed if TDR is performed on a loaded string.
and 4 panels are grouped in a field. The 64 fields are
separately connected to one of two inverters which
interface the plant to power grid.
TDR set-up is shown in Fig. 2. A pulsed voltage -75 I I I I

generator is applied at the inverter side of the field-


\
connecting wires and drives string's anode by negative .!*
pulses with 70 V amplitude, 40 ns rise time, 5 ps
-80
duration and 1000 Hz repetition rate. The generator
(output impedance Zo = 50 Q) is interfaced to the two
A

L
conductor line (characteristic impedance Ze = 300 Q) 0
m -85
by an impedance-matching resistor (RM = 270 Q) and -
c

by a blocking capacitor (100 nF). 8


-90

Oscilloscooe m
-95
0 1 2 3 4 5
Time (ps)

Generator Fig. 3. Waveforms recorded when TDR is


performed on a string containing a module connected
with inverted polarity.

Figures 4 and 5 show the modifications occurring


in voltage waveforms obtained for a string when an
(ideal) fault is moved along the string: in the various
fault positions, both the delay of the reflected pulse
28 and its amplitude are changed. Delay effect is more
RL Modules evident when the fault occurs in the first 15 modules,
while amplitude effect is more evident when it occurs
deep inside the string.

v
N
Strings -70

Fig. 2. Experimental set-up.


-75
In the same point a storage oscilloscope records
the voltage signal by means of a high-impedance h

probe. A standard instrumentation link (GPIB) z


interfaces the oscilloscope to a computer, allowing us & -80
to perform numerical analysis of the measured -d
waveforms. s -85

RESULTS
-90
Fig. 1 shows typical waveforms for a string 1 2 3 4
located about 150 meters far from the point of Time (ps)
observation. The lowest curve reveals a break in the
circuit after ldhmodule. Before repairing the fault, the Fig. 4. Waveforms recorded when TDR is
same point was deliberately short-circuited to ground, performed on a string in the presence of an open circuit
obtaining the waveform shown for comparison in the at different positions.
figure. Also it is reported the waveform obtained after
reparation, when the string was normally operating.

847
be based on comparison approach better than on
analysis of single waveforms: any anomalous feature in
a string can be easily detected by comparison with the
waveforms obtained for other strings under the same
experimental conditions (see Fig. 7).
Improved sensitivity achieved by comparison
method allows to detect a fault even if TDR is performed
-70
on several paralleled strings. Whenever fault detection
occurs, TDR is repeated on single strings, determining
L fault position with a good precision. In order to reduce
& -75 the time required to scan the whole plant, it is important
-Q
CI
0
to keep as high as possible the number of paralleled
strings. On the other hand, when TDR is performed on a
>
-80

-85 :
1
I
2
I
3
+
4 -60

Time (ps)

Fig. 5. Waveforms recorded when TDR is -70


performed on a string in the presence of a short circuit
to ground at different positions. h

L*. -80

The attenuation taking place when the pulse is -*P


0
propagating in the string is influenced by environmental > -90
conditions, like temperature or irradiation level [l]. Fig.
6 shows the modifications occurring in the waveforms
for a string under an irradiation level ranging from 200 to Opened at 15th module I
800 W/m2. The short circuited, normally operating and I I 1 I I
0 1 2 3 4 5
opened strings were nearly located at the same
distance from the observation point. Minor differences Time (ps)
in the waveforms arise from connection details.
Fig. 7a. Waveforms recorded when TDR is
performed on the strings in a panel in the presence of
an open circuit at 1!jthmodule.
-75 I I I I I

'j I
-80 - - -60

L
- -70

5
-90 -
-95 I I I I -90
0 1 2 3 4 5
Time (p)

-1 00 I I
Fig. 6. Irradiation level effect on the waveforms I I I I I I

observed for strings under different operating 0 1 2 3 4 5


conditions (short circuit to ground, normal, open Time (p)
circuit): upper/lower set of curves were recorded under
200/800 W/m2. Fig. 7b. Waveforms recorded when TDR is
performed on the strings in a panel in the presence of a
Former discussion reports some factors affecting short circuit to ground at 15th module.
waveforms shape and illustrate why fault finding could

848
I I I I I
-60 !. -75

-70 - -80

-L h

L
(U -80 - ; -85
m
-0
Y
i -
L

> 3 module inside


-90

-901
-100 I
0 1
I I
2
I
3
I
4 5
-95 ?
0 1 2 3 4 5
Time (ps) Time (ps)

Fig. 8. Waveforms recorded when TDR is Fig. 9a. Waveforms recorded when TDR is
performed on a field consisting of 12 paralleled strings performed on the four panels in the same field in the
in the presence of an open circuit, a short circuit to presence of an open circuit.
ground and in normally operating conditions.

field (12 paralleled strings), in the different operating -75


conditions (normal, short circuit to ground, and open
circuit) the differences observed in resulting
waveforms are too little to allow pratical fault detection
(see Fig. 8). -80
Best compromise between speed and sensitivity A

was obtained by performing TDR on groups of three L


paralleled strings, corresponding to the panels, and 5 -85
comparing each other the waveforms for the four -0
c

panels in each field (see Fig. 9). >


We are recording data under several -90
environmental conditions, in order to synthesise
general criteria allowing us to perform direct fault
detection, without comparison among waveforms. I
-95 { I I I I I
0 1 2 3 4 5

CONCLUSIONS Time (w)

TDR has proven effective for fault finding in large Fig. 9b. Waveforms recorded when TDR is
size PV plants: after detection of breaks in the circuit, performed on the four panels in the same field in the
insulation defects and other wiring anomalies, they presence of a short circuit to ground (each panel
could be localised in a few minutes. The whole 1 MW PV consists of three strings).
plant was tested in two days, revealing five breaks in
modules bus bars and one short circuit to ground in a
switch board. During the test one of two inverters was REFERENCES
switched off, in order to keep interference level as low
as possible. Lost power and man power have a cost, [l] L. Schirone, F.P. Califano, M. Pastena, Progress
which can be balanced in a few months by retrieved in Photovoltaics, 2,1994, pp. 35-44
productivity, amounting to about 10 kW and
corresponding to 1% of plant’s target power. On the [2] S. Ramo, J.R. Winnery and T. Van Duzer, Fields
other hand, the cost of TDR system is neglegible, as and waves in Communication Electronics, Wiley,
the instruments are normally available in our New York, 1965
laboratories.

849

You might also like