AU2016201446A1

AU2016201446A1 - Method and apparatus for recognizing an electric arc

Info

Publication number: AU2016201446A1
Application number: AU2016201446A
Authority: AU
Inventors: Vincent Teichmann; Martin Weiss; Michael Wendl
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2015-03-11
Filing date: 2016-03-04
Publication date: 2016-09-29
Also published as: DE102015204376A1; ITUA20161340A1

Abstract

Abstract The invention relates to a method (200) for detecting an electric arc (102) in a direct-current string (104) of a photovoltaic system (106), with a processing step (202), in which a current-signal (112) acquired from a 5 direct-current string (104) is processed, using a band-pass filter (116), to obtain a detection-signal (114), and with a monitoring step (204), in which the detection-signal (114) is monitored, to detect an electric arc (102) an electric arc (102) being detected when the level of the detection-signal (114) is greater than a threshold value (118, 702) for more than a 10 minimum period. (Fig. 1) 2598188vl Fig. 1 -134 102 -28 136 1307 112 118 116s 108 122 114 110 120 Fig. 2

Description

Specification

Method and Apparatus for Recognising an Electric Arc

Prior Art

This invention relates to a method for detecting an electric arc, and to a corresponding apparatus and a corresponding computer program.

Disclosure of the Invention

If a line carrying direct current in a photovoltaic system is disconnected under load, due to a fault for example, an electric arc may occur and may cause great damage. Since a DC arc will not necessarily extinguish itself, the arc’s ignition and burning needs to be detected, in order to actively interrupt the arc’s circuit.

Disclosure of the Invention

With this background, the approach proposed here provides a method for detecting an electric arc in a direct current string of a photovoltaic system, an apparatus employing this method, and finally, a corresponding computer program, in accordance with the main claims. Advantageous further developments will emerge from the respective dependent claims and from the description below.

An arc causes unwanted signals in the DC side of a photovoltaic system. These unwanted signals caused by an electric arc are distinguishable from interference, and noise signals coupled in from the outside, due to their broad bandwidth.

By band-pass filtering, it is possible to isolate a particular frequency band from the total spectrum of all the frequencies on the direct-current side. In particular, the frequency band here should lie outside of the frequency ranges in which interference is to be expected, e.g. in the range of the pulse frequencies of the inverter or the power supply network frequency.

When the signal intensity in the isolated frequency band is greater than an expected intensity, then an arc can be recognised in the photovoltaic system. A method for detecting an arc in a DC string of a photovoltaic system is proposed, said method having the following steps: - processing a current-signal acquired from the direct current string, to obtain a detection-signal, said processing being performed using a bandpass filter, particularly a broad-band band-pass filter; and - monitoring the detection-signal, to detect an electric arc, an arc being detected when the level of the detection-signal is greater than a threshold value for more than a minimum period.

An electric arc may be understood as being an electric discharge in a gas, between two electrical conductors at different electric potentials. The arc is self-sustaining. A DC string with direct current flowing through it can be a part of the photovoltaic system. In particular, the DC string comprises the photovoltaic modules, the lines between the photovoltaic modules, the lines to the inverter, and the side of the inverter in which direct current flows. A current-signal represents a flow of current in the DC string. A section of a frequency spectrum of the current-signal can pass through the band-pass filter. The level of the detection-signal may be the signal strength or amplitude of the detection-signal.

The method may comprise an adjusting-step, in which the threshold value is adjusted, using the detection-signal. The threshold value can be adjusted proportionately to the background noise of the detection-signal, said proportion being predetermined. The background noise may represent a natural frequency spectrum within the DC string. The background noise may be caused by coupled electromagnetic waves such as radio waves or microwaves. Depending on the location, the background noise may have a higher or lower level. By adjusting the threshold value to the background noise, misrecognition of the arc can be prevented. Furthermore, detection accuracy can be improved.

The arc can be detected using at least one additional threshold value and minimum period of time. The arc can be detected if the detection-signal is greater than the threshold value for more than the minimum period of time. Alternatively or additionally, the arc can be detected if the detection-signal is greater than the additional threshold value for more than the additional minimum period of time. Upon ignition of the arc, the level of the detection-signal may have a peak value for an extremely short time. The time during which the level is far above the threshold value may be shorter than the minimum period. The additional threshold value may be higher than the threshold value, while the additional minimum period is distinctly shorter than the minimum period. As a result, the arc can be detected straight away when it ignites.

The current-signal can also be further processed using a smoothing algorithm, to obtain the detection-signal. Smoothing enables detection-accuracy to be increased.

The current-signal can be averaged quadratically to obtain the detection-signal. The quadratic mean, i.e. the root mean square, may be determined over a sliding interval of time, making it possible to avoid outliers in the detection-signal.

The current-signal can be processed digitally. This makes it possible to perform arc-detection within an inverter control unit, without additional hardware. The digital-processing parameters can be easily adapted to environmental conditions.

The method may include a step of acquiring the current-signal. The current-signal can be sampled from the DC part of an inverter forming part of the photovoltaic system, at a sampling frequency of between 2 kHz and 20 kHz. The current-signal can be acquired with little cost or effort and with a relatively low sampling frequency. Thus, little data-processing capacity is needed to detect arcs.

The approach presented here also provides an apparatus adapted to performing, controlling, and/or implementing the steps of a variant of the method presented here, in corresponding equipment. This variant embodiment of the invention, in the form of an apparatus, likewise enables the objective of the invention to be achieved quickly and efficiently.

An apparatus may be understood here as being an item of electrical equipment that processes sensor signals, and outputs control and/or data signals as a function thereof. The apparatus may have an interface implemented as hardware and/or software. When implemented in hardware, the interfaces may for example be part of a so-called system ASIC containing all sorts of functions of the apparatus. However, the interfaces may also be separate, integrated circuits, or may consist at least partly of discrete components. When implemented in software, the interfaces may be software modules that are provided, along with other software modules, on a microcontroller, for example.

Also advantageous is a computer program product or computer program with program code that can be stored on a machine-readable storage device or medium — such as a semiconductor memory, hard disk, or optical storage — and that serves to perform, implement, and/or control the steps of the method according to one of the forms of implementation described above, particularly when the program-product or program is executed on a computer or an apparatus.

The approach presented here will be described in more detail below through examples, which are illustrated in the accompanying drawings. In the drawings:

Fig. 1 is a block circuit diagram of an apparatus for detecting an electric arc in a DC string of a photovoltaic system, in an example of an embodiment of the invention;

Fig. 2 is a flowchart of a method for detecting an electric arc in a DC string of a photovoltaic system, in an example of an embodiment of the invention;

Fig. 3 represents a photovoltaic system with an apparatus for detecting an electric arc, in an example of an embodiment of the invention;

Fig. 4 represents a frequency spectrum of a current-signal, in an example of an embodiment of the invention;

Fig. 5 represents a frequency spectrum of a detection-signal, in an example of an embodiment of the invention;

Fig. 6 shows fluctuations in current, over time, due to an electric arc, in an example of an embodiment of the invention;

Fig. 7 represents the detection of an electric arc, in an example of an embodiment of the invention; and

Fig. 8 is a program flowchart for a computer program for detecting an electric arc, in an example of an embodiment of the invention.

In the following description of favourable embodiments of the present invention, those elements, in the various Figures, that are similar in function will be given the same or similar reference numbers, and will only be described once.

Fig. 1 is a block circuit diagram of an apparatus 100 for detecting an electric arc 102 in a DC string 104 of a photovoltaic system 106, in an example of an embodiment of the present invention. The apparatus 100 has a processing device 108 and a monitoring device 110. The processing device 108 is adapted to processing a current-signal 112 detected in the DC string 104. The processing, performed in the processing device 108, produces a detection-signal 114. The current-signal 112 is processed, using at least a broadband band-pass filter 116, to obtain the detection-signal 114. The monitoring device 110 is adapted to monitoring the detection-signal 114 so as to detect the arc 102. The arc 102 is detected when the level of the detection-signal 114 is greater than a threshold value 118 for more than a minimum period of time. When the arc is detected 102, an arc signal 120 is outputted.

In one example, the detecting apparatus 100 has an adjusting device 122, to adjust the threshold value 118. In the adjustment device 122, the level of the background noise of the detection-signal 114 is determined. Based on this level, the threshold value 118 is adjusted. In particular, the threshold value is set proportionately to the level of the background noise, said proportion being predetermined.

In the monitoring device 110, more than one threshold value 118 may be used for detecting the arc 102, with each threshold value 118 having its own minimum period of time. In that case, the arc can be detected when, for example, the level of the detection-signal 114 is greater than a first threshold value 118 for longer than a first minimum period; or when the level is greater than a second threshold value 118 for longer than a second minimum period.

In the processing device 108, the current-signal 112 may be smoothed with a smoothing algorithm 124, to obtain the detection-signal 114. For example, the smoothing algorithm 124 may map the quadratic averaging over a defined interval of time.

The signal processing done in the processing device 108 and the monitoring device 110 can be performed digitally. For example, the process and monitoring can be done by executing program-code on a processor.

The photovoltaic system 106 has at least one photovoltaic module 126, connected — by a direct-current and direct-voltage line 128 — to the DC side 130 of an inverter 132. Together, the at least one photovoltaic module 126, the DC line 128, and the DC side 130, form the DC string 104 of the photovoltaic system 106. The inverter 132 also has an AC side 134, which provides AC current 136 for further use.

In other words, Fig. 1 shows an apparatus 100 for arc detection based on low-frequency interference signals in the string current.

Photovoltaic systems 108 generate electric power in the form of direct current, which is converted into alternating current 136 in the inverter 132. Due to the high currents of up to 10 A in a string, together with the high voltages of 300 V to 1000 V, there exists, on the DC side 104, a latent fire risk from arc faults 102 with high thermal energy that occur as soon as a serial connection is interrupted under load. A technical solution for detecting such arcing faults 102 is presented here.

Upon detection, suitable extinguishing measures can be performed by the inverter 132. The photovoltaic system can, for example, be disconnected from the power supply network and the inverter 132.

An arc is detected by detecting low-frequency interference caused by the arc 102, using the current-signal 112 measured in the inverter 132, with typical sampling rates ranging from 2 to 20 kS/s.

For this, the following are performed in the apparatus 100: digital filtering 116 of the current-signal 112 in the low frequency spectrum, smoothing 124 of the filtered signal, and detection 110 of a threshold value being exceeded; and automatic threshold-value adjustment 122, in order to optimise the algorithm with respect to sensitivity and detection 110. If the detection-signal 114 continually exceeds a settable threshold value th 118 for a settable minimum period t_min, an arc 102 is reported.

In one example of an embodiment of the invention, detection 110 is performed using the existing hardware of the inverter 132, and requires no additional pre-processing of the current-signal 112 in hardware. Detection 110 runs continuously, evaluating and analysing the current-signal 112 at given time intervals, in order to promptly report the occurrence of an arc 102.

The algorithm presented here can be applied to any PV system 106, of any size. Configuration to a system is not required: the specific parameters are established by an optimisation method or are calculated as a function of the current-signal 112.

Unlike with other methods, no additional hardware is needed for detection. Moreover, using digital filters 116 does away with the need for computing-intensive calculation based on transformation into the frequency domain. Compared with an analogue bandpass filter, the use of digital filters 116 makes it possible to alter the evaluated frequency range by subsequently adjusting the filter parameters.

Fig. 2 is a flow chart of a method 200 for detecting an electric arc in a DC string of a photovoltaic system, in an example of an embodiment of the present invention. The method 200 may be performed on an apparatus such as the one shown in Fig. 1, for example. The method 200 has a processing step 202 and a monitoring step 204. In the processing step 202, a current-signal detected in the DC string of the photovoltaic system is processed, to obtain a detection-signal. This is done by processing the current-signal with a broadband bandpass filter, in order to obtain the detection-signal. In the monitoring step 204, the detection-signal is monitored, to detect an electric arc. An arc is detected when the level of the detection-signal is greater than a threshold value for more than a minimum period.

In an example of an embodiment of the invention, the method 200 has an adjustment step 206, in which the threshold value is adjusted using the detection-signal. The threshold is adjusted in predetermined proportion to the background noise of the detection-signal.

In an example of an embodiment of the invention, the method 200 includes an acquisition step 208, in which the current-signal is sampled from the DC part of an inverter in the photovoltaic system, with a sampling frequency of between 2 kHz and 20 kHz.

Fig. 3 is a diagram of a photovoltaic system 106 that has an apparatus 100 for detecting an electric arc 102, in an example of an embodiment of the present invention. Here, the current-signal 112 is picked up, by means of an inductive current-sensor 300, from the DC string 104. To detect arcing, it is also possible to monitor the power of high-frequency harmonics in the current: when an electric arc occurs, these increase almost in proportion to the power of the arc. A parallel resonant circuit, for example, may be used here. By means of a toroidal core choke, the AC component of the string current is separated out, with the choke connected in parallel to a capacitor so as to form a parallel resonant circuit. This resonant circuit acts as a band-pass filter with relatively weak rising edges, and with the voltage in the resonant circuit rising as soon as the circuit is excited with a signal near the resonant frequency. The broadband interference spectrum of an arc is sufficient to cause the resonator’s upswing. With subsequent rectification, amplification, and smoothing of the resonator voltage, an analogue DC voltage signal is generated, which can be used as a detection-signal.

Similarly, a spectral analysis can be performed, with the AC component of the string current being separated out by means of a toroidal core choke and then pre-filtered analogue-wise, if necessary, and digitised by means of a rapid A/D converter (up to about 200 kS/s). The digital signal is then converted into an interference power spectrum through transformation into the frequency domain (e.g. by FFT), which is very computingintensive. An arc can then be detected by monitoring the spectrum for any abrupt increase in interference power; and because of the frequency domain representation, it is easy to remove individual frequency ranges with potential interfering-signals due to e.g. internal clock frequencies.

With the parallel resonant circuit, the frequency range monitored is determined by the manufactured components’ values, and thus cannot be modified. Spectral analysis is very computing-intensive, due to the transformation into the frequency domain. These circuits can be implemented with additional hardware.

With the arc-detection approach presented here, the extra hardware and the transformation into the frequency domain can be dispensed with. The inventive approach makes it possible to recognise series arcs 102, parallel arcs 102, and earth fault arcs 102 by algorithmic evaluation of the measurement signal from the current-sensor inside the inverter.

Fig. 4 is a diagram showing a frequency spectrum of a current-signal 112, in an example of an embodiment of the present invention. The frequency spectrum is plotted in a graph, with logarithmic frequency f in Hertz on the abscissa and power density in decibels on the ordinate. A first curve 400 and second curve 402 are shown. The first curve 400 represents the frequency spectrum in the direct current string with no arc. The second curve 402 represents the frequency spectrum in the DC string with an active arc. In large areas, there is a power density difference of approximately ten decibels between the first curve 400 and the second curve 402. In other words, with an arc, the prevailing power density in the DC string is about ten decibels higher than without an arc.

In other words, Fig. 4 shows power spectral density for the current during an active arc and without an arc. Due to the occurrence of an arc in photovoltaic systems, there occurs a broadband increase in the noise level in the current. The signal-to-interference ratio between arc noise 402 and background noise 400 is up to 10 dB, making detection possible. Because of interfering frequencies, e.g. resulting from switching-operations in the inverter or from the mains frequency, and from their respective harmonic oscillations, the signal-to-interference ratio decreases and can affect detection.

Fig. 5 shows a frequency spectrum of a detection-signal 114 in an example of an embodiment of the present invention. The frequency spectrum of the detection-signal 114 is plotted in a frequency/power-density chart, as in Fig. 4. The frequency spectrum of the detection-signal 114 is a section of the entire frequency spectrum of the current-signal.

This section represents a frequency band between an upper frequency limit 500 and a lower frequency limit 502. The frequency limits 500, 502 are defined by the configuration of the band-pass filter during the processing of the current-signal. The frequency limits 500, 502 can be easily adjusted. The larger the frequency band of the detection-signal 114, the greater the reliability of arc detection.

Fig. 6 shows changes in current-intensity over time in a detection-signal 114, due to an electric arc, in an example of an embodiment of the present invention. The fluctuations in current are shown in a graph with time t in seconds on the abscissa and current-intensity I in amperes on the ordinate. The current-intensity I is plotted between minus 0.2 amps and 0.2 amps. Due to the filtering performed in the bandpass filter, the detection-signal 114 now has no DC component. The detection-signal 114 has background noise 600 of low amplitude, as long as no arc has ignited. The background noise 600 is essentially uniform in the frequency band of the detection-signal 114.

When an arc is ignited, the amplitude of the signal noise 602 of the detection-signal 114 rises sharply and remains at an elevated level as long as the arc is burning. Once the arc is extinguished, the detection-signal again shows the background noise 600.

In other words, the Figures show the current, filtered using a bandpass filter with a bandwidth of 2.5 to 5 kHz. By implementing a digital bandpass filter, not only is the DC component removed but also the frequency range is restricted. Attenuation of at least 70 to 80 dB is necessary to remove the entire DC component. The bandwidth should be as broad as possible, so that narrow-band interference does not lead to false positives.

Fig. 7 shows the detection of an electric arc in an example of an embodiment of the present invention. Essentially, Fig. 7 shows an enlargement of the detection-signal 114 of Fig. 6 with its background noise 600 and the signal noise 602. Current-intensity I of zero to 0.03 amperes is shown on the ordinate. Thus, the amplitude of the background noise 600 and the amplitude of the signal noise 602 are shown considerably enlarged. Also shown here are a first threshold value 700, a second threshold value 702, and a third threshold value 704, which are proportional in relation to the amplitude of the background noise 600. These thresholds 700, 702, 704 are dynamically adapted to the amplitude or level of the background noise 600. From the moment that the level of the detection-signal 114 goes above the first threshold 700, at the beginning of the signal noise 602, the threshold values 700, 702, 704 stay constant. If the level is greater than the second threshold value 702, then there is a wait, for a predetermined first minimum period of time, before the arc signal is outputted. Within that minimum period, the level must be above the threshold value, to trigger the arc signal. Since the level here leaps, almost simultaneously, above the third threshold value 704, there is only a much shorter second minimum waiting period until the arc signal is outputted. Thus, the arc can already be detected at the first peak in the level, upon ignition of the arc, even if the level then drops briefly below the second threshold 702 and even below the first threshold 700 before rising again and staying above the second threshold 702.

When the arc is extinguished, the level drops back down to the background noise 600.

In other words, Fig. 7 shows signal processing of the filtered current. Adaptive thresholds 700, 702, 704, for detection purposes, are shown. These are activated in accordance with the operating situation. The signal processing of the filtered current is performed by determining the root mean square over a defined interval of time. The window to be calculated is shifted 10% each time so that weighting with the preceding data occurs and therefore an arc can be detected faster.

Detection occurs when the detection-signal exceeds a definable threshold value th for a minimum period of time t_min. Since the noise power varies depending on the arc, and not all arcs have a pronounced peak at ignition, detection can occur at different threshold values 700, 702, 704, which are adapted to the particular photovoltaic system, due to the different background noise 600 in different systems. Adaptation is done by determining the root mean square of the background noise. The threshold values 700, 702, 704 are then determined using e.g. a fixed multiplier, from the background-noise value 600, or using a combined function of background noise 600 and dispersion.

For detection purposes, three thresholds 700, 702, 704 are required here. The third threshold value 704 is required for the detection of peaks in the detection-signal 114. The second threshold value 702 is required for continuous detection, in the event that there is no peak in the formation of an arc. The first threshold 700 is activated for a fixed, predefined, period of time, once a peak is detected, because the noise power often nosedives after ignition. Successful detection is marked in bold here.

Fig. 8 is a flow chart of a computer program for detecting an electric arc, in an example of an embodiment of the present invention. The computer program is an example of a method for detecting an electric arc as shown for example in Fig. 2.

The process of performing the method begins at starting point 800. Then, a present root mean square of the current-signal is calculated, in calculation step 802. Next, in a query step 804, it is determined whether the time is greater than next_threshold. If the time is greater than next_threshold, new threshold values are calculated, in another calculation step 808, based on the root mean squares. Then, using the present threshold values, checking is performed, in checking step 808, to see whether the root mean square of the present time window is greater than the threshold values.

If the root mean square is less than a normal threshold value, a counter is reset, in resetting-step 810, and an endpoint 812 of the process is arrived at.

If, however, the root mean square is greater than the normal threshold value, but less than a peak threshold value, then the counter is incremented, in incrementation-step 814, and a time is detected. Then, if the time is less than an alarm time, the end point 812 is reached. Otherwise, an alarm is outputted, in output step 816, and then the endpoint 812 is reached. If the root mean square is greater than the peak threshold value, then — in incrementation-step 814 — the normal threshold is also set, for a peak period, to a post-peak threshold value.

In other words, Fig. 8 shows a program flow chart for the arc detector, consisting of threshold adjustment, normal detection and peak detection.

The embodiments selected for description and illustration in the Figures are given by way of example only. Different embodiments can be combined with one another in whole or in part. Also, one embodiment can be supplemented with features of another embodiment.

Furthermore, the steps of the method presented here may be repeated, or may be performed in a different order from that described.

If an example includes an “and/or” relation between a first feature and a second feature, this is to be read as meaning that, in one form of embodiment, both the first feature and the second feature are present, while in another form of embodiment only the first feature or only the second feature is present.

Claims

Claims

1. A method (200) for detecting an electric arc (102) in a direct current string (104) of a photovoltaic installation (106), said method (200) comprising the steps of: - processing (202) a current-signal (112) acquired from a direct current string (104), to obtain a detection-signal (114), said processing (202) being performed using a bandpass filter (116); and - monitoring (204) the detection-signal (114), to detect an electric arc (102), the arc (102) being detected when the level of the detection-signal (114) is greater than a threshold value (118, 702) for more than a minimum period of time.
2. A method (200) as claimed in claim 1, with a step of: - adjusting (208), in which, using the detection-signal (114), the threshold value (118, 702) is adjusted in predetermined proportion to the background noise (600) of the detection-signal (114).
3. A method (200) as claimed in one of the above claims, wherein, in the monitoring step (204), the electric arc (102) is detected, using at least one additional threshold value (704) and one additional minimum period of time; said electric arc (102) being detected when the detection-signal (114) is greater than the threshold value (118, 702) for more than the minimum period of time, and/or is greater than the additional threshold value (704) for more than the additional minimum period of time.
4. A method (200) as claimed in any of the above claims, wherein, in the processing step (202), the current-signal (112) is further processed, using a smoothing algorithm (124), to obtain the detection-signal (114).
5. A method (200) as claimed in any of the above claims, wherein, in the processing step (202), the current-signal (112) is averaged quadratically to obtain the detection-signal (114).
6. A method (200) as claimed in any of the above claims, wherein, in the processing step (202), the current-signal (112) is processed digitally.
7. A method (200) as claimed in any of the above claims, with a step (208) of acquiring the current-signal (112), wherein the current-signal (112) is sampled — from the DC part (104) — in an inverter (132) forming part of the photovoltaic installation (106), at a sampling frequency of between 2 kHz and 20 kHz.
8. An apparatus (100) adapted to performing, implementing, and/or controlling all the steps of a method (200) as claimed in any of the above claims.
9. A computer program adapted to performing, implementing, and/or controlling all the steps of a method as claimed in any of the above claims.
10. A machine-readable storage medium with a computer program, as claimed in claim 9, stored on it.