Direct current fault arc detection method

Abstract

A direct current (DC) fault arc detection method, including performing sampling, filtering and fast Fourier transform (FFT) on an input current of a high frequency power electronics converter, to obtain an amplitude-frequency characteristic curve of a current high frequency component; and selecting, from the amplitude-frequency characteristic curve, at least one frequency band including a switching frequency or a multiple frequency, calculating a peak value D1 of amplitudes and an average value D2 of the amplitudes within the frequency band, and determining, according to a change of a distance between the peak value D1 and the average value D2, whether an arc occurs, where if the average value D2 approaches the peak value D1, an arc occurs; otherwise, no arc occurs.

Claims

1. A direct current (DC) fault arc detection method, comprising the following steps: step 1: performing sampling, filtering and fast Fourier transform (FFT), by using a pulse transformer, a band-pass filter, an AD adapter and an MCU processor, on an input current of a high frequency power electronics converter, to obtain an amplitude-frequency characteristic curve of a current high frequency component; and step 2: selecting, from the amplitude-frequency characteristic curve, at least one frequency band comprising a switching frequency or a multiple frequency, calculating a peak value D1 of amplitudes and an average value D2 of the amplitudes within the frequency band, and determining, according to a change of a distance between the peak value D1 and the average value D2, whether an arc occurs, wherein if the average value D2 approaches the peak value D1, an arc occurs; otherwise, no arc occurs; wherein in the step 2, a difference D3 between the peak value D1 and the average value D2 is calculated, the difference D3 is divided by the average value D2 to obtain a percentage D4, and statistics about a quantity X of the percentage D4 smaller than a first threshold T1 in a predetermined number of pieces of data are collected, and if X is greater than a second threshold T2, the arc occurs; otherwise, no arc occurs; step 3: reporting an error after it is detected that the arc occurs.

2. The DC fault arc detection method according to claim 1, wherein in the step 2, a calculation method of the peak value D1 includes: calculating a moving average value of every Y amplitude of data within the frequency band; and selecting a largest value after comparison to obtain the peak largest value D1; and Y indicates a number of the amplitudes with the frequency band.

3. The DC fault arc detection method according to claim 1, wherein in the step 2, a calculation method of the average value D2 includes: removing a first predetermined percentage of largest amplitudes from the amplitudes within the frequency band, removing a second predetermined percentage of smallest amplitudes from the amplitudes within the frequency band, and obtaining an average value of the remaining amplitudes, so that the average value D2 is obtained.

4. The DC fault arc detection method according to claim 3, wherein the first predetermined percentage is 50%.

5. The DC fault arc detection method according to claim 3, wherein the second predetermined percentage is 30%.

6. The DC fault arc detection method according to claim 3, wherein a percentage of the remaining amplitude is 20%.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 shows a commonly used BOOST circuit;

(2) FIG. 2 shows a drive signal of a BOOST circuit;

(3) FIG. 3 shows an amplitude-frequency characteristic curve when no arc occurs;

(4) FIG. 4 shows an amplitude-frequency characteristic curve when an arc occurs;

(5) FIG. 5 shows an implementation of performing arc detection by using a method provided in the present invention and by using a photovoltaic inverter as an example;

(6) FIG. 6 is a flowchart of an MCU processing part; and

(7) FIG. 7 is an overall flowchart of the present invention.

DETAILED DESCRIPTION

(8) To make the present invention more comprehensible, detailed description is provided below by using preferred embodiments and with reference to the accompanying drawings.

(9) The present invention provides a DC arc detection method combined with an inherent characteristic of a power electronics converter. Most of current power electronics converters use the principle of pulse width modulation (PWM) high frequency chopping, and chop a DC current into a DC impulse current by using high frequency (usually 500 Hz to several kHz) switching of a switching device. In this way, the DC current generates, on a frequency band that is at a switching frequency and at a multiple frequency of the switching frequency, a signal having an amplitude higher than that of a signal generated at a nearby frequency. When an arc occurs, because a frequency spectrum of the current, taken as a whole, is lifted up, changes of noise amplitudes at the switching frequency and the multiple frequency of the switching frequency of the converter are smaller than changes of noise amplitudes on a nearby frequency band.

(10) A commonly used BOOST circuit shown in FIG. 1 is used as an example for description. A switch Q is a high frequency switch, a drive signal of the switch is Vgs shown in FIG. 2, and an inductance current and an input current are respectively i.sub.L and i.sub.in. It can be learned from a waveform that the input DC current has an abundant of switching frequency secondary noises.

(11) FFT is performed on the input current, so that an amplitude-frequency characteristic of the input current is obtained, as shown in FIG. 3. Using a switching frequency 16 kHz as an example, it can be learned that within a frequency band of F1 to F2, a noise peak value D1 at the switching frequency 16 kHz is far higher than an average value D2 within the frequency band, and cases are the same at multiple frequencies of 16 kHz, such as 32 kHz and 48 kHz.

(12) Within the frequency band of F1 to F2, if D3 is defined as a difference between the peak value D1 and the average value D2, a quantization parameter, that is, a ratio D3/D2 of the D3 to the D2 may be used to determine whether an arc occurs, and if no arc occurs, the D3/D2 is relatively large.

(13) The FFT amplitude-frequency characteristic of the input current after an arc occurs is shown in FIG. 4. It can be learned that because the amplitudes, taken as a whole, are lifted up by high frequency noises generated by the arc, the peak value D1 within the frequency band including the switching frequency or a frequency band including a multiple frequency is greatly reduced relative to the average value D2 within the frequency band.

(14) It can be learned from the foregoing analysis that, for a high frequency power electronics converter, an amplitude-frequency characteristic of a current after an arc occurs has two apparent characteristics:

(15) 1. In some frequency bands, amplitudes are apparently lifted up.

(16) 2. At a position near a switching frequency and a multiple frequency of the switching frequency, because an amplitude of a high frequency signal of the converter is drowned by an arc noise or is apparently smaller relative to a peak value, noises of the switching converter at the switching frequency and the multiple frequency of the switching frequency are unchanged, and noises on other frequency bands are bigger because of the arc noise.

(17) Based on the foregoing analysis, an arc detection process provided in the present invention is as follows:

(18) 1. performing sampling, filtering and FFT on an input current, to obtain an amplitude-frequency characteristic of a current high frequency component;

(19) 2. selecting a to-be-compared frequency band according to a switching characteristic of a converter, to determine whether an amplitude is larger.

(20) 3. selecting one or more frequency bands including a switching frequency or a multiple frequency, calculating a largest value D1 and an average value D2 within the frequency band, and determining whether an arc occurs by performing logical calculation according to the phenomenon that an average value is larger and a largest value is substantially unchanged after an arc occurs.

(21) Using a photovoltaic inverter as an example, the following provides an implementation of performing arc detection by using the method in the present invention.

(22) 1) Current data is obtained by sampling a DC current by using a pulse transformer, a band-pass filter, an AD adapter and an MCU processor, as shown in FIG. 5.

(23) 2) In the MCU processing part, Hanning window and FFT are sequentially performed on the sampling data, to convert a time domain into a frequency domain, and a high frequency signal amplitude is calculated before a data logic processing is performed, as shown in FIG. 6.

(24) 3) In the data logic processing part, a noise signal generated according to an inherent characteristic of a machine is mainly used. The noise signal of the machine is mainly located at a multiple frequency of a switching frequency of the machine. FIG. 4 shows a frequency spectrum when a fault arc occurs, and FIG. 3 shows a frequency spectrum when no fault arc occurs.

(25) In the figures, for example, the switching frequency of the machine is 16 kHz. As shown in the frequency spectrums, the machine has apparent noise signals near 16 kHz, 32 kHz, and 48 kHz regardless of whether an arc fault occurs.

(26) Based on the foregoing two points, the MCU selects a frequency band near the multiple frequencies for processing. For example, two frequency bands 27 kHz to 37 kHz and 43 kHz to 53 kHz may be selected in FIG. 3 and FIG. 4 for processing. With reference to FIG. 7, the processing process is as follows (where D1, D2, and D3 are marked in FIG. 3).

(27) Step 1: Calculate a moving average value of Y pieces of data within the frequency band, and select a largest value D1 after comparison. In the present invention, by using the largest moving average value instead of directly using a largest value, interference is considered, and filtering to an extent is reached by using the moving average value.

(28) Step 2: Remove large values and small values of a specific percentage within the frequency band to obtain an average value D2.

(29) Step 3: Subtract D2 from the largest value D1, to obtain a difference D3.

(30) Step 4: Divide the difference D3 by the average value D2, to obtain a percentage D4, and collect statistics about a quantity X of the percentage D4 smaller than a threshold T1 in N pieces of data, where if X is greater than a threshold T2, an arc occurs; otherwise, no arc occurs.

(31) This algorithm has the following advantage over an algorithm in which only amplitude values within some frequency band are determined: first, the difference D3 when an arc fault occurs is smaller than the difference D3 when no arc fault occurs; secondly, the average value D2 when an arc fault occurs is greater than the average value D2 when no arc fault occurs; finally, by dividing D3 by D2, a numerator is smaller and a denominator is larger when an arc fault occurs, so that an obtained numerical value is certainly smaller and can be more easily distinguished.

(32) The present invention uses the noise signal in consideration of the following two points: First, amplitudes of a frequency spectrum, taken as a whole, are lifted up when an arc fault occurs; Secondly, an amplitude of a peak value at a multiple frequency is substantially unchanged.