CN113238165A - Reactor turn-to-turn short circuit fault detection method based on IFRA and SVM - Google Patents
Reactor turn-to-turn short circuit fault detection method based on IFRA and SVM Download PDFInfo
- Publication number
- CN113238165A CN113238165A CN202110183247.1A CN202110183247A CN113238165A CN 113238165 A CN113238165 A CN 113238165A CN 202110183247 A CN202110183247 A CN 202110183247A CN 113238165 A CN113238165 A CN 113238165A
- Authority
- CN
- China
- Prior art keywords
- voltage
- turn
- signal
- reactor
- short circuit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/52—Testing for short-circuits, leakage current or ground faults
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)
Abstract
The invention discloses a reactor turn-to-turn short circuit fault detection method based on IFRA and SVM, which comprises the following steps: 1) building a high-voltage pulse device; 2) injecting excitation pulse voltage to the end part of the winding of the reactor to be tested by the high-voltage pulse device; 3) monitoring an input excitation voltage signal and an output response current signal; 4) drawing an actually measured frequency response curve; 5) calculating the digital statistical index of the actually measured frequency response curve; 6) and the data processing module inputs the digital statistical indexes into the SVM classifier to finish turn-to-turn short circuit fault judgment. According to the invention, nanosecond pulse signals with abundant frequency spectrum components are used as excitation signals, fault diagnosis is carried out by combining with an SVM (support vector machine), and turn-to-turn short circuit faults of the reactor can be detected quickly, sensitively and efficiently.
Description
Technical Field
The invention relates to the technical field of power equipment detection, in particular to a reactor turn-to-turn short circuit fault detection method based on IFRA and SVM.
Background
With the large investment of reactors in power systems, the corresponding inter-turn short circuit fault problem also frequently occurs. The existing method for detecting the turn-to-turn short circuit of the reactor mainly comprises a magnetic field coil method, a temperature detection method, an electrical parameter detection method and a pulse oscillation method. The magnetic field coil method is more effective for the case of multi-turn short circuit; the temperature detection method requires a high-precision sensor; the sensitivity of the electrical parameter detection method is not high; the pulse oscillation method requires expensive equipment and is complicated to operate. In general, the existing detection method generally has the problems of complex operation, low sensitivity and the like. Therefore, an efficient and sensitive turn-to-turn short circuit fault detection method needs to be found.
Disclosure of Invention
The invention aims to provide a reactor turn-to-turn short circuit fault detection method based on IFRA and SVM, which comprises the following steps:
1) and (5) building a high-voltage pulse device.
The high-voltage pulse device comprises a high-voltage direct-current power supply, an FPGA module, a charging protection resistor, a pulse generating circuit, a control circuit and a load resistor.
And the high-voltage direct current power supply charges a capacitor in the pulse forming unit through the charging protection resistor.
The FPGA module receives pulse parameters sent by the upper computer and generates a switch control signal.
The control circuit receives the switch control signal and controls the on-off of the switch in the pulse generating circuit to enable the pulse generating circuit to generate pulse voltage.
The pulse generating circuit comprises a capacitor and a switch which are connected in series.
And the pulse generating circuit injects excitation pulse voltage to the end part of the winding of the reactor to be tested through the load resistor.
2) The upper computer sets pulse parameters and sends the pulse parameters to the high-voltage pulse device.
3) The high-voltage pulse device sends excitation pulse voltage to the end part of the winding of the reactor to be tested.
4) And a voltage sensor placed at the end part of the winding of the reactor to be tested monitors an input excitation voltage signal in real time, and a current sensor placed at the end part of the winding of the reactor to be tested monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
5) The data acquisition module respectively processes the input excitation voltage signal and the output response current signal to obtain an excitation voltage digital signal and a response voltage digital signal, and sends the excitation voltage digital signal and the response voltage digital signal to the data processing module.
The step that the data acquisition module respectively processes the input excitation voltage signal and the output response current signal comprises the following steps:
and 5.1) carrying out analog-to-digital conversion on the input excitation voltage signal to obtain an excitation voltage digital signal.
And 5.2) converting the output response current signal into an output response voltage signal, and then performing analog-to-digital conversion to obtain a response voltage digital signal.
6) The data processing module respectively carries out fast Fourier transform on the input excitation voltage signal and the output response current signal to obtain a frequency domain representation U of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f)。
7) The data processing module represents U according to the frequency domain of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f) And calculating a frequency response transfer function TF, and drawing an actually measured frequency response curve.
The frequency response transfer function TF is as follows:
in the formula I0(f) To output a frequency domain representation of the response current signal. U shapei(f) Is a frequency domain representation of the input excitation voltage signal.
The data processing module calculates a digital statistical index of the measured frequency response curve according to the reference frequency response curve and the measured frequency response curve.
The digital statistical indexes comprise a correlation coefficient CC, a Euclidean distance ED, an absolute value sum ASLE of a logarithmic error and a square sum SSE of the error.
Wherein, the correlation coefficient CC is as follows:
wherein x (i) is a transfer function amplitude sequence of the measured frequency response curve. N is the sequence length. Y (i) is the transfer function amplitude sequence of the reference frequency response curve.
The Euclidean distance ED is as follows:
the absolute value of the log error and ASLE are as follows:
the sum of squared errors SSE is shown below:
in the formula, i is an arbitrary sequence point.
9) The data processing module stores an SVM classifier, inputs the digital statistical indexes into the SVM classifier and finishes turn-to-turn short circuit fault judgment. The judging content of the turn-to-turn short circuit fault comprises the short circuit degree, the short circuit position and whether the turn-to-turn short circuit of the same strand exists.
The SVM classifier is obtained by training different types of training data. The training data types comprise digital statistical indexes of frequency response curves of the reactor under short circuits of different degrees, digital statistical indexes of frequency response curves of the reactor under short circuits of different positions, digital statistical indexes of frequency response curves of the reactor under short circuits of the same turn-to-turn of the same strand and digital statistical indexes of frequency response curves of the reactor under short circuits of different turn-to-turn of the different strand.
The training data acquisition step comprises:
9.1) short-circuiting the resistor with the resistance R between any two turns of coils of the reactor winding.
9.2) injecting excitation pulse voltage into the end part of the reactor winding in the step 9.1) by the high-voltage pulse device.
And 9.3) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
9.5) updating the resistance value R' ═ R + DeltaR, and returning to the step 9.1) until n is collected1A group excitation voltage digital signal and a response voltage digital signal. Δ R is the resistance value increment.
9.6) setting a short-circuit fault point at a position L away from the head end of the reactor winding. L has an initial value of 0.
9.7) injecting excitation pulse voltage into the end part of the reactor winding in the step 9.6) by the high-voltage pulse device.
And 9.8) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
9.9) update L' ═ L + Δ L and return to step 9.6) until n is collected2A group excitation voltage digital signal and a response voltage digital signal. Δ L is the distance increment.
9.10) setting a short-circuit fault point between the same strand coils of the reactor winding.
9.11) the high-voltage pulse device injects excitation pulse voltage into the end part of the reactor winding of the step 9.10).
And 9.12) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
9.13) short-circuit fault points are arranged among different strands of coils of the reactor winding.
9.14) the high-voltage pulse device injects excitation pulse voltage into the end part of the reactor winding of the step 9.10).
9.15) the voltage sensor monitors the input excitation voltage signal in real time, and the current sensor monitors the output response current signal in real time and sends the output response current signal to the data acquisition module.
9.16) the data acquisition module processes the input excitation voltage signal and the output response current signal respectively to obtain an excitation voltage digital signal and a response voltage digital signal, and sends the excitation voltage digital signal and the response voltage digital signal to the data processing module.
9.17) the data processing module respectively carries out fast Fourier transform on the input excitation voltage signal and the output response current signal to obtain a frequency domain representation U of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f)。
9.18) the data processing module represents U from the frequency domain of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f) And calculating a frequency response transfer function TF, and respectively drawing the actually measured frequency response curves of the reactor under short circuits of different degrees, short circuits of different positions, short circuits between turns of the same strand and short circuits between turns of different strands.
9.19) the data processing module calculates the digital statistical index of each measured frequency response curve and uses the digital statistical index as the training data of the SVM classifier.
The technical effect of the invention is undoubtedly that the nanosecond pulse signal with abundant frequency spectrum components is adopted as the excitation signal, fault diagnosis is carried out by combining with the SVM, and the turn-to-turn short circuit fault of the reactor can be detected quickly, sensitively and efficiently.
Drawings
FIG. 1 is a flow chart of the inventive method.
Fig. 2 is a diagram of an implementation of the inventive method.
Fig. 3 is a schematic structural diagram of a high voltage pulse generation module.
Fig. 4 is a frequency response graph of the degree of turn-to-turn short circuit.
Fig. 5 is a graph of the frequency response of the turn-to-turn short circuit position.
Fig. 6 is a graph of turn-to-turn short frequency response for the same strand and different strands.
Fig. 7 is a fault diagnosis diagram of the degree of turn-to-turn short circuit.
Fig. 8 is a fault diagnosis diagram of the turn-to-turn short circuit position.
Fig. 9 is a fault diagnosis diagram of turn-to-turn short circuits of the same strand and different strands.
Detailed Description
The present invention is further illustrated by the following examples, but it should not be construed that the scope of the above-described subject matter is limited to the following examples. Various substitutions and alterations can be made without departing from the technical idea of the invention and the scope of the invention is covered by the present invention according to the common technical knowledge and the conventional means in the field.
Example 1:
referring to fig. 1 to 9, a reactor turn-to-turn short circuit fault detection method based on an IFRA (Impulse Frequency Response) and an SVM (Support Vector Machine) includes the following steps:
1) and (5) building a high-voltage pulse device.
The high-voltage pulse device comprises a high-voltage direct-current power supply, an FPGA module, a charging protection resistor, a pulse generating circuit, a control circuit and a load resistor.
And the high-voltage direct current power supply charges a capacitor in the pulse forming unit through the charging protection resistor.
The FPGA module receives pulse parameters sent by the upper computer and generates a switch control signal.
The control circuit receives the switch control signal and controls the on-off of the switch in the pulse generating circuit to enable the pulse generating circuit to generate pulse voltage.
The pulse generating circuit comprises a capacitor and a switch which are connected in series.
And the pulse generating circuit injects excitation pulse voltage to the end part of the winding of the reactor to be tested through the load resistor.
2) The upper computer sets pulse parameters and sends the pulse parameters to the high-voltage pulse device.
3) The high-voltage pulse device injects excitation pulse voltage to the end part of the winding of the reactor to be tested.
4) And a voltage sensor placed at the end part of the winding of the reactor to be tested monitors an input excitation voltage signal in real time, and a current sensor placed at the end part of the winding of the reactor to be tested monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
5) The data acquisition module respectively processes the input excitation voltage signal and the output response current signal to obtain an excitation voltage digital signal and a response voltage digital signal, and sends the excitation voltage digital signal and the response voltage digital signal to the data processing module.
The step that the data acquisition module respectively processes the input excitation voltage signal and the output response current signal comprises the following steps:
and 5.1) carrying out analog-to-digital conversion on the input excitation voltage signal to obtain an excitation voltage digital signal.
And 5.2) converting the output response current signal into an output response voltage signal, and then performing analog-to-digital conversion to obtain a response voltage digital signal.
6) The data processing module respectively carries out fast Fourier transform on the input excitation voltage signal and the output response current signal to obtain a frequency domain representation U of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f) In that respect Wherein the output response current signal is a response current signal represented in the form of a voltage.
7) The data processing module represents U according to the frequency domain of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f) And calculating a frequency response transfer function TF, and drawing an actually measured frequency response curve.
The frequency response transfer function TF is as follows:
in the formula I0(f) To output a frequency domain representation of the response current signal. U shapei(f) Is a frequency domain representation of the input excitation voltage signal.
The data processing module calculates a digital statistical index of the measured frequency response curve according to the reference frequency response curve and the measured frequency response curve.
The digital statistical indexes comprise a correlation coefficient CC, a Euclidean distance ED, an absolute value sum ASLE of a logarithmic error and a square sum SSE of the error.
Wherein, the correlation coefficient CC is as follows:
wherein x (i) is a transfer function amplitude sequence of the measured frequency response curve. N is the sequence length. Y (i) is the transfer function amplitude sequence of the reference frequency response curve.
The Euclidean distance ED is as follows:
the absolute value of the log error and ASLE are as follows:
the sum of squared errors SSE is shown below:
in the formula, i is an arbitrary sequence point.
9) The data processing module stores an SVM classifier, inputs the digital statistical indexes into the SVM classifier and finishes turn-to-turn short circuit fault judgment. The judging content of the turn-to-turn short circuit fault comprises the short circuit degree, the short circuit position and whether the turn-to-turn short circuit of the same strand exists.
The SVM classifier is obtained by training different types of training data. The training data types comprise digital statistical indexes of frequency response curves of the reactor under short circuits of different degrees, digital statistical indexes of frequency response curves of the reactor under short circuits of different positions, digital statistical indexes of frequency response curves of the reactor under short circuits of the same turn-to-turn of the same strand and digital statistical indexes of frequency response curves of the reactor under short circuits of different turn-to-turn of the different strand.
The training data acquisition step comprises:
9.1) short-circuiting the resistor with the resistance R between any two turns of coils of the reactor winding.
9.2) injecting excitation pulse voltage into the end part of the reactor winding in the step 9.1) by the high-voltage pulse device.
And 9.3) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
9.5) updating the resistance value R' ═ R + DeltaR, and returning to the step 9.1) until n is collected1A group excitation voltage digital signal and a response voltage digital signal. Δ R is the resistance value increment.
9.6) setting a short-circuit fault point at a position L away from the head end of the reactor winding. L has an initial value of 0.
9.7) injecting excitation pulse voltage into the end part of the reactor winding in the step 9.6) by the high-voltage pulse device.
And 9.8) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
9.9) update L' ═ L + Δ L and return to step 9.6) until n is collected2A group excitation voltage digital signal and a response voltage digital signal. Δ L is the distance increment.
9.10) setting a short-circuit fault point between the same strand coils of the reactor winding.
9.11) the high-voltage pulse device injects excitation pulse voltage into the end part of the reactor winding of the step 9.10).
And 9.12) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module.
9.13) short-circuit fault points are arranged among different strands of coils of the reactor winding.
9.14) the high-voltage pulse device injects excitation pulse voltage into the end part of the reactor winding of the step 9.10).
9.15) the voltage sensor monitors the input excitation voltage signal in real time, and the current sensor monitors the output response current signal in real time and sends the output response current signal to the data acquisition module.
9.16) the data acquisition module processes the input excitation voltage signal and the output response current signal respectively to obtain an excitation voltage digital signal and a response voltage digital signal, and sends the excitation voltage digital signal and the response voltage digital signal to the data processing module.
9.17) the data processing module respectively carries out fast Fourier transform on the input excitation voltage signal and the output response current signal to obtain a frequency domain representation U of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f)。
9.18) the data processing module represents U from the frequency domain of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f) And calculating a frequency response transfer function TF, and respectively drawing the actually measured frequency response curves of the reactor under short circuits of different degrees, short circuits of different positions, short circuits between turns of the same strand and short circuits between turns of different strands.
9.19) the data processing module calculates the digital statistical index of each measured frequency response curve and uses the digital statistical index as the training data of the SVM classifier.
Example 2:
referring to fig. 1 to 9, a method for detecting a reactor turn-to-turn short circuit fault based on IFRA and SVM includes the following steps:
1) generating excitation pulse voltage by using a high-voltage pulse generation module, and loading the excitation pulse voltage to the end part of the reactor winding to perform related short-circuit experiments; the purpose of performing the relevant short circuit experiments was: a) observing the frequency response curve change, and b) acquiring SVM training data.
The high-voltage pulse generation module comprises a high-voltage direct-current power supply, a charging protection resistor, a pulse generation circuit, a control circuit and a load resistor. The high-voltage direct-current power supply provides input voltage and charges a capacitor in the pulse generation circuit through the charging protection resistor; the control circuit generates pulse voltage to act on the load resistor by controlling the on-off of the solid-state switch in the pulse generating circuit; the voltage across the load resistor is applied to the reactor winding ends.
2) Acquiring an excitation signal and a response signal using voltage and current sensors;
3) the data processing module analyzes and processes the acquired signals, draws a frequency response curve and calculates mathematical statistical parameters;
the data acquisition is that a voltage sensor is used for carrying out analog-to-digital conversion on a high-voltage nanosecond pulse excitation signal generated by a high-voltage pulse generation module, and meanwhile, a current sensor is used for carrying out I-V conversion on a response current signal on a grounding wire at the tail end of a winding and acquiring the response current signal to a data processing module through analog-to-digital conversion.
The data processing module mainly comprises functions of frequency response curve drawing and mathematical statistic index calculation, and can be realized by programming.
4) And analyzing the frequency response data through the SVM to diagnose turn-to-turn short circuit faults.
The support vector machine is implemented by computer programming to carry out fault diagnosis. Mathematical statistics indicators of the frequency response curves are calculated, respectively, including Correlation Coefficient (CC), Euclidean Distance (ED), sum of absolute values of logarithmic errors (ASLE), Sum of Squared Errors (SSE). And finally, inputting the sample data into the SVM for fault diagnosis.
Example 3:
referring to fig. 2, a reactor turn-to-turn short circuit fault detection system based on IFRA and SVM includes a high voltage pulse generation module, an upper computer, a data acquisition module, a data processing module, a current sensor and a voltage sensor.
The high-voltage pulse generation module includes a high-voltage dc power supply, a charging protection resistor, a pulse generation circuit, a control circuit, and a load resistor, as shown in fig. 3.
The high-voltage direct-current power supply provides input voltage and charges a capacitor in the pulse generation circuit through the charging protection resistor; the control circuit generates pulse voltage to act on the load resistor by controlling the on-off of the solid-state switch in the pulse generating circuit; the voltage across the load resistor is applied to the reactor winding ends. And relevant turn-to-turn short circuit experiments are carried out, including short circuit comparison experiments of different short circuit degrees, different short circuit positions and the same strand and different strands.
Performing analog-to-digital conversion on the high-voltage nanosecond pulse excitation signal generated by the high-voltage pulse generation module by using a voltage sensor, and acquiring the high-voltage nanosecond pulse excitation signal to a data processing module; meanwhile, the current sensor performs I-V conversion on a response current signal on a grounding wire at the tail end of the winding, and the response current signal is acquired to the data processing module through analog-to-digital conversion.
The data processing module mainly comprises functions of frequency response curve drawing and mathematical statistic index calculation, and can be realized by programming; and the frequency response curve drawing is to perform fast Fourier transform on the acquired input excitation voltage signal and the acquired output response current signal respectively, perform frequency response calculation according to the formula (1) and draw a frequency response curve. In the formula I0(f) For outputting a frequency-domain representation of the response current signal, Ui(f) Is a frequency domain representation of the excitation voltage signal. TF is the winding frequency response amplitude curve expressed in gain.
The frequency response data is then compared. And comprehensively comparing and analyzing the change information of the frequency and the amplitude of the wave crest, the wave trough of the actually measured frequency response curve by taking the frequency response data measured by the reactor under the healthy condition as a reference, and when the changes of the frequency and the waveform form of the wave crest, the wave trough and the waveform form of the actually measured frequency response curve are large, determining that the reactor has turn-to-turn short circuit fault. And (5) drawing and analyzing a frequency response curve of the experiment in the step one.
Example 4:
an IFRA and SVM-based reactor turn-to-turn short circuit fault detection method is used for short circuit detection experiments, and the process is as follows:
1) referring to fig. 4, the inter-turn short circuit detection of the reactor based on IFRA and SVM is performed to different degrees by using a method for detecting the inter-turn short circuit fault of the reactor, and the process is as follows:
different short circuit degrees are simulated by connecting resistances (shorts) of 2.2 Ω, 1 Ω and 0 Ω between certain two turns of the reactor, respectively, as shown in fig. 4. The equivalent inductance is smaller along with the deepening of the short circuit degree in 0.8-1.2MHz, and a frequency response curve can move to high frequency according to the formula (2); resonance peaks under different short circuit degrees in 1.8-2MHz all move to low frequency, and because the short circuit point set in the experiment is connected by a lead wire, a parallel stray capacitor is possibly introduced to increase the capacitance and play a leading role under high frequency.
2) Referring to fig. 5, the inter-turn short circuit detection at different positions is performed by using a reactor inter-turn short circuit fault detection method based on IFRA and SVM, and the process is as follows:
short-circuit fault points are provided at the head ends, 1/4, 1/2, 3/4, and the tail ends of the reactors, respectively, and the obtained frequency response curve is shown in fig. 5. It can be seen that the frequency response curve is continuously moving towards high frequency along with the close of the short circuit position to the center between 0.8 and 1.2 MHz. The mutual inductance influence needs to be considered for the turn-to-turn short circuit at different positions, when the short circuit occurs in the middle, the influence of the short circuit turn on other turns is the largest, and the equivalent inductance under the condition is the smallest, so the frequency response curve changes the largest; it can also be seen that the resonance points at different short circuit positions are all shifted to the left in the 1.8-2MHz frequency band.
3) Referring to fig. 6, the method for detecting the turn-to-turn short circuit fault of the reactor based on IFRA and SVM is used for detecting the turn-to-turn short circuit of the same strand and different strands, and the process is as follows:
short-circuit fault points are arranged at the same positions of the inner coil and the outer coil to simulate turn-to-turn short circuit of the same strand and different strands, and the obtained frequency response curve is shown in fig. 6. Within 0.8-1.2MHz, the frequency response curve offset degree of different strand turn-to-turn short circuits is larger, and when the short circuit occurs, the different strand turn-to-turn short circuits can affect more winding branches, so that the reduction amount of the equivalent inductance is larger compared with that of the same strand turn-to-turn short circuit; in addition, the frequency response curve of the outer part of the same strand turn-to-turn short circuit is more deviated than that of the inner coil, because the radius of the outer coil is large, the inductance of each turn of the coil is larger, the reduction of the inductance in the turn-to-turn short circuit is also larger, and the frequency response curve is more shifted to high frequency. In addition, the resonant point frequency under short circuit within 1.8-2MHz is also shifted to the left.
4) The result of the detection
The diagnosis using the support vector machine may be implemented by computer programming. First, 40 samples of each fault are obtained, wherein the number of training samples is 20, the number of test samples is 20, and then mathematical statistical indexes of the frequency response curve, including Correlation Coefficient (CC), Euclidean Distance (ED), sum of absolute values of logarithmic errors (ASLE), and Sum of Squared Errors (SSE), are calculated respectively. And calculating according to the following formula to obtain the mathematical statistical index of the measured frequency response data. And finally, inputting the sample data into the SVM for fault diagnosis.
Fig. 7-9 show the diagnostic results for inter-turn short circuit level, inter-turn short circuit location, inter-turn short circuit for the same strand and for different strands, respectively. Wherein the same inter-turn short circuit in fig. 9 includes both the case of an external same inter-turn short circuit and the case of an internal same inter-turn short circuit. It can be seen that the accuracy of turn-to-turn short circuit degree diagnosis is 100%. The turn-to-turn short position diagnosis accuracy is 98%, there is no error diagnosis at the head end and the tail end, and the error diagnosis is mainly concentrated at the middle portion and 3/4. The turn-to-turn short circuit diagnosis accuracy of the same strand and different strands is 97.5%, and the statistical results are shown in Table 1.
TABLE 1 diagnostic results
Table 1 Diagnostic results
Overall, the method provided by the invention can be used for well detecting the short-circuit fault in the reactor.
Claims (8)
1. A reactor turn-to-turn short circuit fault detection method based on IFRA and SVM is characterized by comprising the following steps:
1) and building the high-voltage pulse device.
2) The upper computer sets pulse parameters and sends the pulse parameters to the high-voltage pulse device;
3) injecting excitation pulse voltage to the end part of the winding of the reactor to be tested by the high-voltage pulse device;
4) a voltage sensor placed at the end part of a winding of the reactor to be tested monitors an input excitation voltage signal in real time, and a current sensor placed at the end part of the winding of the reactor to be tested monitors an output response current signal in real time and sends the output response current signal to a data acquisition module;
5) the data acquisition module is used for respectively processing an input excitation voltage signal and an output response current signal to obtain an excitation voltage digital signal and a response voltage digital signal, and sending the excitation voltage digital signal and the response voltage digital signal to the data processing module;
6) the data processing module respectively carries out fast Fourier transform on the input excitation voltage signal and the output response current signal to obtain a frequency domain representation U of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f)；
7) The data processing module represents U according to the frequency domain of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f) Calculating a frequency response transfer function TF, and drawing an actually measured frequency response curve;
8) the data processing module calculates the digital statistical index of the actually measured frequency response curve according to the reference frequency response curve and the actually measured frequency response curve;
9) the data processing module stores an SVM classifier, inputs the digital statistical indexes into the SVM classifier and finishes turn-to-turn short circuit fault judgment.
2. The IFRA and SVM based reactor turn-to-turn short circuit fault detection method according to claim 1, characterized in that: the high-voltage pulse device comprises a high-voltage direct-current power supply, an FPGA module, a charging protection resistor, a pulse generating circuit, a control circuit and a load resistor;
the high-voltage direct-current power supply charges a capacitor in the pulse forming unit through a charging protection resistor;
the FPGA module receives pulse parameters sent by an upper computer and generates a switch control signal;
the control circuit receives the switch control signal and controls the on-off of a switch in the pulse generating circuit to enable the pulse generating circuit to generate pulse voltage;
the pulse generating circuit comprises a capacitor and a switch which are connected in series;
and the pulse generating circuit injects excitation pulse voltage to the end part of the winding of the reactor to be tested through the load resistor.
3. The IFRA and SVM based reactor turn-to-turn short circuit fault detection method according to claim 1 or 2, characterized in that: the step that the data acquisition module respectively processes the input excitation voltage signal and the output response current signal comprises the following steps:
1) performing analog-to-digital conversion on an input excitation voltage signal to obtain an excitation voltage digital signal;
2) and converting the output response current signal into an output response voltage signal, and performing analog-to-digital conversion to obtain a response voltage digital signal.
4. The IFRA and SVM based reactor turn-to-turn short circuit fault detection method according to claim 1 or 3, characterized in that: the frequency response transfer function TF is as follows:
in the formula I0(f) To output a frequency domain representation of the response current signal; u shapei(f)Is a frequency domain representation of the input excitation voltage signal.
5. The IFRA and SVM based reactor turn-to-turn short circuit fault detection method according to claim 1, characterized in that: the digital statistical indexes comprise a correlation coefficient CC, an Euclidean distance ED, an absolute value sum ASLE of a logarithmic error and a square sum SSE of the error;
wherein, the correlation coefficient CC is as follows:
wherein X (i) is a transfer function amplitude sequence of the measured frequency response curve; n is the sequence length; y (i) is a transfer function amplitude sequence of a reference frequency response curve;
the Euclidean distance ED is as follows:
the absolute value of the log error and ASLE are as follows:
the sum of squared errors SSE is shown below:
in the formula, i is an arbitrary sequence point.
6. The IFRA and SVM based reactor turn-to-turn short circuit fault detection method according to claim 1, characterized in that: the SVM classifier is obtained by training different types of training data; the training data types comprise digital statistical indexes of frequency response curves of the reactor under short circuits of different degrees, digital statistical indexes of frequency response curves of the reactor under short circuits of different positions, digital statistical indexes of frequency response curves of the reactor under short circuits of the same turn-to-turn of the same strand and digital statistical indexes of frequency response curves of the reactor under short circuits of different turn-to-turn of the different strand.
7. The IFRA and SVM based reactor turn-to-turn short circuit fault detection method according to claim 1, wherein the training data acquisition step comprises:
1) a resistor with the resistance value of R is in short circuit between any two turns of coils of the reactor winding;
2) the high-voltage pulse device sends excitation pulse voltage to the end part of the reactor winding in the step 1);
3) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module;
5) updating the resistance value R' ═ R + delta R, and returning to the step 1) until n is collected1A group excitation voltage digital signal and a response voltage digital signal; Δ R is the resistance increment;
6) setting a short-circuit fault point L away from the head end of the reactor winding; the initial value of L is 0;
7) the high-voltage pulse device sends excitation pulse voltage to the end part of the reactor winding in the step 6);
8) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module;
9) update L ═ L + Δ L, and return to step 6) until n is collected2A group excitation voltage digital signal and a response voltage digital signal; Δ L is the distance increment;
10) setting short-circuit fault points among coils with the same strand of the reactor winding;
11) injecting excitation pulse voltage into the end part of the reactor winding in the step 10) by the high-voltage pulse device;
12) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module;
13) setting short-circuit fault points among different strands of coils of the reactor winding;
14) injecting excitation pulse voltage into the end part of the reactor winding in the step 10) by the high-voltage pulse device;
15) the voltage sensor monitors an input excitation voltage signal in real time, and the current sensor monitors an output response current signal in real time and sends the output response current signal to the data acquisition module;
16) the data acquisition module is used for respectively processing an input excitation voltage signal and an output response current signal to obtain an excitation voltage digital signal and a response voltage digital signal, and sending the excitation voltage digital signal and the response voltage digital signal to the data processing module;
17) the data processing module respectively carries out fast Fourier transform on the input excitation voltage signal and the output response current signal to obtain a frequency domain representation U of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f)；
18) The data processing module represents U according to the frequency domain of the input excitation voltage signali(f) And outputting a frequency domain representation I of the response current signal0(f) Calculating a frequency response transfer function TF, and respectively drawing an actually measured frequency response curve of the reactor under short circuits of different degrees, short circuits of different positions, short circuits between turns of the same strand and short circuits between turns of different strands;
19) and the data processing module calculates the digital statistical index of each measured frequency response curve and uses the digital statistical index as training data of the SVM classifier.
8. The IFRA and SVM based reactor turn-to-turn short circuit fault detection method according to claim 1, wherein the turn-to-turn short circuit fault determination content includes a short circuit degree, a short circuit position, and whether the same strand turn-to-turn short circuit exists.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110183247.1A CN113238165A (en) | 2021-02-09 | 2021-02-09 | Reactor turn-to-turn short circuit fault detection method based on IFRA and SVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110183247.1A CN113238165A (en) | 2021-02-09 | 2021-02-09 | Reactor turn-to-turn short circuit fault detection method based on IFRA and SVM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113238165A true CN113238165A (en) | 2021-08-10 |
Family
ID=77130134
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110183247.1A Pending CN113238165A (en) | 2021-02-09 | 2021-02-09 | Reactor turn-to-turn short circuit fault detection method based on IFRA and SVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113238165A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114019298A (en) * | 2021-09-28 | 2022-02-08 | 中电华创(苏州)电力技术研究有限公司 | PCC-SVM-based generator rotor turn-to-turn short circuit online monitoring method |
| CN114994567A (en) * | 2022-06-21 | 2022-09-02 | 国网四川省电力公司电力科学研究院 | Dry-type hollow parallel reactor turn-to-turn short circuit fault online detection method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW201023578A (en) * | 2008-12-15 | 2010-06-16 | Univ Nat Taiwan Science Tech | Method and system for detecting a phishing webpage |
| CN110824389A (en) * | 2019-11-19 | 2020-02-21 | 西南大学 | IFRA-based synchronous generator winding short-circuit fault detection method |
| CN110850340A (en) * | 2019-11-19 | 2020-02-28 | 西南大学 | Transformer winding IFRA curve denoising method based on wavelet transformation |
| CN111400930A (en) * | 2020-04-09 | 2020-07-10 | 武汉大学 | Power equipment small sample fault diagnosis method and system based on virtual and real twin space |
| CN111881781A (en) * | 2020-07-09 | 2020-11-03 | 三峡大学 | Transformer winding deformation classification method based on scanning impedance method and support vector machine |
-
2021
- 2021-02-09 CN CN202110183247.1A patent/CN113238165A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW201023578A (en) * | 2008-12-15 | 2010-06-16 | Univ Nat Taiwan Science Tech | Method and system for detecting a phishing webpage |
| CN110824389A (en) * | 2019-11-19 | 2020-02-21 | 西南大学 | IFRA-based synchronous generator winding short-circuit fault detection method |
| CN110850340A (en) * | 2019-11-19 | 2020-02-28 | 西南大学 | Transformer winding IFRA curve denoising method based on wavelet transformation |
| CN111400930A (en) * | 2020-04-09 | 2020-07-10 | 武汉大学 | Power equipment small sample fault diagnosis method and system based on virtual and real twin space |
| CN111881781A (en) * | 2020-07-09 | 2020-11-03 | 三峡大学 | Transformer winding deformation classification method based on scanning impedance method and support vector machine |
Non-Patent Citations (2)
| Title |
|---|
| S.A. RYDER: "Diagnosing transformer faults using frequency response analysis", 《IEEE ELECTRICAL INSULATION MAGAZINE》, vol. 19, no. 2, XP011096025, DOI: 10.1109/MEI.2003.1192032 * |
| 赵丹: "基于频响曲线特征的变压器绕组变形诊断方法研究", 《中国优秀硕士学位论文全文数据库工程科技II辑》, no. 1, pages 042 - 837 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114019298A (en) * | 2021-09-28 | 2022-02-08 | 中电华创(苏州)电力技术研究有限公司 | PCC-SVM-based generator rotor turn-to-turn short circuit online monitoring method |
| CN114019298B (en) * | 2021-09-28 | 2023-12-05 | 中电华创(苏州)电力技术研究有限公司 | On-line monitoring method for turn-to-turn short circuit of generator rotor based on PCC-SVM |
| CN114994567A (en) * | 2022-06-21 | 2022-09-02 | 国网四川省电力公司电力科学研究院 | Dry-type hollow parallel reactor turn-to-turn short circuit fault online detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106443316B (en) | Multi-information detection method and device for deformation state of power transformer winding | |
| Yao et al. | Transformer winding deformation diagnostic system using online high frequency signal injection by capacitive coupling | |
| CN113238165A (en) | Reactor turn-to-turn short circuit fault detection method based on IFRA and SVM | |
| CN110470465B (en) | Circuit breaker testing method and system based on vibration signal analysis | |
| CN109932053B (en) | State monitoring device and method for high-voltage shunt reactor | |
| CN110824389A (en) | IFRA-based synchronous generator winding short-circuit fault detection method | |
| CN106526409B (en) | A kind of deformation of transformer winding live testing system and its test method | |
| CN108802566B (en) | Grounding line selection method based on HHT signal analysis | |
| WO2019219528A1 (en) | Method and system for fault detection | |
| CN113189515A (en) | IFRA magnetic coupling-based reactor turn-to-turn short circuit fault non-contact type online detection method | |
| CN106124081B (en) | The accurate multiple spot real time temperature measurement method and system of permanent magnet synchronous motor | |
| CN104991160A (en) | Transformer winding fault detection and diagnosis method based on voltage and current characteristics | |
| CN112505415B (en) | Porcelain insulator resistance measuring device and method | |
| CN110929673A (en) | Transformer winding vibration signal identification method based on ITD (inverse discrete cosine transformation) permutation entropy and CGWO-SVM (Carrier-support vector machine) | |
| Naderi et al. | Modeling and detection of transformer internal incipient fault during impulse test | |
| CN112557732A (en) | Magnitude tracing method and system of impulse current measuring device based on induction coil | |
| CN117214777A (en) | Fault positioning method, system, equipment and storage medium for transformer winding | |
| Chen et al. | A new broadband microcurrent transducer for insulator leakage current monitoring system | |
| CN107024613A (en) | High-voltage feedback power cable square wave on-line overvoltage monitor | |
| CN108008262B (en) | Rapid partial discharge detection method | |
| CN111458626B (en) | Circuit system signal analysis and processing method based on symbiotic multi-functional computation | |
| CN115047296A (en) | Power distribution network fault section positioning method | |
| CN113064067B (en) | Frequency detection circuit and method for rapidly calculating electric parameters of motor | |
| Luo et al. | A test method of winding deformation excited by pseudorandom M-Sequences—Part I: Theory and simulation | |
| CN114236254A (en) | Direct current source noise measurement system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |