Health evaluation method for onboard converter of electric locomotive

Abstract

The method for evaluating the health of an onboard converter of an electric locomotive disclosed in the present invention includes the concept of converter output discreteness and its calculation method, and the concept of converter output current discreteness offset index and its calculation method. By measuring the output current discreteness and its offset index, converter fault pre-diagnosis is achieved simply and effectively. The output current discreteness instability threshold evaluation index calculated from the collected data conforms to the normal distribution law. It is a statistical conclusion based on a large amount of data from the railway system locomotive depot, and includes various uncertain factors that may cause converter degradation, which is helpful for the pre-diagnosis of converter faults of electric locomotives. This method is universal for electric locomotives.

Description

Technical Field

[0001] The invention belongs to the technical field of electric locomotive converters in railway traction power supply systems, and in particular relates to a method for evaluating the health of an on-board converter of an electric locomotive. Background Art

[0002] As a key component in the railway traction power supply system, the reliability of the on-board converter is crucial to ensure the safe and efficient operation of the railway system. However, due to the complex operating conditions of electric locomotives and the extremely harsh working environment they often face, the on-board converter system inevitably has parameter perturbations and various nonlinear uncertain disturbance factors, which are prone to various transient and steady-state instability problems, resulting in reduced reliability of the on-board converter.

[0003] Therefore, on-board converter fault pre-diagnosis and health management have become one of the research hotspots of traction power supply system. At present, the research work mainly focuses on the fault monitoring of component parameters, mainly power devices such as IGBT. However, due to the complexity and packaging of converters, it is difficult to measure the internal device parameters, and the measurement results show certain coupling and chaos, which greatly limits the further application of converter fault pre-diagnosis technology. Summary of the invention

[0004] The purpose of the present invention is to provide a method for evaluating the health of an on-board converter of an electric locomotive, and to propose a new evaluation index to address the problems of complex and incomplete converter fault measurement parameters and high coupling.

[0005] The technical solution adopted by the present invention is: a method for evaluating the health of an on-board converter of an electric locomotive, comprising the following steps:

[0006] Step 1: Install a data acquisition module in the locomotive converter cabinet to perform a locomotive loading operation test, and use the fixed sampling frequency of the data acquisition module to obtain the original data of the converter operation output current;

[0007] Step 2: pre-process the acquired data, select the U-phase current, V-phase current and W-phase current outputted from the secondary side and group them, and store each group of data for one minute;

[0008] Step 3, extract data, perform zero-crossing detection on the data and remove pseudo zero points;

[0009] Step 4: Calculate the number of input sequence elements required for fast Fourier transform according to the zero-crossing point detection result, and determine the fast Fourier transform sampling rate according to the Nyquist sampling theorem;

[0010] Step 5: Analyze the data using the interpolation fast Fourier algorithm based on the Blackman-Harris window, and correct the result using the correction formula of the Blackman-Harris window to obtain the fundamental frequency, fundamental amplitude and fundamental phase;

[0011] Step 6: Generate a fundamental wave discrete sequence using the obtained fundamental wave frequency, fundamental wave amplitude and fundamental wave phase;

[0012] Step 7: Subtract the original sampling sequence from the corresponding points of the fundamental discrete sequence to obtain the numerical deviation sequence corresponding to each sampling point;

[0013] Step 8: Calculate the current dispersion of the numerical deviation sequence to measure the variation of the output current of the converter;

[0014] Step 9, respectively collecting and calculating the current dispersion of the output current of the converter with a higher failure rate and the current dispersion of the converter with a more stable output current;

[0015] Step 10, calculating the two current dispersion offsets in step 9;

[0016] Step 11: Statistically calculate the current dispersion offsets obtained from different data groups to obtain the current output dispersion instability threshold λ of the converter. 0.5 ;

[0017] Step 12: Evaluate the stability of the converter. If the output current dispersion offset of the converter is greater than λ 0.5 , the probability of the converter failure is high; if the measured output current dispersion offset of the converter is less than λ 0.5 , the converter health is good.

[0018] The present invention is also characterized in that:

[0019] Before obtaining the original data of the converter operating output current with a fixed sampling frequency in step 1, the locomotive operation time is at least 10 minutes.

[0020] The zero-crossing detection of the data in step 3 specifically includes: assuming that a sampling point of a group of data is Sample[i], the value range of i is [0, (N-1)], and N is the number of discrete data in 1 minute; if Sample[i] satisfies the conditions: Sample[i] is less than 0 and Sample[i+1] to Sample[i+C] is greater than 0, or Sample[i] is greater than 0 and Sample[i-1] to Sample[i+C] is less than 0, then the point is judged to be a zero-crossing point, and the two adjacent zero-crossing points are half of the power frequency cycle time, which is the zero point offset number threshold, and C is not less than 3.

[0021] The pseudo zero point removed in step 3 is a point where the discrete point of the output signal suddenly rises or drops due to high-order harmonic interference.

[0022] The specific method of calculating the number of input sequence elements required for the fast Fourier transform based on the zero-crossing point detection result in step 4 is: if Sample[i1] is the first zero-crossing point and Sample[i2] is the second zero-crossing point, then the number of input sequence elements required for the fast Fourier transform is 2*(i2-i1+1).

[0023] The time domain expression of the interpolation fast Fourier algorithm based on the Blackman-Harris window in step 5 is:

[0024] w bh (n)=0.35875-0.48829cos(2πn / N)+0.14128cos(4πn / N)-0.01168cos(6πn / N)

[0025] Where n = 0, 1, 2…, N-1, and the input sequence is W(n), then the discrete input sequence after the window function is W(n) = W(n)·w bh (n).

[0026] The correction formula adopted in step 5 is:

[0027] v(α)=2.61979085α+0.2865675α 3 +0.1283α 5 +0.08024α 7

[0028] In which, assuming that k0 is the peak point of the spectrum, y1 and y2 are the amplitudes corresponding to the maximum and second maximum discrete spectrum points k1 and k2 respectively, α=k0-k1-0.5, β=(y2-y1) / (y2+y1).

[0029] Step 8 The calculation formula for current dispersion is:

[0030]

[0031] Where δ is the current dispersion, x i is the discrete point value of the sampling signal, is the average value of this group of values.

[0032] The calculation formula of the current dispersion offset in step 10 is:

[0033]

[0034] Among them, δ e is the output current dispersion of the electric locomotive converter with a high failure rate, δs To meet the inverter output current dispersion requirements of electric locomotive maintenance regulations.

[0035] The beneficial effects of the present invention are as follows: the health evaluation method of the on-board converter of an electric locomotive of the present invention proposes a new evaluation index to address the problems of complex and incomplete parameters and high coupling in current converter fault measurement, and considers the complex converter device as a black box and a two-terminal system, thereby avoiding measuring the complex parameters of the internal components of the converter and ignoring the chaos caused by multi-parameter and multi-dimensional coupling. By measuring the concept of output current discreteness and its offset index, the converter fault pre-diagnosis and health management can be simply and effectively achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flow chart of a current dispersion calculation method in a method for evaluating the health of an on-board converter of an electric locomotive according to the present invention;

[0037] Figure 2 It is a flow chart of the current dispersion offset calculation and converter health evaluation method in the electric locomotive on-board converter health evaluation method of the present invention;

[0038] Figure 3 It is a schematic diagram of the implementation status of the electric locomotive on-board converter health evaluation method of the present invention. DETAILED DESCRIPTION

[0039] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] The present invention provides a method for evaluating the health of an onboard converter of an electric locomotive. Figure 1 and Figure 2 As shown, the following steps are included:

[0041] Step 1: Install a data acquisition module in the locomotive converter cabinet to conduct a locomotive loading operation test, and obtain the original data of the converter output current using a fixed sampling frequency. In the specific implementation process, the locomotive operation time is at least 10 minutes;

[0042] Step 2: pre-process the acquired data, select the U-phase current, V-phase current and W-phase current outputted from the secondary side and group them, and store each group of data for one minute;

[0043] Step 3, extract data and perform zero-crossing detection on the data. Suppose a sampling point of a certain set of data is Sample[i], the value range of i is [0, (N-1)], N is the number of discrete data in 1 minute, if the sampling frequency is 12.8kHz (that is, the number of sampling points in each power frequency cycle is 256), then N = 768000, if Sample[i] meets the conditions: Sample[i] is less than 0 and Sample[i+1] to Sample[i+C] is greater than 0, or Sample[i] is greater than 0 and Sample[i-1] to Sample[i+C] is less than 0, then the point is judged to be a zero-crossing point, and the two adjacent zero-crossing points are half of the power frequency cycle time, which is the zero point offset number threshold, and C is generally not less than 3. It is worth noting that it is necessary to remove pseudo-zero points, that is, points where the discrete points of the output signal rise or fall sharply due to high-order harmonic interference.

[0044] Step 4: Calculate the number of input sequence elements required for fast Fourier transform according to the zero-crossing point detection result. If Sample[i1] is the first zero-crossing point and Sample[i2] is the second zero-crossing point, the number of data required for FFT is 2*(i2-i1+1). Determine the appropriate FFT sampling rate according to the Nyquist sampling theorem. The sampling frequency of the fast Fourier transform sampling rate should be greater than twice the signal frequency. The number of sequence elements of the fast Fourier transform input is determined according to the resolution requirement.

[0045] Step 5: Since it is difficult to achieve synchronous sampling and full cycle truncation, there may be spectrum leakage and fence effect when using the FFT algorithm to analyze the signal. In order to improve the analysis accuracy, the analysis sequence is processed by adding the Blackman-Harris window function, and its time domain expression is:

[0046] w bh (n)=0.35875-0.48829cos(2πn / N)+0.14128cos(4πn / N)-0.01168cos(6πn / N)

[0047] Where n = 0, 1, 2…, N-1, let the discrete signal sequence be W(n), then the discrete input sequence after adding the window function is W(n) = W(n)·w bh (n).

[0048] After FFT transformation, the fundamental frequency, fundamental amplitude, and fundamental phase are obtained, and the results are corrected using the interpolation correction formula. The frequency double spectrum correction formula is:

[0049] v(α)=2.61979085α+0.2865675α 3 +0.1283α 5 +0.08024α7

[0050] In which, assuming that k0 is the peak point of the spectrum, y1 and y2 are the amplitudes corresponding to the maximum and second maximum discrete spectrum points k1 and k2 respectively, α=k0-k1-0.5, β=(y2-y1) / (y2+y1).

[0051] Step 6: Generate a discrete sequence of fundamental wave signals, funda-wave[N], by using the fundamental wave frequency, amplitude and phase of the obtained signal;

[0052] Step 7: Subtract the original sampling sequence from the corresponding points of the fundamental discrete sequence to obtain the numerical deviation sequence error[N] corresponding to each sampling point;

[0053] Step 8: Calculate the current dispersion of the deviation sequence to measure the variation of the output current of the converter. The formula has the same physical meaning for the current calculation results of different amplitudes. The current dispersion calculation formula is:

[0054]

[0055] Where δ is the current dispersion, x i is the discrete point value of the sampling signal, is the average value of this group of values.

[0056] Step 9. Collect data on electric locomotives with higher failure rates, such as IGBT power modules that are prone to explosion, and electric locomotives with relatively stable power modules. Since the converters of different models are different, the original data collected for comparison must be based on the same model, for example, all from the HXD1 series electric locomotives. Calculate the current discrete rate of the output current of the converter with a higher failure rate and the current discrete rate of the relatively stable converter.

[0057] Step 10: Calculate the current dispersion offset of the two data. The calculation formula is:

[0058]

[0059] Among them, δ e is the output current dispersion of the electric locomotive converter with a high failure rate, δ s To meet the inverter output current dispersion requirements of electric locomotive maintenance regulations.

[0060] Step 11: Statistically calculate the current dispersion offsets of different data groups to obtain the instability threshold of the current output dispersion of the converter. For a large number of calculated λ, it is in accordance with the normal distribution law. The median λ is taken from the results. 0.5 As a measure of converter health, the result is a statistical conclusion that includes various conditions that may cause converter degradation. The result is universal for a specific vehicle model.

[0061] Step 12: Converter stability evaluation: if the converter output current dispersion offset is detected to be greater than λ 0.5 , indicating that the probability of the vehicle inverter device failure is high. If the measured value is less than λ 0.5 , indicating that the vehicle's converter device is in good health.

[0062] Through the above-mentioned manner, the electric locomotive on-board converter health evaluation method of the present invention is used to evaluate the converter health, regards the complex converter device as a simple two-terminal system, and only considers the input and output of the system, avoiding the measurement of complex parameters of the internal components of the converter, and ignoring the chaos caused by multi-parameter and multi-dimensional coupling. By measuring the output current discreteness and its offset index, the converter fault pre-diagnosis and health management are simply and effectively realized, providing theoretical guidance for the maintenance of electric locomotive converters, and ensuring the safe, reliable and efficient operation of the traction power supply system.

[0063] In the present invention, the current dispersion offset obtained by collecting data is statistically analyzed to obtain the instability threshold of the current output dispersion of the converter, which conforms to the normal distribution law. The measurement standard λ of the converter health 0.5 It is a statistical conclusion, which includes various uncertain factors that may cause converter degradation, and is helpful for the pre-diagnosis of converter faults in electric locomotives. This method is universal for electric locomotives.

[0064] Example

[0065] Taking the health status evaluation method of HXD1 electric locomotive converter as an example, the system structure principle is explained: Figure 3 As shown: a high-speed acquisition unit is installed in the converter device, and a loaded operation test of an electric locomotive is carried out to collect the converter output current signal. The sampling frequency is set to 12800 Hz. During the specific implementation process, the locomotive running time is 15 minutes.

[0066] The acquired raw data is stored in the SD card in groups according to the phase sequence current. The data is read and transferred using software. The data length is stored as one minute. The software sampling frequency is f s Set to 12800Hz.

[0067] The extracted data is detected by software. Assuming that the time of two adjacent zero crossings is T1 and T2, and the cycle length is 2 (T1-T2) seconds, the sequence length that needs to be analyzed within a waveform period T can be calculated as T (1 / f s ), if the result contains a decimal part, the decimal part is discarded directly. In this case, the maximum error is 0.05%.

[0068] In the specific implementation of the example, the software's detection condition for the zero-crossing point is: if the value of a certain sampling point is equal to 0, and the first three points are greater than 0 and the last three points are less than 0; or the value of a certain sampling point is equal to zero, and the first three points are less than 0 and the last three points are greater than 0, then the point is judged to be a zero-crossing point.

[0069] The interpolation fast Fourier algorithm based on the Blackman-Harris window is used to analyze the data. In the specific implementation of the example, the total sequence length obtained in units of 4 cycles is used as the number of Fourier transform input sequences. After the transformation, the fundamental frequency, fundamental amplitude, and fundamental phase are obtained, and the results are corrected using the interpolation correction formula. In this way, all the parameters of the fundamental signal are obtained, the fundamental signal is fitted with data, and then discretized with the original signal frequency of 12800Hz. In the specific implementation of the example, MATLAB software is used for data fitting.

[0070] The original data discrete sequence is subtracted from the corresponding points of the fundamental wave discrete sequence to obtain the numerical deviation sequence corresponding to each sampling point, and the current dispersion of the deviation sequence can be calculated according to the current dispersion calculation formula.

[0071] The current dispersion offsets obtained from different data groups are statistically analyzed to obtain the instability threshold of the current output dispersion of the converter. The median λ is taken from the results. 0.5 As a measure of converter health,

[0072] If the output current dispersion of the converter is detected to be greater than λ 0.5 , indicating that the locomotive onboard converter device has a high probability of failure and needs to be reported for maintenance; if the measured value is less than λ 0.5 , indicating that the locomotive's onboard converter device is in good health and this indicator has passed the inspection.

Claims

1. A method for evaluating the health of an onboard converter of an electric locomotive, characterized in that: The following steps are involved: Step 1: Install a data acquisition module in the locomotive converter cabinet to perform a locomotive loading operation test, and use the fixed sampling frequency of the data acquisition module to obtain the original data of the converter operation output current; Step 2: pre-process the acquired data, select the U-phase current, V-phase current and W-phase current outputted from the secondary side and group them, and store each group of data for one minute; Step 3, extract data, perform zero-crossing detection on the data and remove pseudo zero points; Step 4: Calculate the number of input sequence elements required for fast Fourier transform according to the zero-crossing point detection result, and determine the fast Fourier transform sampling rate according to the Nyquist sampling theorem; Step 5: Analyze the data using the interpolation fast Fourier algorithm based on the Blackman-Harris window, and correct the result using the correction formula of the Blackman-Harris window to obtain the fundamental frequency, fundamental amplitude and fundamental phase; Step 6: Generate a fundamental wave discrete sequence using the obtained fundamental wave frequency, fundamental wave amplitude and fundamental wave phase; Step 7: Subtract the original sampling sequence from the corresponding points of the fundamental discrete sequence to obtain the numerical deviation sequence corresponding to each sampling point; Step 8: Calculate the current dispersion of the numerical deviation sequence to measure the variation of the output current of the converter; Step 9, respectively collecting and calculating the current dispersion of the output current of the converter with a higher failure rate and the current dispersion of the converter with a more stable output current; Step 10, calculating the two current dispersion offsets in step 9; Step 11: Statistically calculate the current dispersion offsets obtained from different data groups to obtain the current output dispersion instability threshold of the converter. ; Step 12: Evaluate the stability of the converter. If the converter output current dispersion offset is greater than , the probability of the converter failure is high; if the measured converter output current discreteness offset is less than , the converter health is good.

2. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 1, characterized in that: Before obtaining the original data of the converter operating output current with a fixed sampling frequency in step 1, the locomotive operation time is at least 10 minutes.

3. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 1, characterized in that: The zero-crossing detection of the data in step 3 specifically includes: assuming that a sampling point of a certain group of data is Sample[i], the value range of i is [0, (N-1)], and N is the number of discrete data in 1 minute; if Sample[i] satisfies the conditions: Sample[i] is less than 0 and Sample[i+1] to Sample[i+C] is greater than 0, or Sample[i] is greater than 0 and Sample[i-1] to Sample[i+C] is less than 0, then the point is judged to be a zero-crossing point, and the two adjacent zero-crossing points are half of the corresponding power frequency cycle time, which is the zero point offset number threshold, and the value of C is not less than 3.

4. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 1, characterized in that: The pseudo zero point removed in step 3 is a point where the discrete point of the output signal suddenly rises or drops due to high-order harmonic interference.

5. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 3, characterized in that: The specific method of calculating the number of input sequence elements required for the fast Fourier transform according to the zero-crossing point detection result in step 4 is: if Sample[i1] is the first zero-crossing point and Sample[i2] is the second zero-crossing point, then the number of input sequence elements required for the fast Fourier transform is 2*(i2-i1+1).

6. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 1, characterized in that: The time domain expression of the interpolation fast Fourier algorithm based on the Blackman-Harris window in step 5 is: in, , let the input sequence be , then the discrete input sequence after the window function is .

7. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 1, characterized in that: The correction formula adopted in step 5 is: Among them, assuming is the peak point of the spectrum, and are the maximum and second maximum discrete spectrum points respectively. and The corresponding amplitude, , .

8. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 1, characterized in that: The calculation formula of the current dispersion in step 8 is: in, is the current dispersion, is the discrete point value of the sampling signal, is the average value of the sampled data set.

9. The method for evaluating the health of an onboard converter of an electric locomotive according to claim 1, characterized in that: The calculation formula of the current dispersion offset in step 10 is: in, The output current dispersion of the electric locomotive converter with a high failure rate is To meet the inverter output current dispersion requirements of electric locomotive maintenance regulations.

Images