Harmonic Detection Method and System for a Vehicle Network Electrical System

Abstract

The present invention provides a harmonic detection method and system for a vehicle-grid electrical system, which relates to the technical field of harmonic detection. The harmonic detection method for the vehicle-grid electrical system includes: acquiring a current signal of a vehicle-grid busbar and performing band-limiting processing on the current signal; constructing a variable window width S-transform kernel function with frequency adaptability; extracting harmonic phase coupling characteristics based on bispectrum analysis of the third-order cumulant, and generating an enhanced spectrum through principal component projection; generating a dynamic detection threshold by using a background noise model in a harmonic-free interval; and performing harmonic parameter fitting by using an exponential decay model with prior constraints in the frequency domain. The method of the present invention realizes the full-process closed-loop optimization from signal preprocessing to parameter optimization through a cascaded architecture of adaptive time-frequency analysis, high-order statistical feature extraction, dynamic threshold detection, and prior constraint fitting, and is particularly suitable for the precise detection of high-frequency, nonlinear, and time-varying harmonics in the vehicle-grid system, and can provide a more reliable diagnostic basis for power quality governance.

Description

Technical Field

[0001] The present invention relates to the technical field of harmonic detection, and particularly relates to a harmonic detection method and system for a vehicle-grid electrical system. Background Art

[0002] With the rapid development of rail transit, the widespread application of power electronic devices in vehicle-grid electrical systems (such as traction power supply systems) has led to an increasingly prominent problem of harmonic pollution. For example, high-frequency switching devices (such as IGBTs), non-linear loads (such as rectifiers, inverters), etc. will generate complex broadband harmonics, which not only affect the power quality, but may also cause equipment overheating, resonance, and even system instability.

[0003] In the existing related technologies, the current harmonic detection schemes have the following problems:

[0004] First, the resolution of high-frequency harmonic detection is insufficient. The traditional Fourier transform (FFT) has poor adaptability to non-stationary signals, and the fixed time-frequency window of the short-time Fourier transform (STFT) is difficult to balance the low-frequency resolution and the high-frequency time localization ability.

[0005] Second, it is difficult to extract non-linear harmonic features. The harmonics generated by non-linear loads have phase coupling characteristics (such as sum-frequency / difference-frequency components), and conventional power spectrum analysis cannot capture such high-order statistical features.

[0006] Third, the set threshold has poor adaptability. Existing methods mostly use fixed thresholds or static models based on empirical values, and it is difficult to cope with the time-varying background noise (such as pulse interference, random fluctuations) in the vehicle-grid system.

[0007] Fourth, the parameter fitting error accumulates. Traditional harmonic decomposition methods (such as Prony algorithm) are easily affected by noise, resulting in parameter estimation deviation. Summary of the Invention

[0008] In order to at least solve some of the technical problems in the related technologies, the present invention provides a harmonic detection method and system for a vehicle-grid electrical system.

[0009] In order to achieve the above purpose, the technical solutions adopted by the present invention include:

[0010] According to the first aspect of the present invention, a harmonic detection method for a vehicle-grid electrical system is provided, including the following steps:

[0011] Step S1: Obtain the current signal of the vehicle-grid busbar and perform band-limiting processing on the current signal;

[0012] Step S2: Construct a variable window width S-transform kernel function with frequency adaptability;

[0013] Step S3: Extract harmonic phase coupling features based on bispectrum analysis of third-order cumulants, and generate an enhanced spectrum through principal component projection;

[0014] Step S4: Generate a dynamic detection threshold using a background noise model in the non-harmonic interval;

[0015] Step S5: Perform harmonic parameter fitting using an exponential decay model with frequency domain prior constraints.

[0016] Optionally, step S1 specifically includes:

[0017] Step S1-1: Collect the current signal of the vehicle-grid bus using a broadband current sensor;

[0018] Step S1-2: Perform band-limiting processing through an anti-aliasing filter:

[0019]

[0020] where, is the complex frequency domain transfer function of the filter, is the complex frequency variable, is the cut-off angular frequency, is the filter order.

[0021] Optionally, the cut-off angular frequency is set to ; the filter order is set to .

[0022] Optionally, step S2 specifically includes:

[0023] Construct a frequency-adaptive variable window width S-transform kernel function according to the following formula:

[0024]

[0025] where, is the time-frequency matrix, is the instantaneous value of the current signal at time , is the frequency-adaptive window width adjustment function, , is the window width proportionality coefficient, is the minimum frequency offset, is the Fourier basis function.

[0026] Optionally, in step S2, it further includes:

[0027] When the detected , the frequency-adaptive window width adjustment function is adjusted to:

[0028]

[0029] wherein, is the frequency judgment threshold.

[0030] Optionally, step S3 specifically includes:

[0031] Step S3-1: Calculate the third-order cumulant:

[0032]

[0033] Step S3-2: Construct the bispectrum feature matrix:

[0034]

[0035] Step S3-3: Implement the principal component projection:

[0036]

[0037] In the formula, is the third-order cumulant of the signal, is the mathematical expectation, and are the time delay amounts, is the instant value of the current signal at time , is the observed value of the current signal shifted to the right by sampling intervals based on the time point , is the observed value of the current signal shifted to the right by sampling intervals based on the time point , is the bispectrum value, and are two independent frequency variables, is the projected spectrum.

[0038] Optionally, step S4 specifically includes:

[0039] Step S4-1: Establish the background noise model:

[0040]

[0041] Step S4-2: Calculate the dynamic threshold:

[0042]

[0043] In the formula, is the estimated value of the background noise standard deviation at frequency , is the number of sampling points in the non-harmonic interval, where the time-frequency matrix is at the frequency , time at the amplitude, ∈ no-harmonic interval, is the detection threshold at the frequency , is the empirical coefficient, is the false alarm rate setting value.

[0044] Optionally, in the step S4, the empirical coefficient is set to , and the false alarm rate setting value is set to .

[0045] Optionally, the step S5 specifically includes:

[0046] Step S5-1: Perform parameter fitting using the following algorithm:

[0047]

[0048] Step S5-2: Optimize the constraint conditions according to the following formula:

[0049]

[0050] In the formula, is the amplitude of the th harmonic, is the attenuation factor, is the th harmonic frequency, is the initial phase, is the model order, is the number of sampling points, is the index value of the sampling point, , is the sampling interval, represents the predicted value of the current signal reconstructed by the algorithm at the th sampling point, is the regularization coefficient, is the prior estimate value of the harmonic frequency obtained according to steps S2 to S4.

[0051] According to the second aspect of the present invention, there is also provided a harmonic detection system for a vehicle-grid electrical system, which is used to execute the harmonic detection method for a vehicle-grid electrical system described in any one of the technical solutions in the first aspect of the present invention. The harmonic detection system for a vehicle-grid electrical system includes:

[0052] A data acquisition and processing module, which is used to acquire the current signal of the vehicle-grid bus and perform band-limiting processing on the current signal;

[0053] An adaptive time-frequency analysis module, which is used to construct a frequency-adaptive variable window width S-transform kernel function;

[0054] Bispectrum feature extraction module, which is used for bispectrum analysis based on third-order cumulants, extracts harmonic phase coupling features, and generates an enhanced spectrum through principal component projection;

[0055] Dynamic threshold generation module, which is used to generate a dynamic detection threshold by using the background noise model in the no-harmonic interval;

[0056] Parameter optimization module, which is used to perform harmonic parameter fitting by using an exponential decay model with frequency-domain prior constraints.

[0057] Beneficial effects:

[0058] 1. Through the above technical solutions, firstly, by performing band-limiting processing on the current signal, it can not only effectively ensure the validity of the Nyquist sampling theorem, but also effectively suppress high-frequency noise aliasing, that is, suppress the interference of high-frequency noise on harmonic detection.

[0059] Secondly, by constructing a frequency-adaptive variable window width S-transform kernel function, it can be well applied to signals in different frequency bands. Among them, for low-frequency band signals, the variable window width S-transform kernel function can effectively ensure its high frequency resolution, and for high-frequency band signals, it can suppress the frequency-domain leakage caused by the widening of the time window through exponential adjustment.

[0060] Thirdly, through the third-order cumulants, the influence of Gaussian noise can be effectively eliminated (this is because the higher-order cumulants of the Gaussian process are zero), the phase coupling features of non-Gaussian harmonic components are extracted to effectively identify interharmonics generated by nonlinear loads and improve the signal-to-noise ratio. At the same time, by generating an enhanced spectrum through principal component projection, the two-dimensional bispectrum can be mapped into a one-dimensional spectrum, highlighting the sum-frequency components of nonlinear harmonics to improve accuracy.

[0061] Fourthly, when statistically averaging the time-frequency amplitude in the no-harmonic interval, a frequency-related noise floor is established to improve the detection sensitivity of high-frequency harmonics, and compared with the fixed threshold method, the false positive rate can be effectively reduced.

[0062] Fifthly, by using an exponential decay model to describe the transient characteristics of harmonics, the fitting deviation of the traditional steady-state model for impulsive harmonics can be effectively avoided. At the same time, by introducing frequency-domain prior constraints, the parameter offset caused by noise can be suppressed.

[0063] Generally speaking, the method of the present invention realizes the full-process closed-loop optimization from signal preprocessing to parameter optimization through a cascaded architecture of adaptive time-frequency analysis (step S2), high-order statistical feature extraction (step S3), dynamic threshold detection (step S4) and prior constraint fitting (step S5), and is particularly suitable for the precise detection of high-frequency, nonlinear, and time-varying harmonics in the vehicle-grid system, and can provide a more reliable diagnostic basis for power quality governance.

[0064] 2. Other beneficial effects or advantages of the present invention will be described in detail in the specific implementation manners. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for description in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0066] Wherein:

[0067] Figure 1 is a schematic flowchart of the steps of the harmonic detection method for the vehicle-grid electrical system provided by an exemplary embodiment of the present invention. SPECIFIC IMPLEMENTATION MANNERS

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0069] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0070] In addition, the terms "include" and "have" and any variations thereof mentioned in the description of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes other steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices. Among them, it should also be noted that in the embodiments of the present invention, words such as "exemplary" or "for example" are used to represent examples, illustrations, or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.

[0071] The following will elaborate on the technical solutions of the present invention with reference to the drawings.

[0072] Embodiment 1

[0073] As shown Figure 1 in the figure, according to the first aspect of the present invention, this embodiment provides a method for detecting harmonics in a vehicle-grid electrical system, including the following steps:

[0074] Step S1: Obtain the current signal of the vehicle-grid busbar and perform band-limiting processing on the current signal;

[0075] Step S2: Construct a variable window width S-transform kernel function with frequency adaptability;

[0076] Step S3: Based on the bispectrum analysis of the third-order cumulant, extract the harmonic phase coupling characteristics, and generate an enhanced spectrum through principal component projection;

[0077] Step S4: Generate a dynamic detection threshold using the background noise model in the harmonic-free interval;

[0078] Step S5: Perform harmonic parameter fitting using an exponential decay model with frequency-domain prior constraints.

[0079] Through the above technical solutions, first, by performing band-limiting processing on the current signal, it can not only effectively ensure the validity of the Nyquist sampling theorem, but also effectively suppress the aliasing of high-frequency noise, that is, suppress the interference of high-frequency noise on harmonic detection.

[0080] Second, by constructing a variable window width S-transform kernel function with frequency adaptability, it can be well applicable to signals in different frequency bands. Among them, for low-frequency band signals, the variable window width S-transform kernel function can effectively ensure its high frequency resolution. For high-frequency band signals, it can suppress the frequency-domain leakage caused by the widening of the time window through exponential adjustment.

[0081] Third, the influence of Gaussian noise can be effectively eliminated through the third-order cumulant (this is because the higher-order cumulants of the Gaussian process are zero), and the phase coupling characteristics of non-Gaussian harmonic components can be extracted to effectively identify the interharmonics generated by nonlinear loads and improve the signal-to-noise ratio. At the same time, by generating an enhanced spectrum through principal component projection, the two-dimensional bispectrum can be mapped into a one-dimensional spectrum, highlighting the sum-frequency components of nonlinear harmonics to improve the accuracy.

[0082] Fourth, by statistically averaging the time-frequency amplitude in the harmonic-free interval, a frequency-related noise floor is established to improve the detection sensitivity of high-frequency harmonics, and compared with the fixed threshold method, the false positive rate can be effectively reduced.

[0083] Fifth, by using the exponential decay model to describe the transient characteristics of harmonics, the fitting deviation of the traditional steady-state model for impulsive harmonics can be effectively avoided. At the same time, by introducing frequency-domain prior constraints, the parameter offset caused by noise can be suppressed.

[0084] Generally speaking, the method of the present invention realizes the full - process closed - loop optimization from signal pre - processing to parameter optimization through a cascaded architecture of adaptive time - frequency analysis (step S2), high - order statistical feature extraction (step S3), dynamic threshold detection (step S4), and prior - constraint fitting (step S5). It is particularly suitable for the precise detection of high - frequency, non - linear, and time - varying harmonics in the vehicle - grid system, and can provide a more reliable diagnostic basis for power quality governance.

[0085] In one embodiment of the present invention, step S1 specifically includes:

[0086] Step S1 - 1: Collect the current signal of the vehicle - grid bus using a broadband current sensor;

[0087] Step S1 - 2: Perform band - limiting processing through an anti - aliasing filter:

[0088]

[0089] Where, is the complex - frequency domain transfer function of the filter (used to describe the amplitude - frequency and phase - frequency responses of the input signal), is the complex - frequency variable, is the cut - off angular frequency (in units of , corresponding to the boundary between the pass - band and stop - band of the filter. For example, 10 kHz corresponds to ), is the filter order (determining the steepness of the transition band. The higher the order, the faster the stop - band attenuation).

[0090] It can be understood that this embodiment is for the signal acquisition and pre - processing environment of harmonic detection in the vehicle - grid electrical system. By combining a broadband current sensor with a high - order anti - aliasing filter, the effective retention of high - frequency harmonics and noise suppression are achieved.

[0091] Specifically, first, a high - order filter with a cut - off frequency of is used. The flatness error of the amplitude - frequency response within the pass - band is smaller, which can ensure that the effective harmonic components from 0 to can be retained without distortion. At the same time, the attenuation rate of the stop - band (greater than ) is larger (taking , as an example, its stop - band attenuation rate can reach - 160 dB / dec), which can effectively suppress high - frequency noise aliasing to avoid interference of false harmonic components in subsequent analysis and improve the accuracy and reliability of the analysis.

[0092] Second, the filter transfer function of the present invention ( ) reduces the ringing effect in the time domain and realizes a steep transition band in the frequency domain through high - order pole configuration, reducing the risk of high - frequency noise folding to the low - frequency band during the sampling process.

[0093] Third, the bandwidth of the broadband current sensor (e.g., Hall effect sensor) covers a wider range (e.g., 0 kHz - 50 kHz), which can meet the requirements of high-frequency harmonic acquisition. Moreover, the signal-to-noise ratio is relatively large (e.g., SNR can be greater than or equal to 80 dB), which can avoid signal saturation or quantization error.

[0094] It can be understood that the pole symmetric distribution characteristic of the transfer function can ensure the maximum flat response within the passband. For example, in an exemplary embodiment, , , whose amplitude-frequency response , near the cut-off frequency (e.g., ), the amplitude will drop rapidly, which can effectively filter out the high-frequency interference generated by the switching device (e.g., IGBT switching noise). At the same time, the present invention collaboratively designs the pre-stage analog filter (anti-aliasing filter) and the digital sampling rate to meet the Nyquist sampling theorem. For example, if the highest harmonic frequency for target detection is 10 kHz, the sampling rate needs to be ≥20 kHz (the actual selection can be above 50 kHz). Combining with the stopband attenuation characteristic of the filter, the aliasing component is suppressed below the system noise floor.

[0095] In an embodiment of the present invention, the cut-off angular frequency can be set to ; the filter order can be set to .

[0096] In this embodiment, the reason for setting the cut-off angular frequency to is to cover the harmonic range and high-frequency resonance components of the vehicle-network system under heavy scenarios. The reason for setting the filter order to is based on the compromise between hardware realizability (filter complexity) and stopband performance, so as to avoid the accumulation of phase nonlinear distortion caused by high-order filters.

[0097] Generally speaking, this embodiment provides a high-fidelity input signal for subsequent time-frequency analysis (such as the adaptive S transform described below) through hardware-level signal preprocessing optimization. For example, after filtering out high-frequency noise, the window width of the S transform kernel function can be adaptively adjusted in the low-frequency band (f ≤ 1 kHz) to more accurately capture the inter-harmonic components, while avoiding the waste of computing resources at invalid frequency points in the high-frequency band (f ≥ 10 kHz).

[0098] In an embodiment of the present invention, step S2 may specifically include:

[0099] Construct a frequency-adaptive variable window width S transform kernel function according to the following formula:

[0100]

[0101] In the formula, is the time-frequency matrix (representing the energy density of the signal at time and frequency ), is the instantaneous value of the current signal at time , is the window width adjustment function with frequency adaptability, , is the window width ratio coefficient, is the minimum frequency offset (which can be set to a small value, for example, to avoid the denominator being zero at the same time and not affecting the window width adjustment in the high-frequency band), is the Fourier basis function (realizing the frequency decomposition of the signal).

[0102] In this embodiment, it is for the time-frequency analysis link of the harmonic detection of the vehicle-grid electrical system. Through the design of the variable window width S-transform kernel function with frequency adaptability, the problem of insufficient resolution in the high-frequency band in the existing related technologies is solved.

[0103] Specifically, first, the window width adjustment function is adopted, so that the window width in the high-frequency band (for example, ) decreases as the frequency increases, and the time-domain window width , thereby improving the time-frequency energy focusing accuracy of high-frequency harmonics to optimize the time-frequency resolution of high-frequency harmonics.

[0104] Second, the time-frequency energy distribution is dynamically adjusted through the exponential Gaussian window function ( ) to suppress the spectrum leakage (such as the spread of IGBT switching noise) caused by the fixed window width in the high-frequency band in the traditional S-transform, thereby reducing the harmonic amplitude estimation error and enhancing the anti-spectrum leakage ability.

[0105] Third, the variable window width S-transform kernel function with frequency adaptability of the present invention can simultaneously capture low-frequency fundamental waves (for example, 50 / 60 Hz) and high-frequency resonance components (for example, 2 - 10 kHz PWM harmonics) to meet the harmonic detection requirements in the vehicle-grid system.

[0106] Among them, it can be understood that in the window width adjustment function , for the low-frequency band, , at this time the window width is large (approximately equal to ), and it can achieve a high frequency resolution for low-frequency harmonics (approximately equal to ). For the high-frequency band, , the window width is inversely proportional to the frequency (approximately equal to ), capable of achieving narrow time-window positioning of high-frequency harmonics (e.g., window width ≈ 1.6 at 10 kHz).

[0107] In addition, in the frequency-adaptive variable window width S-transform kernel function, the Gaussian window and the Fourier basis ( ) act jointly, making the magnitude of the time-frequency matrix represent the instantaneous energy density of the signal at time and frequency . In this way, the adaptive window width adjustment can make the energy distribution more concentrated on the real harmonic components, thereby reducing noise interference.

[0108] Generally speaking, this embodiment realizes high-precision time-frequency analysis of broadband harmonics through the frequency-adaptive variable window width S-transform kernel function, overcomes the defect of insufficient resolution in the high-frequency band of traditional methods, and provides core algorithm support for the detection of high-frequency switching noise and nonlinear harmonics in the vehicle-grid system.

[0109] In one embodiment of the present invention, in step S2, it further includes:

[0110] When the detected is detected, the frequency-adaptive window width adjustment function is adjusted to:

[0111]

[0112] where is the frequency judgment threshold.

[0113] In this way, first, the time resolution of high-frequency harmonics can be improved. Specifically, when the frequency exceeds the threshold , by adjusting the window width adjustment function from to , in this way, the attenuation rate of the window width with frequency is accelerated, the window in the high-frequency band can be made narrower, and the time resolution can be significantly improved, and the transient changes of high-frequency harmonics (e.g., fast harmonics caused by the switching of power electronic devices) can be captured more accurately, reducing time ambiguity.

[0114] Second, high-frequency noise interference is suppressed. Specifically, the noise energy is usually higher in the high-frequency region, and the narrower window can reduce the signal interception duration, reduce the integration effect of noise in time-frequency analysis, thereby improving the signal-to-noise ratio and enhancing the ability to identify real high-frequency harmonic components.

[0115] Third, adapt to the detection requirements of different frequency bands. Specifically, in the vehicle network system, low-frequency harmonics (e.g., fundamental wave, integer harmonics) are usually relatively stable and require a wider window to improve frequency resolution; while high-frequency harmonics (e.g., wide-spectrum switching noise) change rapidly and require higher time resolution. By dynamically adjusting the window width attenuation exponent, it is possible to effectively achieve differential optimization of high and low frequency bands to balance the time-frequency analysis performance.

[0116] In an embodiment of the present invention, step S3 specifically includes:

[0117] Step S3-1: Calculate the third-order cumulant:

[0118]

[0119] Step S3-2: Construct the bispectrum feature matrix:

[0120]

[0121] Step S3-3: Implement principal component projection:

[0122]

[0123] In the formula, is the third-order cumulant of the signal (used to characterize the third-order statistical characteristics of the signal), is the mathematical expectation, and are the time delay amounts, is the time when the instantaneous value of the current signal, is the current signal at the time point shifted to the right by observation values after sampling intervals, is the current signal at the time point shifted to the right by observation values after sampling intervals, is the bispectrum value (reflecting the phase coupling characteristics of the signal at frequencies and ), and are two independent frequency variables, is the projected spectrum (used to highlight the sum-frequency harmonic components generated by the nonlinear system).

[0124] Through this embodiment, first, complex harmonics caused by non-linear loads can be accurately detected. Specifically, traditional Fourier transform cannot capture the phase coupling characteristics between harmonics, while the third-order cumulant of the present invention can effectively identify the sum-frequency and difference-frequency harmonic components generated by non-linear systems (such as power electronic devices) by statistically analyzing high-order correlations.

[0125] Second, Gaussian noise interference can be suppressed. Specifically, the third-order cumulant has a natural suppression effect on Gaussian noise, and the third-order statistic of Gaussian noise in its mathematical expectation is zero. Background noise can be filtered out through bispectrum analysis, improving the harmonic detection accuracy in low signal-to-noise ratio scenarios.

[0126] Third, the discriminability of spectral features can be enhanced. Specifically, principal component projection transforms complex phase-coupled signals into single-frequency energy spectra by integrating the sum-frequency components in the bispectrum plane ( ), significantly improving the spectral resolution. For example, harmonic superposition caused by multiple parallel inverters in a vehicle-grid system can be clearly separated.

[0127] In an embodiment of the present invention, step S4 specifically includes:

[0128] Step S4-1: Establish a background noise model:

[0129]

[0130] Step S4-2: Calculate the dynamic threshold:

[0131]

[0132] In the formula, is the estimated standard deviation of the background noise at frequency , is the number of sampling points in the harmonic-free interval, is the amplitude of the time-frequency matrix at frequency and time , ∈ harmonic-free interval, is the detection threshold at frequency , is the empirical coefficient, is the set value of the false alarm rate.

[0133] Through this embodiment, first, the reliability of detection data in a noisy environment can be improved. Specifically, the background noise statistical model and the dynamic threshold in the harmonic-free interval can adapt to the noise intensity at different frequencies, avoiding misjudgment problems of fixed thresholds in high-frequency noise intervals. For example, broadband noise interference generated by power electronic switches in a vehicle-grid system can be effectively suppressed by this model.

[0134] Second, through the false alarm rate setting value and the empirical coefficient , the optimal trade-off between the false alarm probability and the detection sensitivity can be achieved in a statistical sense, ensuring the recognition accuracy of high-frequency harmonics (such as above 10 kHz).

[0135] Third, it can adapt to the time-varying noise environment. Specifically, the dynamic threshold is updated based on the noise data in the harmonic-free interval collected in real time, and can cope with the background noise fluctuations caused by load switching, locomotive overload, etc. during the operation of the vehicle-network system. For example, it can still maintain stable detection performance in the scenario where the motor noise suddenly increases when the EMU accelerates.

[0136] In this embodiment, it should be noted that extracting noise samples in the frequency bands where harmonics are known to be absent (for example, near the low-frequency fundamental wave or specific idle frequency bands) can ensure that the noise estimation is not contaminated by harmonics.

[0137] In an exemplary practical application scenario, for example, in the high-speed rail traction power supply system, the catenary voltage generates complex harmonic superposition due to the parallel operation of multiple locomotives, and at the same time, the background noise contains random pulse interference (such as pantograph-catenary off-line discharge). This embodiment can distinguish real harmonics from noise through the dynamic threshold. For example, when the time-frequency energy of a certain frequency point is [value], it can be judged as an effective harmonic.

[0138] Generally speaking, this embodiment solves the performance limitations of the traditional fixed threshold method in a complex noise environment through dynamic noise modeling and statistical decision theory. Its core innovation lies in the adaptive estimation of the background noise based on the harmonic-free interval, combined with the threshold generation mechanism constrained by the false alarm rate, significantly improving the reliability and environmental adaptability of harmonic detection in the vehicle-network electrical system.

[0139] In an embodiment of the present invention, in step S4, the empirical coefficient is set to , and the false alarm rate setting value is set to .

[0140] Through this embodiment, first, the balance between the detection sensitivity and the false alarm rate can be optimized. Specifically, by magnifying the noise standard deviation estimation value through the coefficient of , combined with the false alarm rate control of , it is ensured that weak harmonics can still be effectively detected in a low signal-to-noise ratio scenario, while limiting the misjudgment probability within 1%. For example, in the high-speed rail traction power supply system, false judgments caused by random discharge noise of the catenary (such as pantograph-catenary off-line interference) can be avoided, and at the same time, low-amplitude high-frequency harmonics generated by power electronic devices can be accurately identified.

[0141] Second, it can enhance the robustness and adaptability of the system. Specifically, the empirical coefficient Through experiments, it is verified that it can cover the deviation between the actual noise distribution and the ideal Gaussian model, and avoid the threshold misalignment caused by the change of noise statistical characteristics (such as non-Gaussian pulse interference). For example, when the load of the vehicle-grid system changes suddenly, the dynamic threshold can still stably distinguish harmonics from noise.

[0142] Third, it can achieve the dual optimization of high-frequency noise suppression and low-frequency harmonic retention. Specifically, the parameter combination and can dynamically increase the detection threshold according to the characteristic that the noise energy attenuation in the high-frequency band is slow, suppress the IGBT switching noise (for example, interference above 10 kHz), and at the same time retain the effective signals in the low-frequency band (such as 50 Hz fundamental wave and integer harmonics).

[0143] In an embodiment of the present invention, step S5 specifically includes:

[0144] Step S5-1: Use the following algorithm for parameter fitting:

[0145]

[0146] Step S5-2: Optimize the constraint conditions according to the following formula:

[0147]

[0148] In the formula, is the amplitude of the th harmonic, is the attenuation factor (used to describe the change rate of the harmonic amplitude with time), is the frequency of the th harmonic, is the initial phase, is the model order (that is, the total number of harmonic components), is the number of sampling points, is the index value of the sampling point, , is the sampling interval, represents the predicted value of the current signal reconstructed by the algorithm at the th sampling point, is the regularization coefficient, is the prior estimate value of the harmonic frequency obtained according to steps S2 to S4.

[0149] In this embodiment, first, it can improve the accuracy of harmonic parameter estimation. Specifically, through the exponential decay model with prior frequency domain constraints It can accurately describe the time-varying characteristics of harmonic amplitudes (such as the harmonic transient attenuation caused by the switching of power electronic devices), and at the same time combine the prior frequency information provided in steps S2 - S4 , which can significantly reduce the parameter estimation error. For example, in the high-speed rail traction system, the amplitude error of the fast-decaying high-frequency harmonics (such as above 10 kHz) generated by IGBT switching can be controlled within 1% by this method.

[0150] Second, it can suppress noise and overfitting interference. Specifically, the optimization objective function through the regularization term , forces the fitting frequency to be consistent with the prior estimate, avoiding the introduction of false harmonic components caused by noise interference. For example, in an exemplary implementation, in a low signal-to-noise ratio (SNR < 20 dB) scenario, it can reduce the false harmonic misjudgment by more than 50%.

[0151] Third, it can adapt to the time-varying characteristics of complex harmonics. Specifically, the introduction of the exponential decay factor can accurately characterize the harmonic amplitude attenuation characteristics in the vehicle-grid system caused by load mutations (e.g., locomotive acceleration) or the dynamic response of power electronic devices. For example, when the EMU accelerates, the harmonic amplitude of the motor decays exponentially with time, which cannot be captured by the traditional steady-state model, while the time-varying model provided by the present invention can achieve dynamic tracking.

[0152] Fourth, it can enhance the convergence and stability of the model. The introduction of the frequency-domain prior constraint can limit the solution space of the optimization problem near the known harmonic frequencies, avoid local optimal solutions, and at the same time accelerate convergence.

[0153] For the convenience of those skilled in the relevant art to understand, the following describes the process of determining the frequency-domain prior constraint through steps S2 to S4.

[0154] First, in step S2, a time-frequency matrix is generated through the variable window-width S transform , capturing the energy distribution of the signal at different times and frequencies . (For each frequency , count the proportion of time in the time-frequency matrix that exceeds the baseline noise. For example, if is significantly higher than the noise baseline in more than 80% of the time periods, it is marked as a candidate harmonic frequency. At the same time, the adaptive window width ensures that the window width in the high-frequency band is narrower, improving the time resolution of high-frequency harmonics and avoiding frequency aliasing)

[0155] In step S3, based on the third-order cumulant and principal component projection, an enhanced spectrum is generated , highlighting the phase-coupled harmonics caused by non-linear loads. (Calculating the bispectrum , identifying the sum-frequency harmonics that satisfy , enhancing the energy of the true harmonic frequencies and suppressing random noise. Select the frequencies with the top M energy rankings as high-confidence candidates)

[0156] In step S4, calculate the dynamic detection threshold using the background noise model in the harmonic-free interval , and screen for effective harmonics. (Statistical background noise standard deviation in the harmonic-free interval , reflecting the noise intensity at different frequencies. Generate a frequency-related threshold according to . If the time-frequency matrix in step S2 exceeds at multiple consecutive time points, and the value in step S3 is higher than the set threshold, then determine as the true harmonic frequency and add it to the set)

[0157] Finally, only retain the frequencies that continuously appear in multiple time windows (such as sliding window analysis) to avoid misjudgment of transient interference. Combining the time-frequency energy distribution in step S2 and the phase-coupling characteristics in step S3, eliminate isolated frequency points (such as frequencies that only appear at a single time point). Input the final screened list into the exponential decay model in step S5 as the frequency-domain prior constraint.

[0158] In an exemplary implementation, in a high-speed rail traction system, IGBT switches generate high-frequency harmonics (e.g., 2 kHz), while catenary discharges cause random pulse noise.

[0159] In step S2, the time-frequency matrix shows continuous energy at 2 kHz, but there are instantaneous spikes in the frequency band above 10 kHz.

[0160] In step S3, bispectrum analysis reveals the presence of at 2 kHz and phase coupling, enhancing its spectral energy.

[0161] In step S4, the dynamic threshold determines that the 2 kHz frequency exceeds 90% of the time, while the 10 kHz spike only exceeds the threshold at individual time points, so the latter is excluded.

[0162] The result is: is input into step S5 to guide parameter fitting and accurately estimate the harmonic amplitude attenuation characteristics.

[0163] According to the second aspect of the present invention, there is also provided a harmonic detection system for a vehicle-grid electrical system, which is used to execute the harmonic detection method for the vehicle-grid electrical system according to any one of the technical solutions in the first aspect of the present invention. The harmonic detection system for the vehicle-grid electrical system includes:

[0164] A data acquisition and processing module, which is used to acquire the current signal of the vehicle-grid busbar and perform band-limiting processing on the current signal;

[0165] An adaptive time-frequency analysis module, which is used to construct a frequency-adaptive variable window-width S-transform kernel function;

[0166] A bispectrum feature extraction module, which is used to extract harmonic phase coupling features based on the bispectrum analysis of the third-order cumulant and generate an enhanced spectrum through principal component projection;

[0167] A dynamic threshold generation module, which is used to generate a dynamic detection threshold by using a background noise model in a harmonic-free interval;

[0168] A parameter optimization module, which is used to perform harmonic parameter fitting by using an exponential decay model with prior frequency-domain constraints.

[0169] In this embodiment, the harmonic detection system for the vehicle-grid electrical system of the present invention seamlessly integrates the signal acquisition, time-frequency analysis, feature extraction, dynamic threshold determination, and parameter optimization processes through the series cooperation of five functional modules, significantly improving the harmonic detection accuracy. It can also achieve the full-process closed-loop optimization from signal preprocessing to parameter optimization, and is particularly suitable for the precise detection of high-frequency, nonlinear, and time-varying harmonics in the vehicle-grid system, and can provide a more reliable diagnostic basis for power quality governance.

[0170] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A harmonic detection method for a vehicle network electrical system, characterized in that, It includes the following steps: Step S1: Obtain the current signal of the vehicle-network busbar and perform band-limiting processing on the current signal; Step S2: Construct a variable window-width S-transform kernel function with frequency adaptability; Step S3: Based on the bispectrum analysis of the third-order cumulant, extract the harmonic phase coupling characteristics and generate an enhanced spectrum through principal component projection; Step S4: Generate a dynamic detection threshold using the background noise model in the non-harmonic interval; Step S5: Perform harmonic parameter fitting using an exponential decay model with prior frequency-domain constraints.

2. The harmonic detection method of the vehicle network electrical system according to claim 1, wherein The specific content of step S1 includes: Step S1-1: Collect the current signal of the vehicle-network busbar using a broadband current sensor; Step S1-2: Perform band-limiting processing through an anti-aliasing filter: In the formula, is the complex frequency domain transfer function of the filter, is the complex frequency variable, is the cut-off angular frequency, is the filter order.

3. The harmonic detection method for the vehicle network electrical system according to claim 2, wherein The cut-off angular frequency is set to ; the filter order is set to .

4. The harmonic detection method of the vehicle network electrical system according to claim 2, characterized in that, The specific content of step S2 includes: Construct a variable window-width S-transform kernel function with frequency adaptability according to the following formula: In the formula, is the time-frequency matrix, is the instantaneous value of the current signal at time is the window width adjustment function with frequency adaptability, , is the window width ratio coefficient, is the minimum frequency offset, is the frequency, is the Fourier basis function.

5. The harmonic detection method of the vehicle network electrical system according to claim 4, wherein, In step S2, it also includes: When the detected occurs, the frequency adaptive window width adjustment function is adjusted to: Among them, is the frequency judgment threshold value.

6. The harmonic detection method for the vehicle network electrical system according to claim 1, characterized in that, The specific content of step S3 includes: Step S3-1: Calculate the third-order cumulant; Step S3-2: Construct a bispectrum feature matrix; Step S3-3: Implement principal component projection; In the formula, is the third-order cumulant of the signal, is the mathematical expectation, and are the time delay amounts, is the instantaneous value of the current signal at time , is the observed value of the current signal after shifting sampling intervals to the right based on the time point , is the observed value of the current signal after shifting sampling intervals to the right based on the time point , is the bispectrum value, and are two independent frequency variables, is the projected spectrum.

7. The harmonic detection method for the vehicle network electrical system according to claim 1, characterized in that The specific content of step S4 includes: Step S4-1: Establish a background noise model; Step S4-2: Calculate the dynamic threshold; Wherein, is the estimated standard deviation of the background noise at the frequency , is the number of sampling points in the no-harmonic interval, is the amplitude of the time-frequency matrix at the frequency and time , ∈ no-harmonic interval, is the detection threshold at the frequency , is the empirical coefficient, is the set value of the false alarm rate.

8. The harmonic detection method for the vehicle-network electrical system according to claim 7, characterized in that, In the step S4, the empirical coefficient is set to , and the set value of the false alarm rate is set to .

9. The harmonic detection method for the vehicle network electrical system according to claim 1, characterized in that The specific content of step S5 includes: Step S5-1: Perform parameter fitting using the following algorithm: Step S5-2: Optimize the constraint conditions according to the following formula: Wherein, is the amplitude of the -th harmonic, is the attenuation factor, is the frequency of the -th harmonic, is the initial phase, is the model order, is the number of sampling points, is the index value of the sampling point, , is the sampling interval, represents the predicted value of the current signal reconstructed by the algorithm at the -th sampling point, is the regularization coefficient, is the prior estimated value of the harmonic frequency obtained according to steps S2 to S4.

10. A harmonic detection system for a vehicle network electrical system, characterized in that, For implementing the harmonic detection method of the vehicle-network electrical system as described in any one of claims 1-9, the harmonic detection system of the vehicle-network electrical system includes: A data acquisition and processing module, which is used to obtain the current signal of the vehicle-network busbar and perform band-limiting processing on the current signal; An adaptive time-frequency analysis module, which is used to construct a variable window-width S-transform kernel function with frequency adaptability; A bispectrum feature extraction module, which is used to extract the harmonic phase coupling characteristics based on the bispectrum analysis of the third-order cumulant and generate an enhanced spectrum through principal component projection; A dynamic threshold generation module, which is used to generate a dynamic detection threshold using the background noise model in the non-harmonic interval; A parameter optimization module, which is used to perform harmonic parameter fitting using an exponential decay model with prior frequency-domain constraints.

Images