Real-time dynamic compensation method and system for frequency converter harmonic suppression

Abstract

The application discloses a real-time dynamic compensation method and system for frequency converter harmonic suppression, and relates to the technical field of frequency converter harmonic suppression.The method comprises the following steps: constructing a reference signal template; receiving a harmonic-containing output signal of the frequency converter to obtain a compensation execution signal used for executing a compensation operation; analyzing the compensation execution signal by using a deconvolution harmonic compensation prediction model to obtain a predicted output signal of the frequency converter; obtaining a deconvolution compensation signal; and performing deconvolution stripping on the reference signal template by using the deconvolution compensation signal to output a harmonic suppression signal of the frequency converter.The application solves the technical problem that the type of a main signal is often out of synchronization with the type of a harmonic and there is a difference between the types in the prior art, which leads to estimation synthesis error, and further causes the decline of the harmonic suppression effect and stability of the frequency converter, so that the real-time dynamic suppression of the frequency converter harmonic is realized, the harmonic suppression precision is improved, and the estimation deviation caused by type mismatch is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of frequency converter harmonic suppression, and particularly relates to a real-time dynamic compensation method and system for frequency converter harmonic suppression. BACKGROUND

[0002] During the operation of a frequency converter, harmonic interference is a common problem. Traditional harmonic compensation methods mostly directly synthesize and inject compensation sinusoidal components, or use narrowband filters for elimination. However, when the main signal type and the harmonic type are out of synchronization, and estimation or synthesis errors are caused by type differences, the harmonic suppression effect of these traditional methods will obviously decrease, and the stability is also difficult to guarantee, which cannot meet the demand for high-precision compensation.

[0003] In the prior art, the main signal type and the harmonic type are often out of synchronization and have differences, which causes estimation and synthesis errors, and further decreases the harmonic suppression effect and stability of the frequency converter. SUMMARY

[0004] The present application provides a real-time dynamic compensation method and system for frequency converter harmonic suppression, which is used to solve the technical problem that the main signal type and the harmonic type are often out of synchronization and have differences in the prior art, which causes estimation and synthesis errors, and further decreases the harmonic suppression effect and stability of the frequency converter.

[0005] In view of the above problems, the present application provides a real-time dynamic compensation method and system for frequency converter harmonic suppression.

[0006] In a first aspect of the present application, a real-time dynamic compensation method for frequency converter harmonic suppression is provided, and the method comprises:

[0007] A reference signal template is constructed, the reference signal template is a reference signal type output by a frequency converter; a harmonic-containing output signal of the frequency converter is received, the harmonic-containing output signal is compensated based on the reference signal template, and a compensation execution signal used for performing a compensation operation is obtained; the compensation execution signal is analyzed by a deconvolution harmonic compensation prediction model, and a predicted output signal of the frequency converter is obtained; a deconvolution compensation signal is obtained according to the signal difference between the predicted output signal and the reference signal template, the reference signal template is deconvoluted and stripped by using the deconvolution compensation signal, and a harmonic suppression signal of the frequency converter is output, wherein the signal types of the deconvolution compensation signal, the predicted output signal and the reference signal template are the same.

[0008] In a possible implementation, historical non-harmonic interference signal samples of the frequency converter are collected, signal types of the historical non-harmonic interference signal samples are determined, signal normalization processing is performed on the historical non-harmonic interference signal samples to obtain normalized amplitudes and normalized phases, and a reference signal template is constructed according to the normalized amplitudes, the normalized phases, and the signal types, where the signal type of the reference signal template is a pulse signal.

[0009] In a possible implementation, a first reference signal template and a second reference signal template are constructed, the harmonic-containing output signal is compensated once according to the first reference signal template and the second reference signal template respectively to output a first compensation execution signal and a second compensation execution signal, and a first harmonic suppression signal and a second harmonic suppression signal are output according to the first compensation execution signal and the second compensation execution signal, and mean fitting is performed on the first harmonic suppression signal and the second harmonic suppression signal to obtain the harmonic suppression signal of the frequency converter.

[0010] In a possible implementation, the first compensation execution signal and the second compensation execution signal are analyzed by using a deconvolution harmonic compensation prediction model to obtain a first prediction output signal and a second prediction output signal of the frequency converter, a first deconvolution compensation signal is obtained according to signal difference between the first prediction output signal and the first reference signal template, a second deconvolution compensation signal is obtained according to signal difference between the second prediction output signal and the second reference signal template, the first deconvolution compensation signal is used to perform deconvolution stripping on the first reference signal template to output the first harmonic suppression signal of the frequency converter, and the second deconvolution compensation signal is used to perform deconvolution stripping on the second reference signal template to output the second harmonic suppression signal of the frequency converter.

[0011] In a possible implementation, the historical non-harmonic interference signal samples are classified into two categories according to a parameter range of the signal to obtain a first category of historical non-harmonic interference signal samples and a second category of historical non-harmonic interference signal samples, and the first category of historical non-harmonic interference signal samples and the second category of historical non-harmonic interference signal samples are normalized respectively to output a first reference signal template and a second reference signal template.

[0012] In a possible implementation, the compensation execution signal is subjected to band-pass filtering processing to output a preprocessed compensation execution signal, input data of the preprocessed compensation execution signal is extracted, including a time domain sequence, FFT frequency spectrum data, and phase data, the deconvolution harmonic compensation prediction model uses a Tikhonov regularization deconvolution algorithm to take the input data as a convolution result, performs adaptive convolution kernel updating on a signal type of the reference signal template as a deconvolution prior output template to obtain a prediction output signal of the frequency converter.

[0013] In a possible implementation, a frequency domain response parameter of the deconvolution kernel is initialized, and a prior signal frequency domain of the reference signal template is initialized; the prior signal frequency domain of the reference signal template is updated by using a Tikhonov regularization deconvolution formula under the condition of the frequency domain response parameter, to obtain a predicted signal frequency domain; the predicted signal frequency domain is subjected to inverse Fourier transform to obtain a predicted output signal in a time domain; the frequency domain response parameter of the deconvolution kernel is iteratively updated according to the predicted output signal in the time domain, to obtain a converged deconvolution harmonic compensation prediction model, and a predicted output signal of the frequency converter is output.

[0014] In a possible implementation, the frequency domain prior parameter of the reference signal template is updated by using a Tikhonov regularization deconvolution formula; wherein the updating includes adding a regularization parameter to a denominator of the Tikhonov regularization deconvolution formula to suppress noise amplification, and introducing the frequency domain prior of the reference signal template into a numerator of the Tikhonov regularization deconvolution formula to constrain an output signal form.

[0015] In a possible implementation, an amplitude matching coefficient is adjusted according to an amplitude difference between the predicted output signal in the time domain and the reference signal template, to update the prior signal frequency domain; the frequency domain response parameter of the deconvolution kernel is iteratively updated based on the updated prior signal frequency domain, until a prediction error of the deconvolution harmonic compensation prediction model is less than a preset error threshold.

[0016] In a second aspect, the present application provides a real-time dynamic compensation system for harmonic suppression of a frequency converter, the system comprising:

[0017] a signal template construction module configured to construct a reference signal template, the reference signal template being of a reference signal type output by the frequency converter; a compensation execution signal acquisition module configured to receive a harmonic-containing output signal of the frequency converter, and to compensate the harmonic-containing output signal based on the reference signal template to obtain a compensation execution signal used for performing a compensation operation; a predicted output signal acquisition module configured to analyze the compensation execution signal by using a deconvolution harmonic compensation prediction model, and to acquire a predicted output signal of the frequency converter; and a harmonic suppression signal output module configured to acquire a deconvolution compensation signal according to a signal difference between the predicted output signal and the reference signal template, to perform deconvolution stripping on the reference signal template by using the deconvolution compensation signal, and to output a harmonic suppression signal of the frequency converter, wherein the deconvolution compensation signal, the predicted output signal, and the reference signal template are of the same signal type.

[0018] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0019] A reference signal template is constructed; the harmonic-containing output signal of the frequency converter is received to obtain a compensation execution signal for performing compensation operations; the compensation execution signal is analyzed to obtain the predicted output signal of the frequency converter; according to the signal difference between the predicted output signal and the reference signal template, a deconvolution compensation signal is obtained; the reference signal template is deconvolved and stripped using the deconvolution compensation signal to output the harmonic suppression signal of the frequency converter. This achieves the technical effect of realizing real-time dynamic suppression of frequency converter harmonics, improving harmonic suppression accuracy, and reducing estimation bias caused by type mismatch. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A schematic flowchart of the real-time dynamic compensation method for inverter harmonic suppression provided in the embodiments of this application;

[0022] Figure 2 This is a schematic diagram of the real-time dynamic compensation system for inverter harmonic suppression provided in an embodiment of this application.

[0023] Explanation of reference numerals in the attached figures: Signal template construction module 10, compensation execution signal acquisition module 20, prediction output signal acquisition module 30, harmonic suppression signal output module 40. Detailed Implementation

[0024] This application provides a real-time dynamic compensation method and system for inverter harmonic suppression, which addresses the technical problem in the prior art where the main signal type and harmonic type are out of sync and have different types, leading to estimation and synthesis errors, and thus reducing the harmonic suppression effect and stability of the inverter.

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0026] Example 1, as Figure 1 As shown, this application provides a real-time dynamic compensation method for harmonic suppression in frequency converters, the method comprising:

[0027] Step S100: constructing a reference signal template, the reference signal template being a reference signal type output by the frequency converter.

[0028] Specifically, when constructing the reference signal template, first, historical harmonic-free signal samples of the frequency converter are collected, and the signal type (such as pulse signal) of the samples is determined; then, the samples are subjected to signal normalization processing to obtain normalized amplitudes and normalized phases, and the reference signal template is constructed according to the normalized amplitudes, the normalized phases, and the determined signal type. In addition, the historical harmonic-free signal samples are classified into two categories according to the parameter ranges of the signals, to obtain first and second historical harmonic-free signal samples, and the two categories of samples are subjected to normalization processing respectively, thereby generating a first reference signal template and a second reference signal template, to provide a basis for subsequent compensation operations based on different templates.

[0029] Step S200: receiving a harmonic-containing output signal of the frequency converter, and compensating the harmonic-containing output signal based on the reference signal template to obtain a compensation execution signal used for performing compensation operations.

[0030] Specifically, after receiving the harmonic-containing output signal of the frequency converter, the harmonic-containing output signal is subjected to compensation operations based on the constructed reference signal template (including the first reference signal template and the second reference signal template): the first reference signal template is used to perform a first compensation processing on the harmonic-containing output signal, to output a corresponding first compensation execution signal; at the same time, the second reference signal template is used as a reference to perform the same compensation logic on the same harmonic-containing output signal, to obtain a second compensation execution signal, and the two compensation execution signals will be used as input data of a deconvolution harmonic compensation prediction model, to provide a basis for accurate analysis and obtaining of a predicted output signal.

[0031] Step S300: analyzing the compensation execution signal by using the deconvolution harmonic compensation prediction model, to obtain a predicted output signal of the frequency converter.

[0032] Specifically, when analyzing the compensation execution signal by the deconvolution harmonic compensation prediction model, the compensation execution signal is first band-pass filtered to remove noise interference, and the time domain sequence, FFT spectrum data and phase data of the preprocessed signal are extracted as model inputs; the model takes the signal type (such as a pulse signal) of the reference signal template as the deconvolution prior output template, initializes the frequency domain response parameters of the deconvolution kernel and the prior signal frequency domain of the reference signal template, and processes using the Tikhonov regularization deconvolution formula; a regularization parameter is added to the denominator of the formula to suppress noise amplification, and the frequency domain prior of the reference signal template is introduced into the numerator to constrain the output signal form; the predicted signal frequency domain is obtained by updating the prior signal frequency domain; the predicted signal frequency domain is converted into a time domain predicted output signal by inverse Fourier transform, and the amplitude matching coefficient is adjusted according to the amplitude difference between the time domain predicted output signal and the reference signal template; the prior signal frequency domain is updated and the frequency domain response parameters of the deconvolution kernel are iteratively updated until the model prediction error is less than a preset threshold, and a converged deconvolution harmonic compensation prediction model is obtained, and finally the predicted output signal of the frequency converter (such as the first and second predicted output signals corresponding to the first and second compensation execution signals) is output.

[0033] Step S400: According to the signal difference between the predicted output signal and the reference signal template, a deconvolution compensation signal is obtained, the reference signal template is deconvolved using the deconvolution compensation signal, and a harmonic suppression signal of the frequency converter is output, wherein the deconvolution compensation signal, the predicted output signal and the reference signal template are of the same signal type.

[0034] Specifically, according to the signal difference between the predicted output signal (including the first predicted output signal and the second predicted output signal) and the corresponding reference signal template (the first reference signal template and the second reference signal template), the first deconvolution compensation signal and the second deconvolution compensation signal are respectively calculated and obtained; since the deconvolution compensation signal, the predicted output signal and the reference signal template are of the same signal type (all are pulse signals), the first deconvolution compensation signal can be directly used to perform deconvolution stripping operation on the first reference signal template to output the first harmonic suppression signal, and the second deconvolution compensation signal can be directly used to perform the same deconvolution stripping on the second reference signal template to obtain the second harmonic suppression signal; finally, the first harmonic suppression signal and the second harmonic suppression signal are subjected to mean fitting, thereby outputting the final harmonic suppression signal of the frequency converter and realizing accurate suppression of harmonics.

[0035] In one possible implementation manner, step S100 further includes:

[0036] Step S110: Collecting a historical non-harmonic interference signal sample of the frequency converter, and determining the signal type of the historical non-harmonic interference signal sample.

[0037] Step S120: signal normalization processing is performed on the historical non-harmonic interference signal sample to obtain a normalized amplitude and a normalized phase, and a reference signal template is constructed according to the normalized amplitude, the normalized phase and the signal type, wherein the signal type of the reference signal template is a pulse signal.

[0038] Specifically, historical signal samples generated by the frequency converter under normal operation and without harmonic interference are collected, which cover the output signals of the frequency converter under different working loads and operating parameter conditions; the signal type is determined by analyzing the waveform structure, time domain characteristics and frequency components of the sample signals, and the signal type of the historical non-harmonic interference signal sample is a pulse signal, which provides a signal type reference for subsequent construction of the reference signal template.

[0039] When the historical non-harmonic interference signal sample (the signal type is a pulse signal) is subjected to signal normalization processing, the peak amplitude of the pulse signal in the sample is scaled to a preset standard amplitude range (such as 0-1) by an amplitude normalization algorithm, so as to eliminate the influence of amplitude difference under different conditions and obtain a normalized amplitude; at the same time, a phase alignment algorithm is used to take the rising edge or falling edge of the pulse signal as a reference point, and the initial phase of each sample signal is uniformly calibrated to the same reference phase to calculate a normalized phase. Subsequently, a signal template generation tool is called to associate and integrate the normalized amplitude, the normalized phase and the inherent characteristics (such as pulse width and pulse interval parameters) of the pulse signal, the above characteristic parameters are fixed through a template parameter configuration interface, and a reference signal template of the pulse signal type is constructed, which can output standardized pulse signal waveform data through a data interface to provide a reference for subsequent compensation operations.

[0040] In a possible implementation manner, step S100 further includes:

[0041] Step S130: constructing a first reference signal template and a second reference signal template.

[0042] Step S140: performing one compensation on the harmonic-containing output signal according to the first reference signal template and the second reference signal template respectively, and outputting a first compensation execution signal and a second compensation execution signal.

[0043] Step S150: outputting a first harmonic suppression signal and a second harmonic suppression signal according to the first compensation execution signal and the second compensation execution signal, and performing mean value fitting on the first harmonic suppression signal and the second harmonic suppression signal to obtain a harmonic suppression signal of the frequency converter.

[0044] Specifically, when constructing the first reference signal template and the second reference signal template, first, the collected historical non-harmonic interference signal samples are classified according to the parameter range of the signal (such as the amplitude interval of the pulse signal, the phase variation range, etc.), and the first type of historical non-harmonic interference signal samples with parameters meeting the first preset range and the second type of historical non-harmonic interference signal samples with parameters meeting the second preset range are screened out; then, the two types of samples are respectively subjected to signal normalization processing to unify the amplitude and phase scales thereof, and the first reference signal template and the second reference signal template corresponding to the types of the pulse signal (such as the pulse period, the duty cycle, etc.) are respectively constructed, thereby providing differentiated references for subsequent compensation operations based on different parameter ranges.

[0045] After receiving the harmonic-containing output signal of the frequency converter, the compensation operation is performed based on the first reference signal template: the harmonic-containing output signal is compared with the first reference signal template in the time domain to extract the difference values of the pulse amplitude, the phase offset and the pulse width, and a first compensation coefficient is outputted; the harmonic component is corrected by weighting operation of the first compensation coefficient and the harmonic-containing output signal to obtain a first compensation execution signal; at the same time, the same implementation logic is adopted to take the second reference signal template as a reference, difference values are obtained through waveform comparison, a second compensation coefficient is generated, and the harmonic-containing output signal is synchronously weighted and corrected to output a second compensation execution signal. The whole process is completed through a double-channel parallel processing mechanism to ensure the synchronization of the two compensation execution signals and provide standardized input for subsequent analysis of the deconvolution harmonic compensation prediction model.

[0046] The first compensation execution signal and the second compensation execution signal are respectively analyzed through the deconvolution harmonic compensation prediction model, the signal is first subjected to band-pass filter preprocessing to extract the time domain sequence, the FFT spectrum data and the phase data, the Tikhonov regularization deconvolution algorithm is used in combination with the prior information of the corresponding reference signal template to obtain the first prediction output signal and the second prediction output signal; then, according to the signal difference between the two types of prediction output signals and the respective reference signal templates, the first deconvolution compensation signal and the second deconvolution compensation signal are obtained, the first reference signal template and the second reference signal template are respectively subjected to deconvolution stripping to output the first harmonic suppression signal and the second harmonic suppression signal; finally, the mean fitting algorithm is adopted to fuse and calculate the time domain characteristics or amplitude parameters of the two harmonic suppression signals, the amplitude mean value of the two signals at the same time node is obtained to obtain the final harmonic suppression signal of the frequency converter, thereby improving the stability and precision of harmonic suppression.

[0047] In one possible implementation manner, the step S150 further includes:

[0048] Step S151: analyzing the first compensation execution signal and the second compensation execution signal through the deconvolution harmonic compensation prediction model to obtain the first prediction output signal and the second prediction output signal of the frequency converter.

[0049] Step S152: obtaining a first deconvolution compensation signal according to the signal difference between the first prediction output signal and the first reference signal template.

[0050] Step S153: obtaining a second deconvolution compensation signal according to the signal difference between the second prediction output signal and the second reference signal template.

[0051] Step S154: performing deconvolution stripping on the first reference signal template using the first deconvolution compensation signal to output the first harmonic suppression signal of the frequency converter, and performing deconvolution stripping on the second reference signal template using the second deconvolution compensation signal to output the second harmonic suppression signal of the frequency converter.

[0052] Specifically, when analyzing the first compensation execution signal and the second compensation execution signal by the deconvolution harmonic compensation prediction model, first, the two signals are respectively subjected to band-pass filtering processing to complete preprocessing, and the time domain sequence, FFT spectrum data and phase data of the preprocessed signals are extracted as model inputs; the model takes the pulse signal type of the corresponding reference signal template (the first reference signal template corresponds to the first compensation execution signal, and the second reference signal template corresponds to the second compensation execution signal) as the deconvolution prior output template, initializes the frequency domain response parameters of the deconvolution kernel and the prior signal frequency domain of the reference signal template, and under the condition of the frequency domain response parameters, the prior signal frequency domain of the reference signal template is updated by Tikhonov regularization deconvolution formula (the regularization parameter is added to the denominator to suppress noise amplification, and the frequency domain prior of the reference signal template is introduced into the numerator to constrain the output signal form), to obtain the prediction signal frequency domain; the prediction signal frequency domain is subjected to inverse Fourier transform to obtain the prediction output signal in the time domain, and then the frequency domain response parameters of the deconvolution kernel are iteratively updated according to the prediction output signal in the time domain (the amplitude matching coefficient is adjusted by the amplitude difference between the prediction output signal in the time domain and the reference signal template, the prior signal frequency domain is updated, and the frequency domain response parameters of the deconvolution kernel are iteratively updated based on the updated prior signal frequency domain until the model prediction error is less than the preset error threshold), to obtain the converged deconvolution harmonic compensation prediction model, and finally output the first prediction output signal and the second prediction output signal of the frequency converter.

[0053] When the first deconvolution compensation signal is obtained according to the signal difference between the first predicted output signal and the first reference signal template, the two are first compared in time domain and frequency domain, and the signal difference in amplitude fluctuation, phase shift, pulse width deviation and spectral component difference of the pulse signal is specifically analyzed. The amplitude difference at the same time node and the time offset of the rising edge / falling edge of the pulse are calculated in the time domain, and the energy difference of each harmonic component in the FFT spectrum is compared in the frequency domain. Then, based on these difference parameters, the first deconvolution compensation signal of the same type as the first reference signal template is generated through a deviation quantization model combined with the pulse signal type characteristics of the first reference signal template. The signal can accurately map the deviation degree of the first predicted output signal relative to the first reference signal template, and provide a targeted compensation basis for subsequent deconvolution stripping operations.

[0054] When the second deconvolution compensation signal is obtained according to the signal difference between the second predicted output signal and the second reference signal template, the two are first compared in time domain and frequency domain, and the amplitude deviation at the same time node, the rising edge and falling edge time offset of the pulse signal are calculated in the time domain, and the energy difference of each harmonic component is compared through spectral analysis in the frequency domain, and the specific difference between the two in pulse characteristics (such as pulse width, interval) and signal parameters is determined. Then, based on these signal differences, the second deconvolution compensation signal of the same type as the second reference signal template is generated through deviation quantization and signal conversion with the pulse signal type of the second reference signal template as the reference. The signal can accurately reflect the deviation of the second predicted output signal relative to the second reference signal template, and provide an adaptive compensation basis for subsequent deconvolution stripping operations.

[0055] When the first deconvolution compensation signal is used to perform deconvolution stripping on the first reference signal template, the deconvolution operation module is called, the pulse signal type of the first reference signal template is used as the deconvolution prior template, the first deconvolution compensation signal is input as the convolution interference term, the Tikhonov regularization deconvolution algorithm (a regularization parameter is introduced in the frequency domain to suppress noise amplification, combined with the frequency domain prior constraint of the first reference signal template) is used to perform deconvolution operation on the first reference signal template and the first deconvolution compensation signal, to strip the deviation component caused by harmonics in the first reference signal template, and output the first harmonic suppression signal in the time domain. At the same time, the same implementation method is used to input the second deconvolution compensation signal into the deconvolution operation module, the pulse signal type of the second reference signal template is used as the deconvolution prior template, the Tikhonov regularization deconvolution algorithm is used to perform deconvolution stripping operation to remove the harmonic interference component, and the second harmonic suppression signal is obtained, so that the two output signals both maintain the pulse signal type consistent with the corresponding reference signal template.

[0056] In one possible implementation, step S130 further includes:

[0057] Step S131: The historical non-harmonic interference signal samples are classified according to the parameter range of the signal, to obtain first-class historical non-harmonic interference signal samples and second-class historical non-harmonic interference signal samples.

[0058] Step S132: The first-class historical non-harmonic interference signal samples and the second-class historical non-harmonic interference signal samples are normalized respectively, to output a first reference signal template and a second reference signal template.

[0059] Specifically, the collected historical non-harmonic interference signal samples are classified according to the parameter range of the signal (such as the upper and lower limits of the amplitude of the pulse signal, the phase variation range, the pulse period, etc.), and by setting clear parameter thresholds, the samples that meet the first parameter interval are classified as first-class historical non-harmonic interference signal samples, and the samples that meet the second parameter interval are classified as second-class historical non-harmonic interference signal samples, thereby completing the classification and providing a classified sample basis for subsequent construction of different reference signal templates.

[0060] The first-class historical non-harmonic interference signal samples are normalized by calculating the amplitude range and phase distribution of the samples, adjusting the amplitude to a unified standard interval, and calibrating the phase to obtain normalized amplitude and normalized phase; the second-class historical non-harmonic interference signal samples are subjected to the same normalization operation. Then, combined with the pulse signal types of the two types of samples, the normalized amplitude and normalized phase of the first-class samples are integrated with the pulse characteristics to form a first reference signal template and stored as a signal file in a specific format; similarly, the normalized parameters of the second-class samples are integrated with the pulse characteristics to form a second reference signal template and stored in the same format, and the output is completed.

[0061] In one possible implementation manner, step S300 further includes:

[0062] Step S310: The compensation execution signal is subjected to band-pass filtering processing, to output a preprocessed compensation execution signal.

[0063] Step S320: The input data of the preprocessed compensation execution signal is extracted, including a time domain sequence, FFT spectrum data, and phase data.

[0064] Step S330: The deconvolution harmonic compensation prediction model uses a Tikhonov regularization deconvolution algorithm to take the input data as a convolution result, performs adaptive convolution kernel updating on the reference signal template with the signal type of the reference signal template as a deconvolution prior output template, to obtain a predicted output signal of the frequency converter.

[0065] Specifically, when the compensation execution signal is subjected to band-pass filtering, the upper and lower cut-off frequencies of the band-pass filter are set according to the effective frequency range of the frequency converter output signal, the signal components within the frequency range are allowed to pass through the filter, while the noise and interference signals below the lower cut-off frequency and above the upper cut-off frequency are blocked, so that the compensation execution signal is purified, and finally the preprocessed compensation execution signal without irrelevant frequency band interference is output.

[0066] For the preprocessed compensation execution signal, the amplitude information thereof at different time points is continuously collected in a timing sampling manner to form a time domain sequence reflecting the change law of the signal over time; the fast Fourier transform (FFT) is used to convert the time domain sequence into frequency domain data to obtain FFT spectrum data containing the amplitude and energy distribution of each frequency component; at the same time, the phase value of the signal on each frequency component and the phase change information in the time domain are extracted through a phase detection technology to form phase data, and finally the three types of data are integrated as inputs of the deconvolution harmonic compensation prediction model.

[0067] When the deconvolution harmonic compensation prediction model is running, the extracted time domain sequence, FFT spectrum data and phase data are taken as the convolution result, and a Tikhonov regularization deconvolution algorithm is introduced for processing. The algorithm takes the pulse signal type of the reference signal template as the deconvolution prior output template, and realizes adaptive update by continuously adjusting the parameters of the convolution kernel. The frequency domain response of the convolution kernel is constrained based on the signal characteristics of the prior template in the frequency domain, so that the convolution operation result gradually approximates the input data, while the noise interference to the update process is suppressed. After multiple rounds of iterative update of the convolution kernel, the matching error between the signal output by the model and the input data meets the preset requirement, and finally the prediction output signal reflecting the output characteristics of the frequency converter is obtained.

[0068] In one possible implementation manner, step S330 further includes:

[0069] Step S331: initializing the frequency domain response parameters of the deconvolution kernel and the prior signal frequency domain of the reference signal template.

[0070] Step S332: under the condition of the frequency domain response parameters, the prior signal frequency domain of the reference signal template is subjected to regularized update by using a Tikhonov regularization deconvolution formula to obtain a prediction signal frequency domain.

[0071] Step S333: the prediction signal frequency domain is subjected to inverse Fourier transform to obtain a time domain prediction output signal.

[0072] Step S334: the frequency domain response parameters of the deconvolution kernel are iteratively updated according to the time domain prediction output signal to obtain a converged deconvolution harmonic compensation prediction model, and the prediction output signal of the frequency converter is output.

[0073] Specifically, when initializing the frequency domain response parameters of the deconvolution kernel, a set of initial frequency domain response values (such as setting the initial amplitude as 1 and the phase as 0 of the unit response parameters) are set as the initial state of the deconvolution kernel according to the operating frequency range of the frequency converter and the requirements of the deconvolution algorithm; for the prior signal frequency domain of the reference signal template, the reference signal template is converted from the time domain to the frequency domain by performing Fourier transform, and the corresponding frequency domain amplitude and phase information are extracted as the initial values of the prior signal frequency domain.

[0074] Under the limitation of the initialized frequency domain response parameters of the deconvolution kernel, the Tikhonov regularization deconvolution formula is called, and the prior signal frequency domain of the reference signal template is substituted into the formula. The preset regularization parameter is added to the denominator of the formula to reduce the influence of noise amplification, and the frequency domain prior information of the reference signal template is introduced into the numerator to regulate the form of the output signal. Through such operation, the prior signal frequency domain is updated by regularization, and the predicted signal frequency domain is finally calculated.

[0075] For the obtained predicted signal frequency domain, the inverse Fourier transform algorithm is used for processing, and the signal amplitude and phase information represented in the frequency domain are converted back to the time domain. Through this conversion process, the frequency spectrum distribution in the frequency domain is restored to the signal waveform varying with time, and the predicted output signal in the time domain is finally obtained. This signal can intuitively reflect the change characteristics of the frequency converter output signal in the time dimension.

[0076] The amplitude difference between the predicted output signal in the time domain and the reference signal template is calculated, the amplitude matching coefficient is adjusted according to the difference, and then the prior signal frequency domain is updated. Subsequently, based on the updated prior signal frequency domain, the frequency domain response parameters of the deconvolution kernel are adjusted, and the parameter updating and model prediction process are repeatedly executed. Through continuous iteration and optimization, the prediction error of the deconvolution harmonic compensation prediction model is reduced to below the preset error threshold, and the model reaches the convergence state at this time. Finally, the predicted output signal of the frequency converter is output.

[0077] In one possible implementation manner, step S332 further includes:

[0078] The frequency domain prior parameters of the reference signal template are updated by using the Tikhonov regularization deconvolution formula.

[0079] The regularization update includes adding a regularization parameter to the denominator of the Tikhonov regularization deconvolution formula to suppress noise amplification, and introducing the frequency domain prior of the reference signal template into the numerator of the Tikhonov regularization deconvolution formula to constrain the form of the output signal.

[0080] Specifically, when the Tikhonov regularization deconvolution formula is used to update the frequency domain priori parameters of the reference signal template, the frequency domain priori parameters of the reference signal template are input into the formula, and a preset regularization parameter is added to the denominator part of the formula to suppress the amplification effect of noise in the processing process. Meanwhile, the frequency domain priori information of the reference signal template is introduced into the numerator of the formula to constrain the form of the output signal so as to make it conform to the characteristics of the reference signal template, thereby completing the regularization update of the frequency domain priori parameters.

[0081] In the application process of the Tikhonov regularization deconvolution formula, the regularization update is realized through two key operations: one is to add a regularization parameter to the denominator part of the formula, which can effectively suppress the amplification of noise in the deconvolution calculation and reduce the interference of noise on the signal processing result; and the other is to introduce the frequency domain priori information of the reference signal template into the numerator of the formula to constrain the form of the output signal and ensure that it conforms to the characteristics of the reference signal template, thereby completing the regularization update of the related parameters.

[0082] In one possible implementation manner, the step S334 further includes:

[0083] The amplitude matching coefficient is adjusted according to the amplitude difference between the time domain predicted output signal and the reference signal template, and the priori signal frequency domain is updated.

[0084] The frequency domain response parameters of the deconvolution kernel are iteratively updated based on the updated priori signal frequency domain until the prediction error of the deconvolution harmonic compensation prediction model is less than a preset error threshold.

[0085] Specifically, the amplitude difference between the time domain predicted output signal and the reference signal template is calculated, that is, the amplitude difference value at each time point is obtained by comparing the amplitude values of the two at the same time point, and then the overall amplitude difference degree is determined by synthesizing these difference values. According to this difference degree, the amplitude matching coefficient is adjusted, that is, if the overall amplitude of the predicted output signal is higher than that of the reference signal template, the amplitude matching coefficient is appropriately reduced, and if the overall amplitude is lower than that of the reference signal template, the amplitude matching coefficient is correspondingly increased. Subsequently, the adjusted amplitude matching coefficient is applied to the priori signal frequency domain of the reference signal template for correction, thereby completing the update of the priori signal frequency domain, so that the updated priori signal frequency domain is more consistent with the amplitude characteristics of the actual signal.

[0086] Based on the updated prior signal frequency domain, the frequency domain response parameters of the deconvolution kernel are adjusted, the new prior signal frequency domain is substituted into the deconvolution operation process, the response values of the deconvolution kernel at each frequency point are recalculated, and the preliminary update of the parameters is realized. Then, the updated deconvolution kernel is used for prediction to obtain a new predicted output signal, and the prediction error between the signal and the reference signal template is calculated. If the error is greater than or equal to a preset error threshold, the above process of adjusting the frequency domain response parameters of the deconvolution kernel based on the updated prior signal frequency domain is repeated; if the error is less than the preset error threshold, the iteration is stopped, and at this time the deconvolution harmonic compensation prediction model reaches a convergent state.

[0087] In the second embodiment, based on the same inventive concept as the real-time dynamic compensation method for harmonic suppression of the frequency converter in the foregoing embodiments, as shown in the accompanying drawings, the present application provides a real-time dynamic compensation system for harmonic suppression of the frequency converter, and the system and method embodiments in the present application are based on the same inventive concept. The system comprises: Figure 2

[0088] The signal template construction module 10 is configured to construct a reference signal template, and the reference signal template is a reference signal type output by the frequency converter.

[0089] The compensation execution signal acquisition module 20 is configured to receive the harmonic-containing output signal of the frequency converter, compensate the harmonic-containing output signal based on the reference signal template, and obtain a compensation execution signal used for performing a compensation operation.

[0090] The predicted output signal acquisition module 30 is configured to analyze the compensation execution signal by using the deconvolution harmonic compensation prediction model, and acquire a predicted output signal of the frequency converter.

[0091] The harmonic suppression signal output module 40 is configured to acquire a deconvolution compensation signal according to the signal difference between the predicted output signal and the reference signal template, perform deconvolution stripping on the reference signal template by using the deconvolution compensation signal, and output a harmonic suppression signal of the frequency converter, wherein the signal types of the deconvolution compensation signal, the predicted output signal and the reference signal template are the same.

[0092] Further, the system is further configured to implement the following functions:

[0093] The system is further configured to implement the following functions:

[0094] ​Further, the system is also used to realize the following functions:

[0095] The first reference signal template and the second reference signal template are constructed, and the harmonic output signal is compensated according to the first reference signal template and the second reference signal template respectively, to output a first compensation execution signal and a second compensation execution signal; the first harmonic suppression signal and the second harmonic suppression signal are output according to the first compensation execution signal and the second compensation execution signal, and the first harmonic suppression signal and the second harmonic suppression signal are subjected to mean fitting to obtain the harmonic suppression signal of the frequency converter.

[0096] Further, the system is also used to realize the following functions:

[0097] The first compensation execution signal and the second compensation execution signal are analyzed by the deconvolution harmonic compensation prediction model to obtain the first prediction output signal and the second prediction output signal of the frequency converter; the first deconvolution compensation signal is obtained according to the signal difference between the first prediction output signal and the first reference signal template; the second deconvolution compensation signal is obtained according to the signal difference between the second prediction output signal and the second reference signal template; the first harmonic suppression signal of the frequency converter is output by deconvolution stripping of the first reference signal template using the first deconvolution compensation signal, and the second harmonic suppression signal of the frequency converter is output by deconvolution stripping of the second reference signal template using the second deconvolution compensation signal.

[0098] Further, the system is also used to realize the following functions:

[0099] The historical harmonic-free interference signal samples are classified according to the parameter range of the signal to obtain the first type of historical harmonic-free interference signal samples and the second type of historical harmonic-free interference signal samples; the first type of historical harmonic-free interference signal samples and the second type of historical harmonic-free interference signal samples are subjected to normalization processing respectively to output the first reference signal template and the second reference signal template.

[0100] Further, the system is also used to realize the following functions:

[0101] The compensation execution signal is subjected to band-pass filtering processing to output a preprocessed compensation execution signal; the input data of the preprocessed compensation execution signal is extracted, including time domain sequence, FFT frequency spectrum data and phase data; the deconvolution harmonic compensation prediction model uses the Tikhonov regularization deconvolution algorithm to take the input data as a convolution result, and performs adaptive convolution kernel update on the signal type of the reference signal template as a deconvolution prior output template to obtain the prediction output signal of the frequency converter.

[0102] Further, the system is further configured to implement the following functions:

[0103] Initialize a frequency domain response parameter of the deconvolution kernel, and a prior signal frequency domain of the reference signal template; update the prior signal frequency domain of the reference signal template based on the Tikhonov regularization deconvolution formula to obtain a predicted signal frequency domain; perform inverse Fourier transform on the predicted signal frequency domain to obtain a predicted output signal in the time domain; and update the frequency domain response parameter of the deconvolution kernel based on the predicted output signal in the time domain to obtain a converged deconvolution harmonic compensation prediction model, and output the predicted output signal of the frequency converter.

[0104] Further, the system is further configured to implement the following functions:

[0105] Update the frequency domain prior parameter of the reference signal template based on the Tikhonov regularization deconvolution formula; wherein the update includes adding a regularization parameter to the denominator of the Tikhonov regularization deconvolution formula to suppress noise amplification, and introducing the frequency domain prior of the reference signal template into the numerator of the Tikhonov regularization deconvolution formula to constrain the output signal form.

[0106] Further, the system is further configured to implement the following functions:

[0107] Adjust the amplitude matching coefficient based on the amplitude difference between the predicted output signal in the time domain and the reference signal template, update the prior signal frequency domain, and iteratively update the frequency domain response parameter of the deconvolution kernel based on the updated prior signal frequency domain until the prediction error of the deconvolution harmonic compensation prediction model is less than a preset error threshold.

[0108] It should be noted that the above sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. Moreover, the above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0109] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0110] The specification and drawings are, of course, to be regarded in an illustrative rather than a restrictive sense. It is to be understood that any such modifications, variations, combinations or equivalents that fall within the scope of the application are intended to be embraced herein.

Claims

1. A real-time dynamic compensation method for harmonic suppression in frequency converters, characterized in that, The method includes: Construct a reference signal template, wherein the reference signal template is a reference signal type output by the frequency converter; Receive the harmonic output signal of the frequency converter, and compensate the harmonic output signal based on the reference signal template to obtain a compensation execution signal for performing the compensation operation; The compensation execution signal is analyzed by an anti-convolution harmonic compensation prediction model to obtain the predicted output signal of the frequency converter; Based on the signal difference between the predicted output signal and the reference signal template, a deconvolution compensation signal is obtained. The reference signal template is deconvolved and stripped using the deconvolution compensation signal to output the harmonic suppression signal of the frequency converter. The signal types of the deconvolution compensation signal, the predicted output signal, and the reference signal template are the same. Methods for constructing reference signal templates include: Collect historical harmonic-free interference signal samples of the frequency converter and determine the signal type of the historical harmonic-free interference signal samples; The historical harmonic interference-free signal samples are subjected to signal normalization processing to obtain normalized amplitude and normalized phase. A reference signal template is constructed according to the normalized amplitude, normalized phase and the signal type, wherein the signal type of the reference signal template is a pulse signal. Methods for constructing reference signal templates also include: Construct the first reference signal template and the second reference signal template; The harmonic-containing output signal is compensated once according to the first reference signal template and the second reference signal template, and a first compensation execution signal and a second compensation execution signal are output. According to the first compensation execution signal and the second compensation execution signal, the first harmonic suppression signal and the second harmonic suppression signal are output, and the first harmonic suppression signal and the second harmonic suppression signal are subjected to mean fitting to obtain the harmonic suppression signal of the frequency converter.

2. The real-time dynamic compensation method for inverter harmonic suppression as described in claim 1, characterized in that, The method includes outputting a first harmonic suppression signal and a second harmonic suppression signal according to the first compensation execution signal and the second compensation execution signal, wherein the method comprises: The first compensation execution signal and the second compensation execution signal are analyzed by the deconvolution harmonic compensation prediction model to obtain the first prediction output signal and the second prediction output signal of the frequency converter. Based on the signal difference between the first predicted output signal and the first reference signal template, the first deconvolution compensation signal is obtained. Based on the signal difference between the second predicted output signal and the second reference signal template, the second deconvolution compensation signal is obtained; The first reference signal template is deconvolved and stripped using the first deconvolution compensation signal to output the first harmonic suppression signal of the frequency converter, and the second reference signal template is deconvolved and stripped using the second deconvolution compensation signal to output the second harmonic suppression signal of the frequency converter.

3. The real-time dynamic compensation method for inverter harmonic suppression as described in claim 1, characterized in that, The methods for constructing the first reference signal template and the second reference signal template include: The historical harmonic interference-free signal samples are classified into two categories according to the parameter range of the signal to obtain the first category of historical harmonic interference-free signal samples and the second category of historical harmonic interference-free signal samples. The first type of historical harmonic interference-free signal samples and the second type of historical harmonic interference-free signal samples are normalized respectively, and the first reference signal template and the second reference signal template are output.

4. The real-time dynamic compensation method for inverter harmonic suppression as described in claim 1, characterized in that, The method involves analyzing the compensation execution signal using a deconvolution harmonic compensation prediction model to obtain the predicted output signal of the frequency converter. The compensation execution signal is bandpass filtered to output a preprocessed compensation execution signal. Extract the input data of the preprocessed compensation execution signal, including time-domain sequence, FFT spectrum data, and phase data; The deconvolution harmonic compensation prediction model uses the Tikhonov regularized deconvolution algorithm to take the input data as the convolution result, and uses the signal type of the reference signal template as the deconvolution prior output template to perform adaptive convolution kernel update, so as to obtain the predicted output signal of the frequency converter.

5. The real-time dynamic compensation method for inverter harmonic suppression as described in claim 4, characterized in that, The method for obtaining the predicted output signal of the frequency converter includes: Initialize the frequency domain response parameters of the deconvolution kernel, and the frequency domain of the prior signal of the reference signal template; Under the conditions of the frequency domain response parameters, the frequency domain of the prior signal of the reference signal template is updated by regularization using the Tikhonov regularized deconvolution formula to obtain the frequency domain of the predicted signal. The predicted signal is subjected to inverse Fourier transform in the frequency domain to obtain the predicted output signal in the time domain. The frequency domain response parameters of the deconvolution kernel are iteratively updated based on the predicted output signal in the time domain to obtain a converged deconvolution harmonic compensation prediction model, and the predicted output signal of the frequency converter is output.

6. The real-time dynamic compensation method for inverter harmonic suppression as described in claim 5, characterized in that, The frequency domain prior parameters of the reference signal template are updated using the Tikhonov regularized deconvolution formula; The regularization update includes adding a regularization parameter to the denominator of the Tikhonov regularized deconvolution formula to suppress noise amplification, and introducing a frequency domain prior of the reference signal template into the numerator of the Tikhonov regularized deconvolution formula to constrain the output signal shape.

7. The real-time dynamic compensation method for inverter harmonic suppression as described in claim 5, characterized in that, The methods for obtaining a convergent deconvolution harmonic compensation prediction model include: The amplitude matching coefficient is adjusted by the amplitude difference between the predicted output signal in the time domain and the reference signal template, and the frequency domain of the prior signal is updated. The frequency domain response parameters of the deconvolution kernel are iteratively updated based on the updated prior signal frequency domain until the prediction error of the deconvolution harmonic compensation prediction model is less than a preset error threshold.

8. A real-time dynamic compensation system for harmonic suppression in a frequency converter, characterized in that, The system is used to implement the real-time dynamic compensation method for inverter harmonic suppression according to any one of claims 1-7, and the system comprises: The signal template construction module (10) is used to construct a reference signal template, wherein the reference signal template is a reference signal type output by the frequency converter; The compensation execution signal acquisition module (20) is used to receive the harmonic output signal of the frequency converter, and to compensate the harmonic output signal based on the reference signal template to obtain the compensation execution signal for performing the compensation operation. The prediction output signal acquisition module (30) is used to analyze the compensation execution signal through the deconvolution harmonic compensation prediction model to obtain the prediction output signal of the frequency converter; The harmonic suppression signal output module (40) is used to obtain a deconvolution compensation signal according to the signal difference between the predicted output signal and the reference signal template, and to perform deconvolution stripping on the reference signal template using the deconvolution compensation signal to output the harmonic suppression signal of the frequency converter, wherein the signal types of the deconvolution compensation signal, the predicted output signal, and the reference signal template are the same.

Images