Power factor correction method and system of high-power charging module

Abstract

The invention relates to the technical field of charging piles, and provides a power factor correction method and system for a high-power charging module, and the method comprises the steps: obtaining a power grid input voltage, a power grid input current and an output load current, constructing an input power fluctuation map and a load dynamic response map, carrying out the subgraph matching and joint convolution, and obtaining a power factor correction result; and obtaining a response change index table and a fitting weight vector, generating a dynamic current regulation vector by using the response change index table and the fitting weight vector, and performing phase correction and waveform reconstruction on the power grid input current according to the dynamic current regulation vector. Through phase correction and waveform reconstruction, the amplitude and phase information of the power grid input current are strictly aligned, the power factor of the high-power charging module is improved, the harmonic interference of the input side is reduced, the electric energy transmission efficiency and the module operation stability are improved, and the problem that in a complex power grid environment, the power consumption of the high-power charging module is reduced is solved. The problem that the correction effect of the power factor is not stable due to the fact that fast-changing input current and voltage are difficult to deal with is solved.

Description

Technical Field

[0001] This application relates to the technical field of charging piles, and particularly to a power factor correction method and system for high-power charging modules. Background Art

[0002] With the popularization of electric vehicles and the wide application of renewable energy, the load of the power system has gradually increased. Especially in high-power charging equipment, the requirements for the stability of the power grid are getting higher and higher. In the field of power electronics, power factor correction (PFC) technology has become one of the key technologies to improve the efficiency of electric energy use and ensure power quality. As a core component in electric vehicle charging stations, industrial equipment and other high-power applications, high-power charging modules must effectively control the power factor to reduce power losses, improve the efficiency of the power grid, and prevent electric energy waste and system overload.

[0003] In related technical means, the power factor correction of charging modules is usually achieved by using passive or active power factor correction circuits. Common methods include adjusting the waveform of the input current to synchronize it with the grid voltage waveform, thereby reducing the distortion of the current waveform and improving the power factor. In passive power factor correction schemes, passive components such as inductors and capacitors are usually used for filtering, so that the current waveform is as close to a sine wave as possible.

[0004] For the above technical solutions, although the existing active power factor correction schemes can significantly improve the power factor and reduce the grid load, in a complex grid environment, especially when the load fluctuates greatly or the grid quality is poor, the existing technologies often have difficulty coping with the rapidly changing input current and voltage, resulting in unstable power factor correction effects. Summary of the Invention

[0005] In order to improve the problem that in a complex grid environment, it is difficult to cope with the rapidly changing input current and voltage, resulting in unstable power factor correction effects, this application provides a power factor correction method and system for high-power charging modules.

[0006] The present invention provides a power factor correction method for a high-power charging module, including: obtaining the grid input voltage, the grid input current, and the output load current, constructing an input power fluctuation map based on the grid input voltage and the grid input current, and constructing a load dynamic response map based on the output load current; performing sub-map matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and constructing a target correction parameter set by using the response change index table and the fitting weight vector; performing weight scaling and mapping on the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and performing vector interpolation and waveform fitting on the dynamic frequency adjustment matrix by using the dynamic current adjustment vector generation framework to generate a dynamic current adjustment vector; performing phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector to achieve power factor correction of the high-power charging module.

[0007] As a preferred solution, the step of obtaining the grid input voltage, the grid input current, and the output load current, constructing an input power fluctuation map based on the grid input voltage and the grid input current, and constructing a load dynamic response map based on the output load current includes: obtaining the grid input voltage, the grid input current, and the output load current within a preset continuous period, performing normalization processing on the grid input voltage and the grid input current according to the phase period to obtain a normalized input waveform sequence and corresponding periodic synchronization timestamps; constructing a voltage-current phase cross sequence and a voltage transient slope sequence based on the normalized input waveform sequence, performing feature co-occurrence mapping processing on the voltage-current phase cross sequence and the voltage transient slope sequence to obtain a cross-phase map and a slope change map; performing amplitude difference structure superposition processing on the cross-phase map and the slope change map to obtain a voltage disturbance fusion map, and constructing an input power fluctuation map by using the voltage disturbance fusion map and the periodic synchronization timestamps; calculating the load change rate by using the output load current, performing section analysis on the load change rate to obtain load section characteristics and a load continuous change sequence, and constructing a load dynamic response map according to the load section characteristics and the load continuous change sequence.

[0008] As a preferred solution, the step of performing section analysis on the load change rate to obtain a section statistical feature set and a load continuous change sequence, and constructing a load dynamic response map according to the section statistical feature set and the load continuous change sequence includes: performing first-order difference processing and local extreme value analysis on the load change rate to obtain an initial change rate vector and a mutation boundary identification sequence, segmenting the initial change rate vector based on the mutation boundary identification sequence to obtain several load change sections, calculating the maximum change slope, average fluctuation range and duration of each section in all the load change sections to generate a section statistical feature set; performing time normalization processing on the section statistical feature set, constructing a feature interpolation curve based on the processed section statistical feature set, fitting the continuously changing regions in the feature interpolation curve to generate a load continuous change sequence; performing cross-fusion on the load continuous change sequence and the section statistical feature set to construct a hybrid response primitive, constructing a time series tensor mapping based on the hybrid response primitive, and performing spatio-temporal coupling coding on the load change behaviors at different time points by using the time series tensor mapping to obtain a change trend flux map; aligning the change trend flux map with a preset periodic synchronization timestamp to generate a time dimension calibration result, index-binding the time dimension calibration result and the hybrid response primitive to construct a map node and edge weight relationship, and constructing a load dynamic response map based on the map node and the edge weight relationship.

[0009] As a preferred solution, the step of performing sub-graph matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and constructing a target correction parameter set by using the response change index table and the fitting weight vector includes: obtaining an amplitude interval sequence and a phase shift sequence of consecutive multiple periods in the input power fluctuation map, performing spectral feature classification on the amplitude interval sequence and the phase shift sequence to obtain an amplitude structure group and a phase shift pattern group; obtaining a load change section feature set corresponding to the consecutive periods in the input power fluctuation map in the load dynamic response map, performing sub-graph matching on the load change section feature set and the phase shift pattern group to obtain a load interference influence domain and an input perturbation sensitive domain; performing multi-dimensional matrix joint convolution on the amplitude structure group and the load interference influence domain to obtain a perturbation collaborative coupling matrix, and performing section cross-filtering on the perturbation collaborative coupling matrix and the input perturbation sensitive domain to generate a response change index table and a fitting weight vector; performing time series reduction on the response change index table to obtain a load response change trajectory group, and performing error amplitude diffusion processing on the fitting weight vector to obtain a weight perturbation error set; dynamically weighting and fusing the load response change trajectory group and the weight perturbation error set to obtain a target correction parameter set.

[0010] As a preferred solution, the steps of performing multi-dimensional matrix joint convolution on the amplitude structure group and the load interference influence domain to obtain a disturbance cooperative coupling matrix, and performing segment cross filtering on the disturbance cooperative coupling matrix and the input disturbance sensitive domain to generate a response change index table and a fitting weight vector include: constructing a periodic amplitude vector set based on the amplitude structure group, constructing a time synchronous disturbance field based on the load interference influence domain, performing vector domain expansion on the periodic amplitude vector set and the time synchronous disturbance field to obtain an amplitude disturbance tensor and a disturbance density tensor; performing kernel convolution on the amplitude disturbance tensor and the disturbance density tensor to obtain a periodic disturbance. A dynamic induction matrix is prepared, and the input disturbance sensitive domain is mapped based on the periodic disturbance induction matrix to obtain a response coupling factor and a phase response module; a trend comparison mapping is performed on the response coupling factor, and a change intensity spectrum is generated based on the mapping result; an error expansion transformation is performed on the phase response module to obtain a phase anomaly weight spectrum; the change intensity spectrum and the phase anomaly weight spectrum are jointly filtered to output a response change index table; a weight fitting analysis is performed on the response change index table to obtain a frequency band disturbance residual vector, and the weight value of the disturbance source that best matches the frequency band disturbance residual vector is calculated according to the frequency band disturbance residual vector, and a fitting weight vector is generated by aggregation.

[0011] As a preferred solution, the step of performing weight scaling and mapping on the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and using the dynamic current adjustment vector generation framework to perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix to generate a dynamic current adjustment vector includes: parsing out a frequency distribution vector based on the target correction parameter set, splitting the frequency distribution vector into frequency bands to obtain a low-frequency subset and a high-frequency subset, applying a band-pass screening network to extract the edge perturbation factor of the high-frequency subset, applying a multiple integral filter to extract the period offset factor of the low-frequency subset, performing a mixed feature deconstruction on the edge perturbation factor and the period offset factor to obtain a coupled frequency-domain projection array and a frequency perturbation overlapping region; mapping the coupled frequency-domain projection array to a preset perturbation phase diagram space to obtain a frequency slope field and a response fine-tuning domain, mapping the frequency perturbation overlapping region to a preset adjustment amplitude space to obtain an amplitude adjustment coefficient group and a slope interference term; performing an inter-domain factor interleaving fusion based on the frequency slope field and the amplitude adjustment coefficient group to obtain a high-frequency frequency-domain modulation factor and a low-frequency frequency-domain weighting factor, performing a matching optimization on the slope interference term and the response fine-tuning domain to generate a dynamic response delay vector; performing a joint weight scaling on the high-frequency frequency-domain modulation factor, the low-frequency frequency-domain weighting factor, and the dynamic response delay vector to obtain a dynamic frequency adjustment matrix, performing an inverse modulation mapping on the dynamic frequency adjustment matrix and the target correction parameter set to construct a dynamic current adjustment vector generation framework; performing vector interpolation and waveform fitting on the dynamic frequency adjustment matrix based on the dynamic current adjustment vector generation framework to generate a dynamic current adjustment vector.

[0012] As a preferred solution, the step of performing phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector includes: performing a period broadening process on the dynamic current adjustment vector, and extracting a sub-period current vector group based on the broadening result; sampling and segmenting the grid input current to obtain a plurality of current waveforms, applying a multi-layer wavelet packet decomposition to decompose all the current waveforms to generate an amplitude perturbation reference set; performing a sub-segment convolution matching on the sub-period current vector group and the amplitude perturbation reference set to extract a perturbation fitting factor; constructing an amplitude fitting conversion matrix using the perturbation fitting factor, applying the amplitude fitting conversion matrix to perform phase correction and amplitude correction on each segment of the grid input current to obtain corrected phase information and corrected amplitude information; fusing the corrected phase information and the corrected amplitude information, and constructing a phase-amplitude coupling envelope based on the fusion result, applying a fast Fourier reconstruction method to perform waveform reconstruction on the phase-amplitude coupling envelope.

[0013] The present application also provides a power factor correction system for a high-power charging module, including: an acquisition module, configured to acquire the grid input voltage, the grid input current, and the output load current, construct an input power fluctuation map based on the grid input voltage and the grid input current, and construct a load dynamic response map based on the output load current; a construction module, configured to perform sub-graph matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and construct a target correction parameter set by using the response change index table and the fitting weight vector; a fitting module, configured to perform weight scaling and mapping on the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix by using the dynamic current adjustment vector generation framework to generate a dynamic current adjustment vector; a correction module, configured to perform phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector to achieve power factor correction of the high-power charging module.

[0014] Compared with the prior art, the present application has the following beneficial effects: The correction effect is accurate and stable. By acquiring the input grid voltage, current, and load current signals of the high-power charging module, constructing an input power fluctuation map and a load dynamic response map respectively as the basic inputs, and using sub-graph matching joint convolution, the coupling interference characteristics between the load and the input side are extracted, and a target correction parameter set is generated; Subsequently, through the dynamic frequency adjustment matrix and the dynamic current adjustment vector generation framework, the correction angle of the high-power harmonics and the current adjustment direction are determined. Through phase correction and waveform reconstruction, the amplitude and phase information of the grid input current are strictly aligned, effectively improving the power factor of the high-power charging module, reducing the harmonic interference on the input side, improving the power transmission efficiency and the module operation stability, and solving the problem that it is difficult to cope with the rapidly changing input current and voltage in a complex grid environment, resulting in unstable correction effect of the power factor. Description of the Drawings

[0015] 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 the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] The structures, ratios, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those skilled in this technology to understand and read, and are not used to limit the limiting conditions for the implementation of the present invention. Therefore, they do not have substantial technical significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that can be covered by the technical content disclosed in the present invention.

[0017] Figure 1 It is a schematic flowchart of the power factor correction method of the high-power charging module provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of the power factor correction system of the high-power charging module provided by an embodiment of the present invention.

[0018] Explanation of reference numerals: 10. Power factor correction system of the high-power charging module; 11. Acquisition module; 12. Construction module; 13. Fitting module; 14. Correction module. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] The flowchart shown in the drawings is only an example illustration, and does not necessarily include all contents and operations / steps, nor does it necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined, or partially merged. Therefore, the actual execution order may be changed according to the actual situation.

[0021] It should also be understood that the terms used in this specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this specification of the present application and the appended claims, unless the context clearly indicates otherwise, the singular forms of "a", "an", and "the" are intended to include the plural forms.

[0022] It should be further understood that the term "and / or" used in this specification of the present application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0023] Next, the technical solutions of the present invention will be further described in conjunction with the drawings and through specific implementation manners.

[0024] Embodiment 1: As Figure 1 shown, the present application provides a power factor correction method for a high-power charging module, including steps S100 to S400.

[0025] Step S100: Obtain the grid input voltage, grid input current, and output load current, construct an input power fluctuation map based on the grid input voltage and grid input current, and construct a load dynamic response map based on the output load current.

[0026] In this step, by controlling the high-frequency signal acquisition module, continuously sample the grid input voltage, grid input current, and output load current of the high-power charging module, and at the same time bind the time synchronization signal to ensure that the voltage and current signal data have a strict time synchronization relationship. Specifically, normalize the collected grid input voltage and grid input current according to a preset phase period to obtain a normalized input waveform sequence and a periodic synchronization timestamp; based on this normalized waveform sequence, further construct a voltage-current phase cross sequence and a voltage transient slope sequence, and perform feature co-occurrence mapping processing on these two sequences to generate a cross-phase map and a slope change map; subsequently, perform amplitude difference superposition processing on the cross-phase map and the slope change map to generate a voltage disturbance fusion map with frequency band characteristics; combine the periodic synchronization timestamp of the input signal, and use this fusion map to construct an input power fluctuation map. At the same time, perform time series analysis on the output load current, calculate its load change rate, and extract load change segment features and continuous change sequences through section analysis; combine the load change segment features and continuous change sequences, and map to generate a load dynamic response map.

[0027] For example, for the grid input voltage and grid input current signals of each complete sampling period, they can be divided into multiple equal time segments, and the normalized value and phase difference of each time segment are calculated in turn to obtain a complete normalized waveform sequence. At the same time, use the periodic synchronization timestamp to mark the occurrence positions of the transient voltage change (voltage slope) and intersection points (voltage and current zeros) recorded in each segment. Based on this, the amplitude changes of each segment can be superimposed and statistically analyzed to form a more refined voltage disturbance fusion map output. The section analysis of the load current can be marked by setting the interval length where the load change rate exceeds a preset threshold, and a dynamic response block map with significant features is generated.

[0028] Step S200: Perform sub-graph matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and use the response change index table and the fitting weight vector to construct a target correction parameter set.

[0029] In this step, the input power fluctuation map and the load dynamic response map are subjected to sub-map matching processing through a map matching algorithm. Among them, the voltage oscillation region of the input power fluctuation map is connected and matched with the load change section of the load dynamic response map to extract the interference response characteristics that occur repeatedly. Specifically, by applying multi-dimensional joint kernel convolution operations to each sub-map region, the amplitude superposition result of strong interference frequency components is calculated to form a cooperative coupling matrix for the strong interference frequency band. At the same time, the unmatched weak interference blocks are filtered to generate a response change index table. Through the above convolution and matching processing, the periodic exponential weight, amplitude perturbation data, and phase true value characteristics are extracted to generate a fitting weight vector. The response change index table and the fitting weight vector are further combined to coordinately adjust the interference response characteristics on the input side and the load side, and finally, a target correction parameter set is output.

[0030] For example, in map matching, a strong coupling occurs between the voltage amplitude mutation section shown in the input power fluctuation map and the continuous change section with the highest load change rate in the load dynamic response map. This matching relationship can be evaluated by calculating the correlation index between the voltage transient slope and the load change rate, and then reflected in the target correction parameters through the normalization adjustment of the exponential weight. The result of convolution filtering can also be further used to strengthen the matching processing of adjacent load and input symmetric perturbations, thereby optimizing the generation of the index table and the weight vector.

[0031] Step S300: Scale and map the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and use the dynamic current adjustment vector generation framework to perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix to generate a dynamic current adjustment vector.

[0032] In this step, by analyzing the distribution characteristics of the strong interference frequency band and the weak interference frequency band in the target correction parameter set, the correction weight values corresponding to different frequency bands are scaled to generate a dynamic frequency adjustment matrix. Specifically, the phase offset factor and the amplitude mutation factor in the target correction parameter set are respectively mapped to the matrix rows and columns of the dynamic frequency adjustment matrix to construct a two-dimensional frequency-amplitude adjustment framework. Subsequently, local optimization interpolation is performed on the dynamic frequency adjustment matrix through vector interpolation technology to generate smoother spectrum data. Based on the dynamic current adjustment vector generation framework, the current vector input end is constructed using the frequency correspondence relationship in the above matrix to realize the generation of the dynamic current adjustment vector.

[0033] For example, a certain strong interference frequency band recorded in the target correction parameter set can be parsed by the dynamic frequency adjustment matrix as the row with the highest weight in the matrix, and this row corresponds to the dynamic current response of the target frequency region. In the generation framework, the frequency region corresponding to this high weight is further extended to the adjacent weak interference segments in the matrix through interpolation, thereby enhancing the data continuity of the interpolation region. The current adjustment vector generated by the fitting technique contains more refined frequency band correction factors and current adjustment coefficients.

[0034] Step S400: Perform phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector to achieve power factor correction of the high-power charging module.

[0035] In this step, by applying the inverse multi-dimensional Fourier transform method, the phase spectrum data in the grid input current is split using the dynamic current adjustment vector, and at the same time, the frequency band data with a large phase delay is corrected in advance, and the waveform section with a large amplitude mutation is reconstructed. Specifically, the corrected phase data and amplitude characteristic data are fused to generate a dynamic coupling envelope curve; the envelope curve is processed by fast Fourier transform to complete the frequency domain reduction of the reconstructed waveform, and the calculation of the sampling points in the next cycle is iteratively optimized according to the reconstructed waveform to ensure the correction coherence.

[0036] For example, in the process of reconstructing a certain typical waveform, if the dynamic vector shows that the phase drift of a certain frequency band in the input current exceeds the preset threshold, the data of this frequency band will be preferentially processed, and its corrected dynamic envelope curve will smoothly transition to the next frequency band in a continuously changing form, thereby achieving smooth correction in the output waveform.

[0037] In this embodiment, by acquiring the grid input voltage, grid input current, and output load current signals, an input power fluctuation map is constructed based on the grid input voltage and grid input current, and a load dynamic response map is constructed based on the output load current. Then, sub-map matching and joint convolution are performed on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and a target correction parameter set is constructed using these two results. Next, weight scaling and mapping are performed on the target correction parameter set to obtain a dynamic frequency regulation matrix and a dynamic current regulation vector generation framework. Under this framework, vector interpolation and waveform fitting are performed on the dynamic frequency regulation matrix, and finally, a dynamic current regulation vector is generated. Finally, based on the generated dynamic current regulation vector, phase correction and waveform reconstruction are performed on the grid input current, and finally, power factor correction of the high-power charging module is achieved. By combining the input power fluctuation map and the load dynamic response map and using the methods of joint convolution and sub-map matching, the relationship between the grid input voltage, grid input current, and load current can be accurately captured, and the key features required for correction can be extracted. Through the generation framework of the dynamic frequency regulation matrix and the dynamic current regulation vector, efficient and accurate current regulation is achieved, the power factor is optimized, and the power transmission efficiency of the high-power charging module is significantly improved. Through phase correction and waveform reconstruction, the fluctuations of the grid and the load can be adapted, the stability of the system can be improved, grid interference can be reduced, and thus the overall energy utilization rate and system reliability can be improved, and the problem that it is difficult to cope with the rapidly changing input current and voltage in a complex grid environment, resulting in unstable power factor correction effect, can be improved.

[0038] Embodiment 2: In step S100, the grid input voltage, grid input current, and output load current within a preset continuous period are acquired, and the grid input voltage and grid input current are normalized according to the phase period to obtain a normalized input waveform sequence and the corresponding period synchronization time stamps.

[0039] By successively performing high-precision continuous sampling on the grid input voltage and current signals, and through digital filtering and synchronization processing, the original data of the voltage signal and the current signal within a set of preset continuous periods is obtained, ensuring that the sampling frequency meets the Nyquist criterion. Specifically, first, a high-frequency data collector is used to sample the voltage signal and the current signal within a preset period T and record the time, and the timestamp of each sampling point is marked with the assistance of a hardware synchronization module; then, by cutting the period boundaries of the sampled signals, a continuous periodic waveform that is complete and has no breakpoints is extracted; a period normalization algorithm is applied to the voltage signal and the current signal to map the signal amplitude, time dimension, and phase information to a standardized axis, generating a normalized voltage input waveform sequence and a normalized current input waveform sequence; an accurate period synchronization timestamp is generated based on the normalized signal sequence, and the phase drift of the input signal is corrected at the secondary level in combination with the timestamp to ensure that the time starting point of the normalized waveform is synchronized with the standard period.

[0040] For example, in the analysis of a certain input signal, assume that the grid operating frequency is 50 Hz and the period T is 20 ms. The voltage and current signals of 50 periods can be sampled within a 1 s time window. Assuming the sampling frequency is 100 kHz, each period will contain 2,000 sampling points. After acquisition, the amplitude of the original data of the voltage and current signals is normalized, with the maximum amplitude standardized to 1 and the minimum amplitude normalized to 0; subsequently, the phase offset of each period is calculated based on phase analysis, and the phase is adjusted by sliding forward or backward to align the time points of each period, generating a normalized waveform sequence. Finally, a starting timestamp record is generated for each period through the time synchronization module, such as 20 ms, 40 ms, which are accumulated in sequence as the output of the period synchronization timestamp sequence.

[0041] Based on the normalized input waveform sequence, a voltage-current phase cross sequence and a voltage transient slope sequence are constructed, and feature co-occurrence mapping processing is performed on the voltage-current phase cross sequence and the voltage transient slope sequence to obtain a cross-phase spectrum and a slope change spectrum.

[0042] Sequence information of voltage-current phase crossover points is obtained by extracting features from the time position and amplitude characteristics of the normalized input waveform sequence, and the slope of the normalized voltage signal is calculated to generate a transient slope value matrix. Specifically, by extracting the positions where the normalized voltage and current signals cross zero, recording the crossover points of the two on the time axis, and the relative phase values at the time of crossover, a time-phase data sequence is formed; the continuous point difference of the normalized voltage input sequence in the time axis direction is calculated to obtain the voltage change rate at each moment; combined with a preset threshold, the points with a change rate exceeding the amplitude threshold are selected as significant mutation points, and the transient characteristics of the mutation points are recorded; the phase crossover sequence and the transient slope sequence are subjected to feature joint mapping, and based on the alignment relationship on the time axis of the two, a two-dimensional map containing the simultaneous distribution of phase crossover characteristics and mutation slope characteristics is generated, namely the cross-phase map and the slope change map.

[0043] For example, in the normalized input voltage and current signals, if the zero-crossing time of the voltage signal in a certain period T is t1 = 5ms, and the zero-crossing time of the current signal is t2 = 7ms, then the phase difference between the two on the time axis is (50Hz, Δt = 2ms), and the corresponding phase offset angle is 36 degrees. The slope change rate of the voltage signal at t = 6ms in the same period is calculated as ΔV / Δt = 0.5V / 0.1ms, which exceeds the preset threshold (such as 0.4V / ms), so it is marked as a voltage transient point. By mapping the phase characteristics and slope characteristics of these points on the map, the dynamic change distribution law of the transient slope value when the current lags the phase by more than 30 degrees can be clearly observed.

[0044] The cross-phase map and the slope change map are subjected to amplitude difference structure superposition processing to obtain a voltage disturbance fusion map, and the input power fluctuation map is constructed by using the voltage disturbance fusion map and the cycle synchronization timestamp.

[0045] By performing structured superposition on the amplitude change rates of the cross-phase map and the slope change map, combined with the joint weight assignment of the time axis and phase offset, a complete map data representation of voltage disturbance fusion is generated. Specifically, for each position point in the cross-phase map and the slope change map, the weighted average of the cross-point amplitude and the transient change rate value is calculated in turn to reduce the influence of single-point deviation in the original map, and linear interpolation or extension is performed on the non-overlapping points in the two maps. The combined values generated by superposition are classified by frequency interval and rearranged according to the voltage periodic characteristics to obtain a voltage disturbance characteristic fusion map. The cycle synchronization timestamp is introduced into the fusion map to mark the voltage disturbance intensity in a specific time period in real time, and a complete input power fluctuation map is constructed.

[0046] For example, in a certain period, assuming that the weight value of a certain point in the cross-phase spectrum is 0.7 and the corresponding point value in the slope change spectrum is 0.5, then through the amplitude difference structure superposition calculation (such as the weighted formula is 0.6×phase value + 0.4×slope value), the final fusion value of this point is obtained as 0.62. In the fusion spectrum, this value is mapped to the voltage disturbance area where the 50 Hz frequency is located, and then by inserting the cycle synchronization timestamp information to clarify the actual time point corresponding to this disturbance (such as: t = 12 ms).

[0047] Calculate the load change rate using the output load current, perform a section analysis on the load change rate, obtain the load section characteristics and the load continuous change sequence, and construct a load dynamic response spectrum based on the load section characteristics and the load continuous change sequence.

[0048] By continuously extracting the change rate of the output load current signal and combining the local feature clustering analysis method, the section change pattern and time series in the load signal are finely analyzed and modeled. Specifically, perform a first-order difference calculation on the output load current signal, obtain a complete change rate sequence through the current mutation rate extraction, divide the entire change rate sequence into sections by setting a change rate threshold, and extract the maximum change slope, average change range, and time length for each section to generate a section feature set containing multiple section description attributes. Utilize the time distribution characteristics of the section feature set to jointly encode the section features and the time continuous sequence of the load change signal, and display the load dynamic change concentration areas in different time periods on the time axis to form a load dynamic response spectrum.

[0049] For example, on a load current signal curve, assuming that the current rises from 10 A to 20 A in the time period from t1 = 100 ms to t2 = 120 ms, calculate its change rate as 0.5 A / ms. Subsequently, the curve suddenly drops to 5 A in the time period from t3 = 200 ms to t4 = 220 ms, and the change rate is -0.75 A / ms. According to the change rate value, load section characteristics such as slope, average change amplitude, and section time can be extracted, significant changes in different time periods can be marked, and these attributes are combined with the section sequence to construct a load dynamic response spectrum.

[0050] Among them, the steps of performing a section analysis on the load change rate, obtaining the section statistical feature set and the load continuous change sequence, and constructing a load dynamic response spectrum based on the section statistical feature set and the load continuous change sequence include: performing a first-order difference processing and local extreme value analysis on the load change rate to obtain an initial change rate vector and a mutation boundary identification sequence, performing boundary segment segmentation on the initial change rate vector based on the mutation boundary identification sequence to obtain several load change sections, and calculating the maximum change slope, average fluctuation range, and duration of each section in all load change sections to generate a section statistical feature set.

[0051] By performing a first-order difference calculation on the output load current signal, a change rate vector of the load current signal is obtained, and it is calibrated in combination with the local extreme value analysis method. Specifically, first, the change rate of the load current signal is calculated, and the formula is: The speed information of the load current changing with time is obtained through this formula.

[0052] Then, the local extreme value analysis algorithm is applied to identify the regions with large fluctuations in the load current. These regions usually represent the time periods when the load undergoes sudden changes or drastic changes, and are called "mutation boundaries". Based on the mutation boundary identification sequence, the change rate vector is segmented, and the load change signal on the time axis is divided into several change segments. Each segment represents a different stage of the load current change. Calculate the maximum change slope, average fluctuation range, and duration of each segment to obtain the statistical characteristics of the segment, and these characteristics constitute the segment statistical characteristic set.

[0053] For example, assume that in the sampled load current signal, the load current changes significantly between the 10th ms and the 50th ms, and the current increases from 10 A to 20 A. The change rate calculated for this period is: Through local extreme value analysis, it is identified that there is a peak in the current change at 50 ms, which is the mutation boundary. Next, this mutation boundary is used to segment the change rate vector, generating a load change segment. For this segment, calculate its maximum change slope to be 0.25 A / ms, average fluctuation range to be 10 A, and duration to be 40 ms, and these characteristics constitute the segment statistical characteristic set.

[0054] Perform time normalization on the segment statistical characteristic set, construct a characteristic interpolation curve based on the processed segment statistical characteristic set, and fit the continuously changing regions in the characteristic interpolation curve to generate a load continuous change sequence.

[0055] By performing time normalization on the segment statistical characteristic set, the characteristic data of different segments are mapped to a unified time scale. Specifically, the duration of each segment is standardized, and the durations of all segments are scaled to the same range (for example, normalized to between 0 and 1). Then, based on the time-normalized statistical characteristic set, an interpolation algorithm is used to construct a characteristic interpolation curve. The interpolation curve can accurately describe the details of the load current changing with time, especially for the regions where the load change is relatively gentle. Finally, the continuously changing regions are extracted from the characteristic interpolation curve, and a fitting algorithm (such as polynomial fitting or spline interpolation) is used to generate a load continuous change sequence, which is used to reflect the gradual change trend of the load current over a long period of time.

[0056] For example, assume that within a section of the load current, with a duration ranging from 0 ms to 100 ms, through time normalization, this section is mapped to a time interval between 0 and 1. Then, based on the characteristic data of this normalized section (such as the maximum change slope, average fluctuation range, etc.), the spline interpolation algorithm is used to generate the interpolation curve for this section. This curve can smoothly reflect the changing trend of the load current over time. Next, for the relatively flat regions of change, the change sequence in this region can be further optimized through fitting methods.

[0057] Cross-fuse the continuous load change sequence and the section statistical feature set to construct a hybrid response primitive. Based on the hybrid response primitive, construct a time series tensor mapping, and use the time series tensor mapping to perform spatio-temporal coupling encoding on the load change behavior at different time points to obtain a change trend flux map.

[0058] By cross-fusing the continuous load change sequence and the section statistical feature set, a hybrid response primitive with temporal and spatial characteristics is extracted. Specifically, each data point in the continuous load change sequence and the section statistical feature set is cross-fused to capture the key temporal characteristics during the load change process; construct a time series tensor mapping, and use the characteristics and statistical information of each time point in the load change sequence to map it into a multi-dimensional space to form a data structure containing spatio-temporal coupling relationships. Finally, through spatio-temporal coupling encoding, these mapping results are converted into a change trend flux map, and the flux map represents the trend and key characteristics of the load current changing over time.

[0059] Assume that in the load current change sequence, for a load section, the current value corresponding to a certain point on the interpolation curve is 15 A, the duration is 50 ms, and the section characteristic of this point is that the change slope is 0.3 A / ms. Through cross-fusion, these characteristic data are combined with the corresponding time points in the continuous change sequence to form a hybrid response primitive. Then, these primitives are mapped into the time series tensor, where each time point corresponds to the change rate of the load current and other characteristics. Through spatio-temporal coupling encoding, these data form a detailed trend map of the load current change in the flux map.

[0060] Align the change trend flux map with the preset periodic synchronization timestamp to generate the time dimension calibration result. Index-bind the time dimension calibration result with the hybrid response primitive to construct the graph node and edge weight relationship, and construct the load dynamic response graph based on the graph node and edge weight relationship.

[0061] By aligning the change trend flux diagram with the preset periodic synchronization timestamps, the change characteristics of the load current at different time points are accurately calibrated. Specifically, the timestamps are made to correspond one by one with each data point in the change trend flux diagram to ensure that the time information of each node in the diagram is synchronized with the actual measurement data. Based on these calibration results, the hybrid response primitives are associated with the timestamps through index binding to construct the diagram nodes. The relationship between the diagram nodes is quantified by constructing edge weights, and the dependency relationship between the nodes is represented in the form of a weighted graph to form the load dynamic response diagram.

[0062] For example, assume that within a certain load current section, the change trend of the load current is relatively stable at t = 10 ms, while a large fluctuation occurs at t = 50 ms. By aligning the change trend flux diagram with the timestamps, the position of the node at t = 50 ms in the diagram can be determined, and through connection with weighted edges to other nodes, the complete diagram structure of the load dynamic response diagram can be obtained.

[0063] In step S200, the amplitude interval sequences and phase shift sequences of multiple consecutive periods in the input power fluctuation diagram are obtained, and the spectral characteristics of the amplitude interval sequences and phase shift sequences are classified to obtain the amplitude structure group and the phase shift mode group.

[0064] Through segmented analysis of the input power fluctuation diagram, the amplitude interval sequences and phase shift sequences of each period are extracted and classified. Specifically, by performing periodic segmentation on the amplitude in the input power fluctuation diagram, the amplitude data of each period is divided into several intervals, and the amplitude range and fluctuation conditions of each interval are recorded. The phase difference between the input voltage and current is sampled periodically to obtain the phase shift data in each period, and the phase change between different periods of the voltage and current is recorded. Based on the extracted amplitude intervals and phase shift sequences, frequency domain analysis is performed on them. The amplitude interval sequences are classified according to different amplitude ranges, and at the same time, the phase shift sequences are classified according to the change patterns to obtain the amplitude structure group and the phase shift mode group.

[0065] For example, assume that within a sampling period, the amplitude change range of the voltage and current is from 10 V to 50 V. After classification, the amplitude interval of this period is divided into 5 sub-intervals, and the maximum amplitude range and minimum amplitude range of each sub-interval are recorded. At the same time, by calculating the phase difference between the voltage and current, assume that the phase shift change increases from 30° to 60° during a certain period in the cycle, which is classified as a phase shift mode. After analyzing multiple periods, all the amplitude intervals and phase shift modes are classified and grouped into the amplitude structure group and the phase shift mode group.

[0066] Obtain the feature set of the load change segments within consecutive periods in the input power fluctuation map corresponding to the load dynamic response map. Perform subgraph matching between the load change segment feature set and the phase shift pattern group to obtain the load interference influence domain and the input disturbance sensitive domain.

[0067] By analyzing the corresponding periodic data in the load dynamic response map and the input power fluctuation map, extract the characteristics of the load current change segments and match them with the phase shift pattern group. Specifically, calculate the increasing and decreasing trends of the load current using the change rate of the load current, thereby identifying the significant change segments of the load current. These change segments reflect the key characteristics of the load response. Extract the maximum change slope, fluctuation amplitude, and duration for each change segment to form the load change segment feature set. By matching the load change segment feature set with the data in the phase shift pattern group, identify the corresponding relationship between the regions with strong frequency response in the input power fluctuation map and the load changes, and obtain the load interference influence domain and the input disturbance sensitive domain.

[0068] For example, assume that within one period, the change segment of the load current ranges from 10A to 15A, and the change rate is 0.1A / ms. This change segment will be marked as a load change segment, and its maximum change slope of 0.1A / ms, fluctuation range of 5A, and duration of 50ms will be extracted. During subgraph matching, it is found that this load change segment has a significant correlation with a certain pattern in the phase shift pattern group (phase shift from 40° to 80°), identifying the load interference influence domain and the input disturbance sensitive domain.

[0069] Perform multi-dimensional matrix joint convolution on the amplitude structure group and the load interference influence domain to obtain the perturbation co-coupling matrix. Perform section cross-filtering on the perturbation co-coupling matrix and the input disturbance sensitive domain to generate the response change index table and the fitting weight vector.

[0070] By performing multi-dimensional matrix convolution on the amplitude structure group and the load interference influence domain, extract the co-characteristics between the amplitude and the load changes. Specifically, perform convolution operation on the feature matrices of the amplitude structure group and the load interference influence domain to obtain the perturbation co-coupling matrix, which represents the spatio-temporal correlation between the amplitude changes and the load changes. Perform cross-filtering on the frequency components in the perturbation co-coupling matrix to extract the frequency band information related to the input disturbance sensitive domain, reduce the influence of irrelevant perturbations, and generate the response change index table. Based on the results of cross-filtering, further analyze the influence of each frequency band and generate the fitting weight vector for subsequent dynamic adjustment and optimization.

[0071] For example, within one period, the amplitude variation range of a certain section of the amplitude structure group is from 0.5V to 1.5V, and the corresponding frequency band in the load interference influence domain is from 50Hz to 100Hz. After multi-dimensional convolution, a perturbation co-coupling matrix is obtained, which reflects the coupling effect between amplitude variation and load variation in this frequency band. Then, through cross filtering, high-frequency noise unrelated to load variation is removed, a response variation index table is generated, and a fitting weight vector is generated according to the contribution degree of different frequency bands to the perturbation.

[0072] Among them, the steps of performing multi-dimensional matrix joint convolution on the amplitude structure group and the load interference influence domain to obtain a perturbation co-coupling matrix, and performing section cross filtering on the perturbation co-coupling matrix and the input perturbation sensitive domain to generate a response variation index table and a fitting weight vector include: constructing a periodic amplitude vector set based on the amplitude structure group, constructing a time-synchronized perturbation field based on the load interference influence domain, and performing vector domain expansion on the periodic amplitude vector set and the time-synchronized perturbation field to obtain an amplitude perturbation tensor and an interference density tensor.

[0073] By performing joint convolution on the amplitude structure group and the load interference influence domain, and using the time-synchronized perturbation field to map these signals, an amplitude perturbation tensor and an interference density tensor are generated. Specifically, based on the periodic data in the amplitude structure group, the amplitude change vector of each period is extracted to construct a periodic amplitude vector set. Time synchronization processing is performed on the load interference influence domain to generate a perturbation field, ensuring that the influence of load perturbation on the amplitude is synchronized at different time points. The periodic amplitude vector set is combined with the time-synchronized perturbation field and vector domain expansion is performed to obtain an amplitude perturbation tensor and an interference density tensor.

[0074] For example, assume that within a certain period, the amplitude vector is [0.1, 0.3, 0.5], and the perturbation field within the load interference influence domain is [1.0, 0.8, 0.6]. Through vector domain expansion, they are mapped into a common time domain to obtain an amplitude perturbation tensor [0.1, 0.3, 0.5] and an interference density tensor [1.0, 0.8, 0.6], which are used for further co-convolution and subsequent processing.

[0075] Perform kernel convolution processing on the amplitude perturbation tensor and the interference density tensor to obtain a periodic perturbation induction matrix, and map the input perturbation sensitive domain based on the periodic perturbation induction matrix to obtain a response coupling factor and a phase response module.

[0076] By performing kernel convolution processing on the amplitude perturbation tensor and the interference density tensor, a periodic perturbation induction matrix is extracted, and based on this matrix, the mapping of the input perturbation sensitive domain is carried out. Specifically, a convolution operation is performed on the amplitude perturbation tensor and the interference density tensor to obtain a periodic perturbation induction matrix, which reflects the temporal correlation between the perturbation source and the load current. Based on the periodic perturbation induction matrix, it is mapped to the input perturbation sensitive domain to identify the frequency band information closely related to the change in the load current. According to the mapping result, a response coupling factor and a phase response module are generated to further optimize the regulation strategy of the load current.

[0077] For example, by performing kernel convolution processing on the amplitude perturbation tensor [0.1, 0.2, 0.3] and the interference density tensor [0.5, 0.4, 0.3], a periodic perturbation induction matrix [0.15, 0.2, 0.25] is obtained, and these matrices reflect the correlation between the perturbation source and the load current at different time points. Based on these results, a response coupling factor and a phase response module are further generated to ensure the coordination between the load current and the grid signal.

[0078] Perform a trend comparison mapping on the response coupling factor, generate a change intensity spectrogram based on the mapping result, perform an error expansion transformation on the phase response module to obtain a phase anomaly weight spectrogram, and perform a joint filtering process on the change intensity spectrogram and the phase anomaly weight spectrogram to output a response variation index table.

[0079] By comparing the response coupling factor with the input power fluctuation and the load change trend, a change intensity spectrogram is generated; at the same time, an error process is performed on the phase response module to obtain a phase anomaly weight spectrogram, and the two are combined to generate a response variation index table. Specifically, a time series analysis is performed on the response coupling factor and compared with the trend data of the reference input power fluctuation spectrum and the load dynamic response spectrum. By calculating the similarity between the two (for example, using the correlation coefficient or cosine similarity), local intensity peaks are extracted to generate a change intensity spectrogram. The phase response module is subjected to an expansion transformation using an error model, including position offset and random error adjustment, to generate a phase anomaly weight spectrogram based on the frequency and time dimensions. The change intensity spectrogram and the phase anomaly weight spectrogram are weighted and superimposed, and the influence of the noise frequency band and the low contribution frequency band is reduced through a filter to generate an output response variation index table with weight marks.

[0080] For example, assume that the response coupling factor is a sequence: [0.5, 0.7, 0.9], which has a significant correlation with the intensity curve of the input power fluctuation spectrum in the comparison analysis. The generated change intensity spectrogram shows that the intensities of the second and third segments are 0.8 and 0.9 respectively. At the same time, an extended transformation is performed on the phase response module [30°, 45°, 60°], and the processed phase anomaly weight spectrogram is [0.2, 0.35, 0.5]. After joint filtering processing, the low-frequency components with smaller weights (such as the first segment) are eliminated, and the final response change index table is generated, marking the second and third segments as the main interference response regions.

[0081] Perform a weight fitting analysis on the response change index table to obtain the frequency band perturbation residual vector. Calculate the perturbation source weight value that best matches the frequency band perturbation residual vector, and aggregate and generate the fitting weight vector.

[0082] By analyzing the frequency band characteristics of the change points in the response change index table and performing weight fitting on each frequency band, a frequency band perturbation residual vector and a fitting weight vector are generated. Specifically, extract and match the characteristics of the main frequency bands in the response change index table, calculate the deviation between the weight value of each frequency band and the actual perturbation, and form the frequency band perturbation residual vector; based on the deviation value provided by the frequency band perturbation residual vector, find the perturbation source weight value that best matches it (for example, calculate through the least squares fitting or the residual regression model); normalize and aggregate the matching weights of each frequency band, and finally generate the fitting weight vector for correction.

[0083] For example, assume that the weight value of the second frequency band in the response change index table is 0.85, while the actually measured weight value is 0.9, and the difference is 0.05, which is recorded as a part of the frequency band perturbation residual vector; while the weight value of the third frequency band is 0.75, and the difference is 0 after matching with the actual value. Based on these data, the matching perturbation source weight values are obtained through the fitting method: [0.9, 0.75], and the normalized fitting weight vector is generated: [0.6, 0.4], indicating the contribution ratio of the weights of these two segments in the correction.

[0084] Perform a time series reduction on the response change index table to obtain the load response change trajectory group, and perform an error amplitude diffusion process on the fitting weight vector to obtain the weight perturbation error set.

[0085] The change characteristics of the response change index table are reduced in time series, and the error distribution of multiple frequency bands is calculated for the fitting weight vector, and the results are used to generate a weight perturbation error set. Specifically, the characteristics of each frequency band in the response change index table are reduced along the time axis to form a load response change trajectory group, which reflects the cumulative characteristics of the response change in different time periods; the error diffusion value of each component of the fitting weight vector in the frequency and time dimensions is calculated, and the influence of amplitude error and frequency deviation on the correction accuracy is considered, and these diffusion values are aggregated to generate a weight perturbation error set.

[0086] For example, in a certain time series, the response change index table data of the second frequency band shows that the index value gradually increases from 0.7 to 0.9. The load response change trajectory group formed after reduction represents the change trend of the frequency band from 20ms to 40ms. At the same time, the fitting weight vector shows that the error of the second frequency band has a deviation of 0.1. After diffusion processing, it ensures that the influence of the error on the adjacent frequency band is timely diffused and suppressed, and finally generates a weight perturbation error set, such as [0.02, 0.05, 0.01].

[0087] The load response change trajectory group and the weight disturbance error set are dynamically weighted fused to obtain the target correction parameter set.

[0088] By dynamically weighting and fusing the characteristics of the load response change trajectory group and the weight disturbance error set, the optimized weight value is calculated for each frequency band, and finally the target correction parameter set is generated. Specifically, the load response change trajectory group and the weight disturbance error set are standardized respectively, and the two sets of data are combined in a weighted superposition manner. The accuracy of the correction weights of each frequency band is improved through optimization calculation; combined with specific correction goals (such as phase adjustment, amplitude amplification, etc.), the weighted results are adjusted for multi-band distribution to generate the final target correction parameter set as the basis for subsequent current regulation and waveform reconstruction.

[0089] For example, for the second frequency band in the load response change trajectory group, its response weight is 0.85, combined with the error compensation value of 0.05 in the weighted disturbance error set, the final generated weighted value is 0.9; while the trajectory group weight of the third frequency band is 0.7, combined with the error compensation value of 0.1 to generate a weighted value of 0.8. Through dynamic weighted fusion, the correction contribution of each frequency band is optimized as a whole, and finally a target correction parameter set is generated, for example: [0.9, 0.8, 0.7], which is used for the dynamic adjustment of subsequent power factor correction.

[0090] In step S300, a frequency distribution vector is parsed based on the target correction parameter set. The frequency distribution vector is split into frequency bands to obtain a low-frequency subset and a high-frequency subset. A band-pass screening network is applied to extract the edge perturbation factors of the high-frequency subset, and a multiple integral filter is applied to extract the periodic offset factors of the low-frequency subset. The edge perturbation factors and the periodic offset factors are subjected to mixed feature deconstruction to obtain a coupled frequency-domain projection array and a frequency perturbation overlapping region.

[0091] By analyzing the frequency information in the target correction parameter set, a frequency distribution vector is extracted and split into frequency bands, and the signal is segmented and processed according to low frequency and high frequency. Specifically, the frequency information is extracted from the target correction parameter set, and a frequency distribution vector is constructed, which represents the power distribution of each frequency band of the input current or voltage signal. Based on the statistical analysis of the frequency distribution vector, the signal is divided into a low-frequency subset and a high-frequency subset according to a preset frequency demarcation line. The low-frequency band generally contains relatively stable frequency components, while the high-frequency band contains sharp or drastic change components. A band-pass screening network is applied to limit the frequency band of the high-frequency subset and extract the edge perturbation factors of the high-frequency band, which reflect the influence of instantaneous fluctuations in the high-frequency band. A multiple integral filter is applied to filter the low-frequency band and extract the periodic offset factors, which are mainly used to describe the overall offset trend of the signal in the low-frequency band. The extracted edge perturbation factors and periodic offset factors are subjected to feature deconstruction, and these factors are mapped to the coupled frequency-domain projection array through numerical decoupling to form a frequency perturbation overlapping region.

[0092] For example, in the frequency distribution vector of the target correction parameter set, the frequency range of the low-frequency band is 0 - 200 Hz, and the frequency range of the high-frequency band is 200 - 1000 Hz. In the high-frequency band, the edge perturbation factor extracted through band-pass filtering is 0.15 (indicating the abrupt perturbation in the high-frequency band), while in the low-frequency band, the periodic offset factor of 0.05 is extracted through multiple integral filtering (indicating the stable change in the low-frequency band). After the mixed deconstruction of these factors, a coupled frequency-domain projection array is obtained, and an intersection is generated in the frequency perturbation overlapping region, and this intersection is the part of the signal that needs special correction.

[0093] The coupled frequency-domain projection array is mapped to a preset perturbation phase diagram space to obtain a frequency slope field and a response fine-tuning domain, and the frequency perturbation overlapping region is mapped to a preset adjustment amplitude space to obtain an amplitude adjustment coefficient group and a slope interference term.

[0094] By mapping the coupled frequency-domain projection array to the preset perturbation phase diagram space, the frequency slope field and the response fine-tuning domain are calculated. At the same time, the adjustment amplitude mapping is performed on the frequency perturbation overlapping region to obtain the amplitude adjustment coefficient group and the slope interference term. Specifically, the frequency components in the coupled frequency-domain projection array are mapped to the perturbation phase diagram space according to the phase relationship, thereby obtaining the frequency slope field. The frequency slope field represents the frequency change rate of different frequency bands. Based on the frequency slope field, the response fine-tuning domain is extracted by the local maximum search method, that is, those frequency regions that have a greater impact on the signal waveform. The frequency perturbation overlapping region is mapped to the preset adjustment amplitude space, and the amplitude adjustment coefficient of each frequency band is calculated. These coefficients represent the gain or attenuation of the signals in each frequency band. According to the mapping results, the amplitude adjustment coefficient group and the slope interference term are obtained. The amplitude adjustment coefficient is used to adjust the amplitude of the signal, while the slope interference term is used to optimize the phase of the signal.

[0095] For example, a certain frequency band value in the coupled frequency-domain projection array is 0.3. After being mapped to the perturbation phase diagram space, the frequency slope of this frequency band is 0.2 / s, and the fine-tuning value of 0.1 is obtained in the response fine-tuning domain. For the frequency perturbation overlapping region, the amplitude adjustment coefficient of this region is 1.5, and the slope interference term is 0.03, indicating that the amplitude of this region needs to be increased by 1.5 times, and the phase will be fine-tuned according to the slope interference term.

[0096] Based on the frequency slope field and the amplitude adjustment coefficient group, the inter-domain factor interleaving fusion is carried out to obtain the high-frequency frequency-domain modulation factor and the low-frequency frequency-domain weighting factor. The slope interference term and the response fine-tuning domain are matched and optimized to generate the dynamic response delay vector.

[0097] By performing the inter-domain factor interleaving fusion on the frequency slope field and the amplitude adjustment coefficient group, the adjustment effect between each frequency band is optimized, and the high-frequency frequency-domain modulation factor and the low-frequency frequency-domain weighting factor are generated. Specifically, the frequency slope field and the amplitude adjustment coefficient group are used as inputs, and weighted fusion is performed in the time and frequency domains to achieve the collaborative optimization between different frequency bands. According to the fused results, the high-frequency frequency-domain modulation factor and the low-frequency frequency-domain weighting factor are calculated. The high-frequency factor mainly controls the adjustment of the fast-changing part, while the low-frequency factor is used for the adjustment of the steady signal. Further optimization is carried out according to the slope interference term and the response fine-tuning domain to ensure the matching of frequency and amplitude to reduce waveform distortion, and the dynamic response delay vector is generated.

[0098] For example, in the high-frequency band, after the factor interleaving fusion, the generated high-frequency frequency-domain modulation factor is 0.9, and the low-frequency frequency-domain weighting factor is 1.1, which means that the signal modulation in the high-frequency part will be enhanced by 90%, and the low-frequency part will be weighted and increased by 10%. After the matching optimization, the generated dynamic response delay vector is [0.1, 0.05, 0.02], and these values represent the phase delay amounts that need to be introduced in the adjustment process of different frequency bands.

[0099] Jointly weight-scale the high-frequency frequency-domain modulation factor, the low-frequency frequency-domain weighting factor, and the dynamic response delay vector to obtain a dynamic frequency adjustment matrix, and perform inverse modulation mapping on the dynamic frequency adjustment matrix and the target correction parameter set to construct a dynamic current adjustment vector generation framework.

[0100] By jointly weight-scaling the high-frequency frequency-domain modulation factor, the low-frequency frequency-domain weighting factor, and the dynamic response delay vector, a dynamic frequency adjustment matrix is generated, and inverse modulation is performed based on this matrix and the target correction parameter set to construct a dynamic current adjustment vector generation framework. Specifically, the high-frequency frequency-domain modulation factor, the low-frequency frequency-domain weighting factor, and the dynamic response delay vector are scaled proportionally to ensure that the influence of each factor on dynamic adjustment is balanced across the entire frequency spectrum. Using the scaled factors, a dynamic frequency adjustment matrix is constructed, and each element in the matrix represents the adjustment coefficient for different frequency bands. Perform inverse modulation mapping on the dynamic frequency adjustment matrix and the target correction parameter set to generate the final dynamic current adjustment vector for current correction.

[0101] For example, assume that the high-frequency frequency-domain modulation factor is 0.9, the low-frequency weighting factor is 1.1, and the dynamic response delay vector is [0.1, 0.05, 0.02]. Then, the dynamic frequency adjustment matrix obtained through weight scaling is: Then, using inverse modulation mapping, this matrix is combined with the target correction parameter set to generate the final dynamic current adjustment vector.

[0102] Perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix based on the dynamic current adjustment vector generation framework to generate a dynamic current adjustment vector.

[0103] Optimize the dynamic frequency adjustment matrix by applying vector interpolation and waveform fitting techniques to generate the final dynamic current adjustment vector. Specifically, use linear interpolation or spline interpolation methods to smooth the frequency adjustment matrix to ensure that there are no sudden changes in the transition between different frequency bands during the actual current adjustment process. Based on the interpolation results, further fit the waveform to ensure that the adjusted current waveform is close to the ideal target waveform.

[0104] For example, assume that some frequency band values in the dynamic frequency adjustment matrix are smoothly adjusted to [0.85, 1.05, 1.00] after interpolation, and then a adjusted dynamic current adjustment vector is obtained through waveform fitting method. Finally, this vector can precisely control the adjustment of the current waveform to meet the predetermined grid requirements.

[0105] In step S400, perform a periodic broadening process on the dynamic current adjustment vector, and extract a sub-cycle current vector group based on the broadening result.

[0106] By performing periodic broadening processing on the dynamic current regulation vector, it is expanded into data containing multiple periods, thereby extracting a current vector group related to each sub-period. Specifically, the dynamic current regulation vector is subjected to periodic broadening, that is, it is periodically repeated to form a continuous time series, ensuring that the time interval between each period is consistent. By cutting the broadened data, the sub-period current vector group within each period is extracted, and each vector group represents the fluctuation characteristics of the current within a complete period.

[0107] For example, assume that the dynamic current regulation vector is 1.5, 2.0, 1.7 (unit: A), representing the regulation value of the current within a certain period. After periodic broadening processing, the broadened vector obtained is 1.5, 2.0, 1.7, 1.5, 2.0, 1.7, 1.5, 2.0, 1.7. Next, it is cut into the sub-period current vector group 1.5, 2.0, 1.7, representing the current values of each sub-period, thereby providing data support for subsequent analysis.

[0108] The grid input current is sampled and segmented to obtain several current waveforms, and multi-layer wavelet packet decomposition is applied to decompose all the current waveforms to generate an amplitude perturbation reference set.

[0109] By segmenting and sampling the grid input current, multiple current waveforms are obtained, and each waveform is decomposed by multi-layer wavelet packet decomposition technology to extract the amplitude perturbation characteristics. Specifically, the grid input current is segmented and sampled, and the current signal is divided into several segments according to a set time window, and the complete waveform of each segment is obtained. Using the multi-layer wavelet packet decomposition method, each current waveform is decomposed into multiple frequency components, and the amplitude information of each frequency band is obtained through successive decomposition. The amplitude perturbation characteristics are extracted from the results of wavelet decomposition, such as the spike changes in the high-frequency part and the steady changes in the low-frequency part, to form an amplitude perturbation reference set.

[0110] For example, assume that the sampled waveform of the grid input current is 0.5, 1.2, 1.8, 2.1, 2.5. After multi-layer wavelet packet decomposition, the frequency components of this waveform are obtained: the low-frequency part is 0.5, 1.0, 1.5, and the high-frequency part is 1.2, 1.8, 2.0. The amplitude perturbation reference set consists of the amplitude change characteristics of these two frequency bands and is used to describe the fluctuation characteristics of the signal.

[0111] The sub-period current vector group and the amplitude perturbation reference set are subjected to sub-segment convolution matching to extract the perturbation fitting factor.

[0112] By convolving and matching the sub - cycle current vector group with the amplitude perturbation reference set, the perturbation fitting factors are extracted. Specifically, the sub - cycle current vector group is convolved with each element in the amplitude perturbation reference set to calculate their similarity, and the matching degree between the sub - cycle current waveform and the amplitude perturbation is identified. Based on the convolution results, the fitting factors that best conform to specific perturbation characteristics are extracted, and these factors represent the main contributing parts of the amplitude perturbation in the current waveform.

[0113] For example, if the sub - cycle current vector group is 1.5, 2.0, 1.7 and the amplitude perturbation reference set is 0.5, 1.0, 1.5, after performing the convolution operation on these two vectors, the obtained matching degree is 0.85. This value is used as the perturbation fitting factor, indicating the matching degree between the current waveform and the amplitude perturbation during this period.

[0114] Using the perturbation fitting factors, an amplitude fitting transformation matrix is constructed, and the amplitude fitting transformation matrix is applied to each segment of the grid input current for phase correction and amplitude correction to obtain the corrected phase information and the corrected amplitude information.

[0115] By constructing the amplitude fitting transformation matrix from the perturbation fitting factors and applying this matrix to each segment of the grid input current for correction. Specifically, according to the values of the perturbation fitting factors, a transformation matrix is constructed, which is used to adjust the amplitude of the current signal to make it close to the target waveform. Through this matrix, phase correction and amplitude correction are performed on each segment of the current waveform to ensure that the shape and size of the current signal meet the preset requirements and achieve the goal of power factor correction.

[0116] For example, the perturbation fitting factor extracted in the previous stage is 0.85, and this factor will be used to construct the amplitude fitting transformation matrix, such as 0.85, 1.0, 1.2. Then, this matrix is applied to each segment of the grid input current to gradually adjust the amplitude and phase of the current, ensuring that the amplitude of each segment of the current waveform is consistent with the target and fine - tuning its phase.

[0117] The corrected phase information and the corrected amplitude information are fused, and based on the fusion result, a phase - amplitude coupled envelope is constructed, and the fast Fourier reconstruction method is applied to reconstruct the waveform of the phase - amplitude coupled envelope.

[0118] By fusing the corrected phase information and amplitude information, a phase - amplitude coupled envelope is constructed, and waveform reconstruction is performed using the fast Fourier transform (FFT). Specifically, the corrected phase information and amplitude information are combined to generate a phase - amplitude coupled envelope, and the envelope reflects the variation of the phase and amplitude of the signal over time. Through the fast Fourier transform (FFT), frequency - domain reconstruction of the phase - amplitude coupled envelope is performed to obtain the corrected current waveform.

[0119] For example, assuming the corrected phase information is 30°, 45°, 60° and the amplitude information is 1.5, 2.0, 1.7, after fusing these two pieces of information, the phase-amplitude coupled envelope 1.5, 2.0, 1.7 at the phases of 30°, 45°, 60° is obtained. After reconstructing it using the Fast Fourier Transform (FFT), the generated waveform can better match the requirements of the power grid.

[0120] In this embodiment, by performing real-time sampling and normalization processing on the grid input voltage, grid input current, and output load current, the input power fluctuation map and the load dynamic response map are extracted, providing the basic data for subsequent calibration work. Then, combining the input power fluctuation map and the load dynamic response map, using map matching and convolution techniques, a target calibration parameter set is constructed. On this basis, through frequency distribution analysis and frequency band splitting, the characteristics of the low-frequency band and high-frequency band are further extracted, the frequency disturbance factor is extracted using band-pass filtering and multiple integral filtering techniques, and a coupled frequency domain projection array is generated through mixed feature deconstruction. By performing weighted fusion on these data, a frequency domain adjustment matrix is obtained, and a dynamic current adjustment vector generation framework is generated in combination with the target calibration parameter set. Then, through cycle broadening and sampling segmentation techniques, the disturbance reference set of the current waveform is extracted, and further amplitude and phase calibration of the grid input current is performed through convolution matching and fitting transformation matrix. Finally, through the construction of the phase-amplitude coupled envelope and the reconstruction using the Fast Fourier Transform (FFT), the accurate calibration of the current waveform is successfully achieved, significantly improving the power factor, reducing harmonic interference, and enhancing the stability and efficiency of the system. Through this comprehensive signal analysis and adjustment method, the optimized performance of the high-power charging module during actual operation is ensured.

[0121] Embodiment 3: As Figure 2 shown, the present application also provides a power factor correction system 10 for a high-power charging module, including an acquisition module 11, a construction module 12, a fitting module 13, and a calibration module 14.

[0122] The acquisition module 11 is mainly used to acquire the grid input voltage, grid input current, and output load current, construct an input power fluctuation map based on the grid input voltage and the grid input current, and construct a load dynamic response map based on the output load current.

[0123] The construction module 12 is mainly used to perform sub-map matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and construct a target calibration parameter set using the response change index table and the fitting weight vector.

[0124] The fitting module 13 is mainly used to perform weight scaling and mapping on the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and use the dynamic current adjustment vector generation framework to perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix to generate a dynamic current adjustment vector.

[0125] The correction module 14 is mainly used to perform phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector to achieve power factor correction of the high-power charging module.

[0126] In this embodiment, by dividing the power factor correction process of the high-power charging module into four modules: acquisition, construction, fitting, and correction, the accurate correction of the input grid current and the load current is effectively achieved. First, the acquisition module 11 collects the grid input voltage, grid input current, and output load current in real time, and uses these data to construct an input power fluctuation map and a load dynamic response map, providing key basic information for subsequent correction. Then, the construction module 12 deeply fuses the input power fluctuation map and the load dynamic response map by means of sub-map matching and joint convolution, extracts a response change index table and a fitting weight vector from them, and then constructs a target correction parameter set. Next, the fitting module 13 performs weight scaling and mapping on the target correction parameter set to generate a dynamic frequency adjustment matrix, and performs vector interpolation and waveform fitting on the dynamic current adjustment vector based on this matrix to generate an accurate dynamic current adjustment vector. Finally, the correction module 14 uses this dynamic current adjustment vector to perform phase correction and waveform reconstruction on the grid input current to ensure that the current waveform highly matches the grid requirements, significantly improving the power factor. Through these modular steps, this embodiment realizes refined current waveform adjustment, not only optimizing the power factor, but also effectively reducing harmonic interference and improving the overall operation efficiency of the grid and the high-power charging module.

[0127] It should be noted that those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described system and each module can refer to the corresponding processes in the foregoing Embodiment 1, and will not be elaborated herein.

[0128] The structures, ratios, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those skilled in this technology to understand and read, and are not used to limit the limited conditions under which the present invention can be implemented. Therefore, they do not have technical essence. Any modification of the structure, change of the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that the technical content disclosed in the present invention can cover.

[0129] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power factor correction method for a high-power charging module, characterized in that, Including: Obtain the grid input voltage, grid input current, and output load current, construct an input power fluctuation map based on the grid input voltage and the grid input current, and construct a load dynamic response map based on the output load current; Perform sub-graph matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and construct a target correction parameter set using the response change index table and the fitting weight vector; Perform weight scaling and mapping on the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix using the dynamic current adjustment vector generation framework to generate a dynamic current adjustment vector; Perform phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector to achieve power factor correction of the high-power charging module.

2. The power factor correction method of the high-power charging module according to claim 1, characterized in that The steps of obtaining the grid input voltage, grid input current, and output load current, constructing an input power fluctuation map based on the grid input voltage and the grid input current, and constructing a load dynamic response map based on the output load current include: Obtain the grid input voltage, grid input current, and output load current within a preset continuous period, perform normalization processing on the grid input voltage and the grid input current according to the phase period to obtain a normalized input waveform sequence and corresponding period synchronization timestamps; Construct a voltage-current phase cross sequence and a voltage transient slope sequence based on the normalized input waveform sequence, perform feature co-occurrence mapping processing on the voltage-current phase cross sequence and the voltage transient slope sequence to obtain a cross-phase map and a slope change map; Perform amplitude difference structure superposition processing on the cross-phase map and the slope change map to obtain a voltage disturbance fusion map, and construct an input power fluctuation map using the voltage disturbance fusion map and the period synchronization timestamps; Calculate the load change rate using the output load current, perform section analysis on the load change rate to obtain load section characteristics and a load continuous change sequence, and construct a load dynamic response map according to the load section characteristics and the load continuous change sequence.

3. The power factor correction method of the high-power charging module according to claim 2, characterized in that The steps of performing section analysis on the load change rate to obtain a section statistical feature set and a load continuous change sequence, and constructing a load dynamic response map according to the section statistical feature set and the load continuous change sequence include: Perform first-order difference processing and local extreme value analysis on the load change rate to obtain an initial change rate vector and a mutation boundary identification sequence, perform boundary segment segmentation on the initial change rate vector based on the mutation boundary identification sequence to obtain several load change sections, and calculate the maximum change slope, average fluctuation range, and duration of each section in all the load change sections to generate a section statistical feature set; Perform time normalization on the statistical feature set of the section, construct a feature interpolation curve based on the processed statistical feature set of the section, fit the continuously changing region in the feature interpolation curve to generate a load continuous change sequence; Perform cross-fusion on the load continuous change sequence and the statistical feature set of the section to construct a hybrid response primitive, construct a time series tensor mapping based on the hybrid response primitive, and use the time series tensor mapping to perform spatio-temporal coupling encoding on the load change behavior at different time points to obtain a change trend flux map; Align the change trend flux map with a preset periodic synchronization timestamp to generate a time dimension calibration result, index-bind the time dimension calibration result with the hybrid response primitive to construct a graph node and edge weight relationship, and construct a load dynamic response graph based on the graph node and the edge weight relationship.

4. The power factor correction method of the high-power charging module according to claim 1, characterized in that The step of performing subgraph matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and constructing a target correction parameter set using the response change index table and the fitting weight vector includes: Obtain the amplitude interval sequence and phase shift sequence of multiple consecutive periods in the input power fluctuation map, and classify the spectral features of the amplitude interval sequence and the phase shift sequence to obtain an amplitude structure group and a phase shift mode group; Obtain the load change segment feature set in the load dynamic response map corresponding to the continuous periods in the input power fluctuation map, perform subgraph matching on the load change segment feature set and the phase shift mode group to obtain a load interference influence domain and an input perturbation sensitive domain; Perform multi-dimensional matrix joint convolution on the amplitude structure group and the load interference influence domain to obtain a perturbation co-coupling matrix, and perform section cross-filtering on the perturbation co-coupling matrix and the input perturbation sensitive domain to generate a response change index table and a fitting weight vector; Perform time series reduction on the response change index table to obtain a load response change trajectory group, and perform error amplitude diffusion processing on the fitting weight vector to obtain a weight perturbation error set; Perform dynamic weighted fusion on the load response change trajectory group and the weight perturbation error set to obtain a target correction parameter set.

5. The power factor correction method of the high-power charging module according to claim 4, characterized in that, The step of performing multi-dimensional matrix joint convolution on the amplitude structure group and the load interference influence domain to obtain a perturbation co-coupling matrix, and performing section cross-filtering on the perturbation co-coupling matrix and the input perturbation sensitive domain to generate a response change index table and a fitting weight vector includes: Construct a periodic amplitude vector set based on the amplitude structure group, construct a time-synchronized perturbation field based on the load interference influence domain, expand the periodic amplitude vector set and the time-synchronized perturbation field in the vector domain to obtain an amplitude perturbation tensor and an interference density tensor; Perform kernel convolution processing on the amplitude perturbation tensor and the interference density tensor to obtain a periodic perturbation induction matrix, and map the input perturbation sensitive domain based on the periodic perturbation induction matrix to obtain a response coupling factor and a phase response module; Perform trend comparison mapping on the response coupling factor, generate a change intensity spectrogram based on the mapping result, perform error expansion transformation on the phase response module to obtain a phase anomaly weight spectrogram, and perform joint filtering processing on the change intensity spectrogram and the phase anomaly weight spectrogram to output a response change index table; Perform weight fitting analysis on the response change index table to obtain a frequency band perturbation residual vector, calculate the perturbation source weight value that best matches the frequency band perturbation residual vector, and aggregate and generate a fitting weight vector.

6. The power factor correction method of the high-power charging module according to claim 1, characterized in that, The step of performing weight scaling and mapping on the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and using the dynamic current adjustment vector generation framework to perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix to generate a dynamic current adjustment vector includes: Based on the target correction parameter set, parse out a frequency distribution vector, perform frequency band splitting on the frequency distribution vector to obtain a low-frequency subset and a high-frequency subset, apply a band-pass screening network to extract the edge perturbation factor of the high-frequency subset, apply a multiple integral filter to extract the periodic offset factor of the low-frequency subset, and perform mixed feature deconstruction on the edge perturbation factor and the periodic offset factor to obtain a coupled frequency domain projection array and a frequency perturbation overlapping region; Map the coupled frequency domain projection array to a preset perturbation phase diagram space to obtain a frequency slope field and a response fine-tuning domain, and map the frequency perturbation overlapping region to a preset adjustment amplitude space to obtain an amplitude adjustment coefficient group and a slope interference term; Perform inter-domain factor interleaving fusion based on the frequency slope field and the amplitude adjustment coefficient group to obtain a high-frequency frequency domain modulation factor and a low-frequency frequency domain weighting factor, and perform matching optimization on the slope interference term and the response fine-tuning domain to generate a dynamic response delay vector; Perform joint weight scaling on the high-frequency frequency domain modulation factor, the low-frequency frequency domain weighting factor, and the dynamic response delay vector to obtain a dynamic frequency adjustment matrix, and perform inverse modulation mapping on the dynamic frequency adjustment matrix and the target correction parameter set to construct a dynamic current adjustment vector generation framework; Based on the dynamic current adjustment vector generation framework, perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix to generate a dynamic current adjustment vector.

7. The power factor correction method of the high-power charging module according to claim 1, characterized in that, The step of performing phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector includes: Perform period broadening processing on the dynamic current adjustment vector, and extract a sub-period current vector group based on the broadening result; Perform sampling segmentation on the grid input current to obtain a plurality of current waveforms, and apply multi-layer wavelet packet decomposition to decompose all the current waveforms to generate an amplitude perturbation reference set; Perform sub-segment convolution matching on the sub-period current vector group and the amplitude perturbation reference set to extract a perturbation fitting factor; Use the perturbation fitting factor to construct an amplitude fitting conversion matrix, and apply the amplitude fitting conversion matrix to perform phase correction and amplitude correction on each segment of the grid input current to obtain the corrected phase information and the corrected amplitude information; Fuse the corrected phase information and the corrected amplitude information, construct a phase-amplitude coupling envelope based on the fusion result, and apply the fast Fourier reconstruction method to reconstruct the waveform of the phase-amplitude coupling envelope.

8. A power factor correction system for a high-power charging module, characterized in that, It includes: An acquisition module, configured to acquire the grid input voltage, the grid input current, and the output load current, construct an input power fluctuation map based on the grid input voltage and the grid input current, and construct a load dynamic response map based on the output load current; A construction module, configured to perform subgraph matching and joint convolution on the input power fluctuation map and the load dynamic response map to obtain a response change index table and a fitting weight vector, and construct a target correction parameter set by using the response change index table and the fitting weight vector; A fitting module, configured to perform weight scaling and mapping on the target correction parameter set to obtain a dynamic frequency adjustment matrix and a dynamic current adjustment vector generation framework, and perform vector interpolation and waveform fitting on the dynamic frequency adjustment matrix by using the dynamic current adjustment vector generation framework to generate a dynamic current adjustment vector; A correction module, configured to perform phase correction and waveform reconstruction on the grid input current according to the dynamic current adjustment vector to achieve power factor correction of the high-power charging module.

Images