High-speed power conversion control method

Abstract

The invention discloses a high-speed power supply conversion control method, which relates to the technical field of power supply conversion control, and comprises the following steps of: synchronously acquiring transient voltage waveforms and transient current waveforms at two ends of a power device in each pulse width modulation period through an analog-to-digital conversion module; and constructing a high-frequency waveform data frame containing a plurality of modulation periods based on a fixed time sliding window mechanism. According to the invention, through frequency domain monitoring and frequency spectrum characteristic quantitative analysis, a resonance response index is output, and a closed-loop mechanism from high-frequency waveform sensing to risk early warning is realized; the PWM dominant frequency and harmonic distribution are dynamically adjusted in combination with a digital controller, a frequency modulation avoidance and harmonic suppression cooperative control strategy is constructed, the adaptive capacity, stability and electromagnetic compatibility of the system when dealing with parasitic resonance interference are remarkably improved, and the method has outstanding engineering practicability and technical innovation.

Description

Technical Field

[0001] The present invention relates to the technical field of power conversion control, and in particular to a high-speed power conversion control method. Background Art

[0002] High-speed power conversion control refers to a control technology that achieves efficient and stable conversion of input power to output power in a power supply system through a high-frequency, fast-response control strategy. This control method typically uses high-switching frequency power electronic devices (such as MOSFETs, IGBTs, etc.) and high-speed digital controllers (such as DSPs, FPGAs, etc.), combined with refined feedback control algorithms (such as current loop, voltage loop control, or predictive control). With dynamic response capabilities in milliseconds or even microseconds, it can quickly adapt to load changes and input disturbances, ensuring excellent steady-state and transient performance of the output voltage or current. High-speed control is particularly suitable for application scenarios that require fast energy conversion speed, high stability and accuracy, and sensitive dynamic response, such as server power supplies, automotive power systems, industrial high-frequency power supplies, and new energy power conversion equipment, significantly improving the robustness, efficiency, and control accuracy of the power supply system.

[0003] The existing technology has the following shortcomings: In existing high-speed power conversion systems, high-frequency PWM modulation is commonly used to improve power conversion efficiency and control response speed. However, in certain power supply topologies, due to the inevitable parasitic parameters within the system (including parasitic inductance and capacitance in PCB wiring, distributed capacitance in magnetic rings or power device winding structures, etc.), high-frequency modulation may form hidden closed-loop resonant paths. When the PWM modulation signal frequency or its higher harmonic frequencies approach the natural frequency of this parasitic resonant circuit, it will trigger a nonlinear resonant response, which in turn leads to abnormal localized high-frequency energy accumulation and a phenomenon called "resonance penetration." This phenomenon can cause high-frequency energy to break through the limitations of the system's original filtering structure and directly couple into the load side or feedback control loop, causing output voltage fluctuations, control misjudgment, abnormally enhanced EMI, and even serious consequences such as mis-conduction of power devices, loop oscillation, and device functional degradation.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a high-speed power conversion control method, which outputs a resonance response index through frequency domain monitoring and quantitative analysis of spectrum characteristics, thereby realizing a closed-loop mechanism from high-frequency waveform perception to risk warning; combined with a digital controller to dynamically adjust the PWM main frequency and harmonic distribution, a frequency modulation avoidance and harmonic suppression collaborative control strategy is constructed, which significantly improves the system's adaptability, stability and electromagnetic compatibility when dealing with parasitic resonance interference, and has outstanding engineering practicality and technological innovation to solve the problems in the above-mentioned background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a high-speed power conversion control method, comprising the following steps:

[0007] Through the analog-to-digital conversion module, the transient voltage waveform and transient current waveform at both ends of the power device in each pulse width modulation cycle are synchronously collected, and a high-frequency waveform data frame containing multiple modulation cycles is constructed based on a fixed time sliding window mechanism;

[0008] Perform fast Fourier transform on high-frequency waveform data frames to extract the spectral amplitude distribution characteristics of the main frequency component and multiple high-order harmonic frequency components in the frequency domain, construct an energy spectrum density curve corresponding to the frequency, identify the peak frequency segments with abnormally amplified amplitudes in the energy spectrum density curve, extract the frequency value and corresponding amplitude of each peak frequency segment, and form a spectrum response feature vector;

[0009] A frequency matching operation is performed on the peak frequency segment in each spectral response eigenvector. The peak frequency value is compared item by item with the pre-modeled parasitic resonance frequency list. The frequency offset and frequency amplitude magnification are calculated. The offset and magnification are combined to calculate the frequency proximity. The frequency coupling factor is introduced to construct the resonance coupling strength score, which quantifies the energy excitation intensity between the peak frequency segment and the parasitic resonance path.

[0010] The frequency offset, frequency amplitude magnification and resonance coupling strength score are input into a nonlinear mapping function, the input parameters are processed by a normalization function, an excitation function is introduced to amplify and map the high-order response characteristics in the critical frequency range, and the output is a nonlinear resonance response index limited to between 0 and 1;

[0011] When the nonlinear resonance response index is greater than the reference threshold, the main switching frequency of the pulse width modulation and its higher harmonic distribution are dynamically adjusted. The modulation frequency is controlled by a digital controller to drift to a frequency band that does not overlap with the parasitic resonance frequency, avoiding the resonance excitation frequency point and suppressing the high-frequency energy superposition effect.

[0012] Preferably, the specific steps of constructing a high-frequency waveform data frame containing multiple modulation cycles are as follows:

[0013] Based on the pulse width modulation cycle clock signal, the analog-to-digital conversion module is synchronously triggered so that the sampling starting points of the voltage sampling channel and the current sampling channel are strictly aligned within each modulation cycle, ensuring that the sampling starting points of the voltage waveform and the current waveform on the time axis are consistent;

[0014] Within each modulation cycle, multiple equally spaced sampling points are evenly divided to fully cover the rising edge stage, stable conduction stage, falling edge stage and complete shutdown stage of the power device conduction process, and construct single-cycle transient voltage waveform units and transient current waveform units with structural consistency;

[0015] A plurality of continuous single-cycle waveform units are sequentially spliced ​​according to a fixed sliding interval to generate a high-frequency waveform data frame containing a plurality of modulation cycles, and the data frame is cached in a high-speed cache.

[0016] Preferably, the steps for generating the spectrum response feature vector are as follows:

[0017] Perform fast Fourier transform operations on high-frequency waveform data frames to obtain spectrum amplitude data within the complete frequency range, and extract the amplitude distribution information of the main frequency component and multiple high-order harmonic frequency components in the frequency domain;

[0018] The frequency components and corresponding amplitude data points are mapped into a one-dimensional energy spectrum density curve in ascending order of frequency. The peak response area with abnormally amplified amplitude is identified in the curve. The central frequency value and maximum amplitude are obtained by calculating the local maximum value of each peak area to form a peak frequency segment with response characteristics.

[0019] All peak frequency segments are arranged in order according to the frequency dimension, and the center frequency value, maximum amplitude and relative position of each peak frequency segment in the entire spectrum are combined to construct a three-dimensional vector structure to generate a spectrum response feature vector.

[0020] Preferably, a fast Fourier transform operation is performed on the high-frequency waveform data frame to obtain spectrum amplitude data within the complete frequency range, and the amplitude distribution information of the main frequency component and multiple high-order harmonic frequency components in the frequency domain is extracted. The specific steps are as follows:

[0021] Based on the constructed high-frequency waveform data frame, the input parameters of the fast Fourier transform are initialized, and the transient voltage value and transient current value of each sampling point are respectively sorted into two groups of real time domain signal sequences of equal length;

[0022] The two sets of time domain signal sequences were subjected to window function weighting processing respectively, and the Blackman window function was used to smooth the transition of the signal boundary and reduce the spectrum leakage effect caused by edge discontinuity.

[0023] Call the high-precision fast Fourier transform operation module to convert the weighted real signal sequence into complex spectrum data arranged in ascending frequency, calculate the amplitude corresponding to each frequency point, and output a set of frequency domain amplitude arrays;

[0024] The frequency points where the main frequency components are located and their corresponding energy amplitudes are identified from the frequency domain amplitude array, and multiple physically meaningful higher harmonic frequency points and their amplitudes are located to construct a complete frequency domain amplitude distribution set.

[0025] Preferably, the frequency matching operation is performed on the peak frequency segment in each spectral response characteristic vector to construct the resonance coupling strength score. The specific steps are as follows:

[0026] For each peak frequency segment in the spectral response eigenvector, extract its center frequency value and corresponding amplitude, perform frequency matching comparison with each natural frequency value in the pre-modeled parasitic resonance frequency list, and calculate the frequency offset between each peak frequency value and the corresponding natural frequency value;

[0027] Combine the amplitude of the peak frequency segment with the average amplitude level of the frequency region where it is located to calculate the corresponding frequency amplitude magnification factor. Combine the frequency offset and the frequency amplitude magnification factor as input factors and calculate the frequency proximity index through ratio normalization.

[0028] The frequency proximity index is weighted and fused with the pre-set frequency coupling factor to construct a resonance coupling strength score that reflects the energy coupling relationship between the peak frequency segment and the parasitic resonance path.

[0029] Preferably, a normalization function is used to process the input parameters, an excitation function is introduced to amplify and map the high-order response characteristics in the critical frequency range, and the specific steps of outputting the nonlinear resonance response index are as follows:

[0030] The frequency offset, frequency amplitude magnification, and resonance coupling strength score corresponding to each peak frequency band are used as three input parameters and uniformly input into the nonlinear mapping function model, and each input parameter is normalized.

[0031] Based on the normalized input, a multivariate combination relationship is constructed through a mapping function to identify input samples in the critical frequency range. For parameter combinations in highly sensitive areas, an excitation function is introduced to perform nonlinear amplification processing, so that the excitation signals in the critical frequency band receive higher weight in the calculation results.

[0032] The results after nonlinear combination and excitation transformation are reduced and mapped to form a nonlinear resonance response index that reflects the excitation risk level of the peak frequency band.

[0033] Preferably, when the nonlinear resonance response index is greater than the reference threshold, spectrum dynamic control is performed, and a frequency adjustment offset factor is constructed based on the nonlinear resonance response index and the nonlinear resonance response index reference threshold. The frequency adjustment offset factor determines the adjustment amplitude and direction of the pulse width modulation main frequency, and the calculation expression is as follows:

[0034]

[0035] , where Δf a is the frequency adjustment offset factor, α is the frequency adjustment sensitivity coefficient, β is the nonlinear amplification order, χ r is the nonlinear resonant response index, χ th is the reference threshold of the nonlinear resonance response index, (χ r -χ th ) + is the positive difference after the index exceeds the threshold;

[0036] Adjust the offset factor Δf based on the generated frequency a , dynamically reconstruct the main switching frequency of pulse width modulation and its harmonic energy distribution, the calculation expression is as follows:

[0037]

[0038] , where f new is the adjusted main switching frequency, f base is the original set PWM main switching frequency, γ(t) is the frequency drift direction factor based on time change, H new (n) is the energy amplitude corresponding to the adjusted nth-order higher harmonic frequency, H base (n) is the original amplitude of the nth-order higher harmonic frequency before adjustment, e is the natural base, and λ is the harmonic suppression coefficient.

[0039] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0040] By performing fine-grained monitoring of high-frequency energy accumulation in the frequency domain and utilizing the spectrum response eigenvector combined with a nonlinear mapping function to output a resonant response index, a complete quantitative mechanism from waveform perception to risk warning has been established. Furthermore, a digital controller dynamically adjusts the PWM main frequency and its harmonic energy distribution to achieve coordinated control of frequency modulation avoidance and harmonic suppression. This solution not only improves the system's adaptive regulation capability and robustness in the face of frequency-domain resonant interference, but also significantly enhances the stability, electromagnetic compatibility, and control accuracy of the power conversion process, demonstrating significant engineering practicality and innovation. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0042] Figure 1 This is a flow chart of the high-speed power conversion control method of the present invention. DETAILED DESCRIPTION

[0043] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0044] The present invention provides Figure 1 The high-speed power conversion control method shown includes the following steps:

[0045] Through a high-sampling-rate analog-to-digital conversion module, the transient voltage and current waveforms at both ends of the power device are synchronously acquired within each pulse-width modulation cycle. A high-frequency waveform data frame containing multiple modulation cycles is constructed based on a fixed-time sliding window mechanism to capture the transient energy variation characteristics generated by pulse-width modulation.

[0046] The specific steps for constructing a high-frequency waveform data frame containing multiple modulation cycles are as follows:

[0047] Based on the pulse width modulation cycle clock signal, the analog-to-digital conversion module is synchronously triggered, so that the sampling starting points of the voltage sampling channel and the current sampling channel are strictly aligned within each modulation cycle, ensuring that the sampling starting points of the voltage waveform and the current waveform on the time axis are consistent, thereby improving the timing symmetry of the dual-channel sampling data;

[0048] The purpose of synchronously triggering the analog-to-digital conversion module based on the pulse width modulation cycle clock signal is to ensure that the voltage sampling channel and the current sampling channel strictly align the sampling start point within each modulation cycle, thereby ensuring the synchronization consistency of the two sampling data on the time axis, so that subsequent waveform analysis can accurately reflect the dynamic response relationship between voltage and current. This synchronous triggering mechanism uses the pulse width modulation cycle clock as the global synchronization reference and configures the trigger control logic of the analog-to-digital conversion module to start the voltage and current sampling process at the same time at the beginning of each modulation cycle, avoiding phase errors or data distortion problems caused by sampling start point offsets. In this way, the integrity of the sampled data in the timing dimension can be significantly improved, providing high-precision, low-error basic data support for subsequent spectrum transformation, resonance identification, and dynamic control.

[0049] Within each modulation cycle, multiple equally spaced sampling points are evenly divided to fully cover the rising edge stage, stable conduction stage, falling edge stage and complete shutdown stage of the power device conduction process, and construct single-cycle transient voltage waveform units and transient current waveform units with structural consistency;

[0050] The method for evenly dividing multiple equally spaced sampling points within each modulation cycle is to first divide the entire modulation cycle into several time segments based on the duration of the pulse width modulation cycle, with each time segment having equal intervals, and determine the number and distribution density of the sampling points according to the set sampling accuracy. Subsequently, the internal timer or sampling clock of the analog-to-digital conversion module is configured to trigger a voltage and current sampling operation at each equally spaced time point in sequence, thereby obtaining continuous waveform data covering the entire conduction process within a complete modulation cycle. In this way, not only can the dynamic behavior characteristics of the power device at the rising edge, stable conduction, falling edge, and complete shutdown stages be fully captured, but also the key conduction nodes can be accurately identified in subsequent waveform analysis, providing high-resolution raw data support for energy change trend modeling and frequency domain feature extraction.

[0051] Multiple continuous single-cycle waveform units are sequentially spliced ​​at a fixed sliding interval to generate high-frequency waveform data frames containing multiple modulation cycles. The data frames are then cached in a high-speed cache and used as input data sets for subsequent spectrum conversion and feature extraction operations, enhancing the temporal continuity and frequency domain identification accuracy of transient energy change characteristics.

[0052] A high-sampling-rate analog-to-digital conversion module synchronously acquires the transient voltage and current waveforms across the power device during each pulse-width modulation cycle. A fixed-time sliding window mechanism is used to construct a high-frequency waveform data frame encompassing multiple modulation cycles. This provides accurate, continuous, and structured raw data with high temporal resolution for subsequent frequency-domain analysis and resonant feature extraction. During high-speed power conversion control, power devices switch extremely rapidly, resulting in strong transient characteristics and nonlinear dynamic responses in the voltage and current waveforms. Low-speed or single-point sampling alone cannot fully capture the changing trends of these critical signals. Synchronously acquiring the complete waveform within each pulse-width modulation cycle ensures that the voltage and current are observed on the same time scale, thereby reconstructing the true power transfer process and device conduction behavior. Furthermore, a fixed-time sliding window mechanism is used to concatenate waveform data from multiple consecutive modulation cycles into a consistent waveform data frame. This preserves the continuity and trend of the signal over time, providing a stable data foundation for operations such as fast Fourier transforms and energy spectral density extraction, avoiding analysis distortion or response offset. In addition, continuous data frame construction also helps to identify local resonance events, spike responses and high-frequency disturbances under non-steady-state operating conditions, thereby improving the ability to predict the "resonance penetration" phenomenon and the accuracy of dynamic control. It is an important basic step in building a highly robust, high-resolution and high-real-time power control system.

[0053] Perform fast Fourier transform on high-frequency waveform data frames to extract the spectral amplitude distribution characteristics of the main frequency component and multiple high-order harmonic frequency components in the frequency domain, construct an energy spectrum density curve corresponding to the frequency, identify the peak frequency segments with abnormally amplified amplitudes in the energy spectrum density curve, extract the frequency value and corresponding amplitude of each peak frequency segment, and form a spectrum response feature vector;

[0054] The steps for generating the spectral response feature vector are as follows:

[0055] Perform fast Fourier transform operations on high-frequency waveform data frames to obtain spectrum amplitude data within the complete frequency range, extract the amplitude distribution information of the main frequency component and multiple high-order harmonic frequency components in the frequency domain, and use it to reflect the distribution intensity of high-frequency energy in each frequency dimension;

[0056] The frequency components and corresponding amplitude data points are mapped into a one-dimensional energy spectrum density curve in ascending order of frequency. The peak response area with abnormally amplified amplitude is identified in the curve. The central frequency value and maximum amplitude are obtained by calculating the local maximum value of each peak area to form a peak frequency segment with response characteristics.

[0057] All peak frequency segments are arranged in order according to the frequency dimension, and the center frequency value, maximum amplitude and relative position of each peak frequency segment in the entire spectrum are combined to construct a three-dimensional vector structure to generate a spectrum response feature vector, which is used as the basic data input for subsequent resonant path frequency matching and high-frequency energy aggregation identification.

[0058] A fast Fourier transform (FFT) is performed on high-frequency waveform data frames, further extracting the spectral amplitude distribution characteristics of the main frequency component and multiple higher harmonic frequency components in the frequency domain. Its core function is to convert complex transient waveform signals, originally in the time domain, into easily identifiable and analyzable frequency-domain energy distribution information, thereby revealing possible frequency response anomalies in the system under high-frequency modulation conditions. By constructing frequency-dependent energy spectral density curves, the energy concentration trends of power devices at each frequency point can be clearly observed, further revealing abnormal peak responses in certain frequency regions. In power supply systems, hidden resonant paths formed by certain parasitic parameters are often difficult to directly identify in the time domain, but their excitation responses in the frequency domain often manifest as peak frequency segments with significant amplitude amplification. By identifying these peak frequency segments and extracting their corresponding frequency and amplitude information, it is possible to effectively determine whether the system is at risk of resonant excitation. These spectral characteristics are not only key to assessing the strength of resonant coupling and predicting the risk of resonant penetration, but also provide the foundational data for the subsequent construction of spectral response feature vectors, further supporting the implementation of dynamic frequency modulation avoidance mechanisms. Therefore, this step not only realizes the effective conversion of data dimensions, but also builds a bridge between time domain observation and frequency domain identification. It is the prerequisite for realizing high-resolution, high-accuracy and high-real-time control strategies.

[0059] Frequency components and corresponding amplitude data points are mapped in ascending order of frequency into a one-dimensional energy spectral density curve, and peak response regions with abnormally amplified amplitudes are identified within this curve. The specific process includes four key steps: data sorting, function mapping, curve construction, and peak identification. First, each frequency point and its corresponding amplitude data are extracted from the spectrum amplitude array obtained by fast Fourier transform. These data points are then constructed into a two-dimensional coordinate set of frequency-amplitude pairs. Next, this set is sorted in ascending order of frequency value, ensuring monotonicity along the horizontal axis, facilitating subsequent data fitting and anomaly detection. The sorted frequency-amplitude pairs are then mapped onto a one-dimensional energy spectral density curve, plotting the energy distribution trajectory across the entire frequency domain in a continuous coordinate system with frequency as the horizontal axis and amplitude as the vertical axis. In this process, frequency, as the independent variable, locates the signal frequency band, while amplitude, as the dependent variable, reflects the changing trend of the signal energy intensity. After the curve is drawn, it is scanned using derivative rate analysis, sliding window analysis, or dynamic thresholding to identify frequency bands where the amplitude increase significantly exceeds the neighborhood average within a local area, marking these regions as peak response areas. This process structures and continuates discrete spectral data, accurately locating high-energy concentration points in the system's frequency domain and providing clear, quantifiable input for subsequent frequency matching, resonance intensity scoring, and frequency modulation control.

[0060] Perform a fast Fourier transform operation on the high-frequency waveform data frame to obtain the spectrum amplitude data within the complete frequency range, and extract the amplitude distribution information of the main frequency component and multiple high-order harmonic frequency components in the frequency domain. The specific steps are as follows:

[0061] Based on the constructed high-frequency waveform data frame, the input parameters of the fast Fourier transform are initialized, and the transient voltage and transient current values ​​of each sampling point are sorted into two sets of real time domain signal sequences of equal length, which serve as the original input data for the frequency domain conversion operation;

[0062] Window function weighted processing is performed on the two sets of time domain signal sequences obtained respectively. The Blackman window function is used to smooth the transition of the signal boundary, reduce the spectrum leakage effect caused by edge discontinuity, ensure the continuity of signal energy distribution during the frequency domain conversion process, and improve the accuracy of amplitude recognition.

[0063] Call the high-precision fast Fourier transform operation module to convert the weighted real signal sequence into complex spectrum data arranged in ascending frequency, calculate the amplitude corresponding to each frequency point, and output a set of frequency domain amplitude arrays, where each array element represents the signal energy intensity at a specific frequency point;

[0064] Identify the frequency points of the main frequency components and their corresponding energy amplitudes from the frequency domain amplitude array, locate multiple physically significant higher harmonic frequency points and their amplitudes, and construct a complete frequency domain amplitude distribution set, providing basic data support for subsequent energy spectrum density curve construction and peak frequency segment extraction.

[0065] Window function weighting is performed on each of the two acquired time-domain signal sequences, and a Blackman window function is used to smooth the signal boundaries. The core purpose is to suppress spectral leakage and improve the accuracy of frequency-domain analysis. In the Fast Fourier Transform (FFT), the actual acquired time-domain signal is typically a truncated sequence of finite length. Since the signal often exhibits mutations or discontinuities at the truncation points, direct transformation can lead to the diffusion of unrealistic frequency components in the spectrum, known as spectral leakage, which obscures or interferes with the identification of true frequency features. To alleviate this problem, a window function is introduced before the frequency-domain transformation to weight the time-domain signal so that it gradually converges to zero at the boundaries, thereby reducing frequency interference caused by mutations at the edges. The Blackman window function, with its superior edge smoothness and strong sidelobe suppression, can effectively suppress the leakage of higher-order frequency components while reducing distortion of the main frequency component, making it an ideal choice for processing highly dynamic transient signals. Through this processing, not only the resolution and signal-to-noise ratio of the spectrum amplitude distribution are improved, but also the accuracy and robustness of subsequent resonance peak identification, spectrum feature extraction and other links are significantly enhanced.

[0066] A fast Fourier transform (FFT) operation is performed on high-frequency waveform data frames to obtain spectrum amplitude data within the complete frequency range. The amplitude distribution information of the main frequency component and multiple higher harmonic frequency components in the frequency domain is extracted. Its main function is to achieve accurate conversion from time domain signals to frequency domain information, thereby revealing the frequency response characteristics and energy distribution structure of the power supply system during high-frequency modulation. In high-speed power conversion control, the transient voltage and current waveforms at both ends of the power device usually contain a large number of high-frequency components and complex nonlinear characteristics. These characteristics appear as spikes, mutations, or periodic perturbations in the time domain, making them difficult to accurately quantify using time domain methods. However, through the fast Fourier transform, these transient time domain signals can be mapped to the frequency domain, making the amplitude of each frequency component a measurable energy indicator, thereby clearly displaying the main frequency and harmonic structure contained in the signal in the frequency domain. In particular, the main frequency reflects the normal operating frequency of the system, while higher harmonics may be related to switching devices, electromagnetic interference or parasitic effects. Therefore, complete acquisition of frequency domain amplitude distribution information not only helps to determine whether the system is operating within the expected frequency range, but also provides a key basis for identifying possible resonant excitation points, abnormal frequency concentration areas and interference sources. In addition, this frequency domain data is also the basic input for constructing energy spectrum density curves, extracting peak frequency segments, performing resonant path matching and controller frequency modulation response, so that the entire control process has calculable and traceable spectrum support. Therefore, this step is not only the starting point of spectrum analysis, but also the prerequisite for the entire high-speed control system to achieve dynamic perception and precise adjustment capabilities.

[0067] A frequency matching operation is performed on the peak frequency segment in each spectral response eigenvector. The peak frequency value is compared item by item with the pre-modeled parasitic resonance frequency list. The frequency offset and frequency amplitude magnification are calculated. The offset and magnification are combined to calculate the frequency proximity. The frequency coupling factor is introduced to construct the resonance coupling strength score, which quantifies the energy excitation intensity between the peak frequency segment and the parasitic resonance path.

[0068] Perform frequency matching on the peak frequency segments in each spectral response eigenvector to construct a resonant coupling strength score. The specific steps are as follows:

[0069] For each peak frequency segment in the spectral response eigenvector, extract its center frequency value and corresponding amplitude, perform frequency matching comparison with each natural frequency value in the pre-modeled parasitic resonance frequency list, and calculate the frequency offset between each peak frequency value and the corresponding natural frequency value;

[0070] The above steps are used to identify whether the peak frequency segment overlaps with potential parasitic resonant frequencies in the system. By calculating the frequency offset, a preliminary judgment is made as to whether there is a possibility of resonant excitation.

[0071] For each peak frequency segment in the spectral response characteristic vector, the process of extracting its center frequency value and corresponding amplitude includes three steps: local range scanning, peak location, and value extraction. First, in the constructed one-dimensional energy spectrum density curve, all frequency points and corresponding amplitude data in the frequency segment are extracted according to the identified peak frequency segment range to form a local spectrum data set; then, an amplitude comparison operation is performed on the local data set to find the frequency point where the amplitude reaches the maximum value, and the frequency value of this point is used as the center frequency value of the peak frequency segment, and its corresponding maximum amplitude is recorded at the same time; finally, the center frequency value and amplitude value are standardly formatted to generate a data pair with consistent structure for constructing the spectral response characteristic vector. The purpose of this process is to extract the characteristic points representing the strongest energy concentration in the continuous frequency segment, so that each peak frequency segment has a unique, clear, and quantifiable core indicator in the subsequent frequency matching and resonance analysis, thereby improving the accuracy and discrimination efficiency of the system in identifying abnormal energy excitation areas.

[0072] Combine the amplitude of the peak frequency segment with the average amplitude level of the frequency region in which it is located to calculate the corresponding frequency amplitude magnification. Combine the frequency offset and the frequency amplitude magnification as input factors, and calculate the frequency proximity index by ratio normalization to quantify the local similarity between the peak frequency segment and a certain natural frequency value.

[0073] The implementation process for calculating the frequency proximity index through ratio normalization uses the frequency offset and corresponding frequency amplitude magnification between each peak frequency segment and the target parasitic resonant frequency as joint input variables to establish a normalized mapping relationship to measure whether the peak frequency is in an excitable state. Specifically, the absolute offset between the peak frequency value and the reference resonant frequency value is first calculated and divided by the maximum acceptable offset threshold to obtain the normalized frequency offset component. Simultaneously, the amplitude magnification of the peak frequency segment is divided by the maximum allowable magnification within the reference amplitude range to obtain the normalized amplitude intensity component. These two components are then combined using a weighting function or product function to output a frequency proximity index limited to 0 to 1. The closer this index value is to 1, the closer the current frequency point is to the natural frequency point of the parasitic resonant path and the stronger the energy excitation effect. Its purpose is to provide a unified metric for establishing comparability of resonance risk across multiple peak frequency segments, thereby providing a mathematical foundation and quantitative basis for subsequent coupling strength scoring and dynamic frequency modulation decision-making.

[0074] The above steps quantify the significance of the spike response by evaluating the relationship between the amplitude amplification and the frequency offset. The frequency proximity metric can indicate whether the frequency band has a high risk of resonant triggering.

[0075] The frequency proximity index is weighted and fused with the pre-set frequency coupling factor to construct a resonance coupling strength score that reflects the energy coupling relationship between the peak frequency segment and the parasitic resonance path. This score serves as the criterion input for subsequent nonlinear response index evaluation and dynamic frequency modulation control triggering, thereby achieving quantitative identification and precise response to potential resonance risks.

[0076] The implementation process of weighted fusion of the frequency proximity index and the pre-set frequency coupling factor to construct a resonant coupling strength score includes three steps: factor normalization, weighting coefficient setting, and fusion calculation. First, the calculated frequency proximity index is normalized to ensure that its value is between 0 and 1. At the same time, the frequency coupling factor is amplitude compressed or mapped to make it additive or multiplicative in the same numerical space as the frequency proximity. Then, based on the power system's sensitivity to frequency offset and the weight of the coupling path, a set of weighting coefficients is set, acting on the proximity index and coupling factor respectively to reflect the system's response priority to the two types of factors. Finally, the two indicators are combined into a single numerical output through weighted averaging or nonlinear function fusion to form the final resonant coupling strength score. The higher the score, the stronger the energy coupling caused by the current peak frequency segment in the parasitic resonant path and the greater the risk of resonant excitation.

[0077] This process aims to integrate multiple factors from different physical sources into a unified, comparable metric, enhancing the system's ability to identify potential resonance hazards. By scoring the resonance coupling strength, it is possible to quickly identify peak frequency bands with the greatest excitation risk in complex spectral environments. This provides direct criteria for subsequent dynamic control strategies such as nonlinear resonance response index calculation and PWM modulation frequency adjustment, significantly improving the system's ability to prevent and control resonance penetration and its accuracy.

[0078] The above steps combine multiple key factors to generate a unified metric for characterizing the strength of energy coupling. The resonant coupling strength score provides a quantifiable basis for determining whether dynamic frequency modulation is required.

[0079] A frequency matching operation is performed on each peak frequency segment in the spectral response eigenvector. The peak frequency values ​​are compared item by item with a list of pre-modeled parasitic resonant frequencies. The frequency offset and frequency amplitude magnification are then calculated. The frequency proximity is then calculated by combining these offsets and magnifications. A frequency coupling factor is then introduced to construct a resonant coupling strength score. Ultimately, the energy excitation intensity between the peak frequency segment and the parasitic resonant path is quantified. This method's core function is to quantitatively correlate frequency domain characteristics with the power system resonant model, identify high-risk frequency segments, and provide early warning signals for the control system. In power systems, resonant paths formed by parasitic parameters have fixed natural frequency characteristics. The presence of peak frequency segments may indicate energy coupling between the PWM modulation frequency or its harmonics and these natural frequencies. By comparing the peak frequency values ​​with the parasitic resonant frequencies one by one, their offsets can be accurately calculated and the degree of proximity can be determined. The amplitude magnification reflects the energy response intensity, thus establishing a dual-metric system of "proximity + intensity." Further weighted fusion of the frequency coupling factor yields a resonant coupling strength score that not only unifies the conversion relationship between physical properties and mathematical quantification but also provides a direct triggering basis for subsequent modulation frequency avoidance, response index assessment, and controller frequency modulation actions. This mechanism significantly enhances the system's awareness of hidden resonance risks, shifting power supply control from passive compensation to active prevention, improving system stability, robustness, and responsiveness.

[0080] The frequency offset, frequency amplitude magnification, and resonance coupling strength score are input into a nonlinear mapping function. The input parameters are processed using a normalization function, and an excitation function is introduced to amplify and map the high-order response characteristics within the critical frequency range. The output is a nonlinear resonance response index limited to between 0 and 1. The closer the index value is to 1, the closer the current pulse width modulation frequency is to the inherent excitation frequency of the parasitic resonance path, and the stronger the high-frequency energy concentration effect.

[0081] The normalization function is used to process the input parameters, and the excitation function is introduced to amplify and map the high-order response characteristics in the critical frequency range. The specific steps for outputting the nonlinear resonance response index are as follows:

[0082] The frequency offset, frequency amplitude magnification, and resonant coupling strength score corresponding to each peak frequency segment are used as three input parameters and uniformly input into the nonlinear mapping function model. Each input parameter is normalized so that its value is limited to between 0 and 1 to ensure the numerical compatibility and controllability of each parameter in the fusion calculation process.

[0083] The above steps unify the numerical ranges of frequency offset, amplitude amplification, and coupling strength score, ensuring the comparability and computational stability of input parameters of different dimensions in nonlinear mapping.

[0084] A nonlinear mapping function model is a computational structure that integrates and transforms the complex relationships between multiple input parameters through nonlinear mathematical expressions, reflecting the nonlinear influence weights of the input parameters on the output results. In power system spectrum analysis, the frequency offset, frequency amplitude magnification, and resonant coupling strength score in the peak frequency band do not have a linear cumulative relationship, but rather exhibit nonlinear characteristics such as coupling amplification and critical mutation. By constructing a nonlinear mapping function model, these three types of input factors can be fused according to a set functional relationship (such as a weighted power function, exponential function, or logistic S-type function), resulting in an intermediate calculation result that accurately reflects the excitation potential of the peak frequency band. The purpose of this model is to break through the limitations of traditional linear weighting methods and truly reproduce the complex dynamic laws of resonance caused by the superposition of frequency proximity, energy mutation, and path coupling in power systems, providing a reasonable calculation framework and physical basis for outputting a more discriminative resonant response index.

[0085] Based on the normalized input, a multivariate combination relationship is constructed through a mapping function to identify input samples in the critical frequency range. For parameter combinations in highly sensitive areas, an excitation function is introduced to perform nonlinear amplification processing, so that the excitation signal in the critical frequency band receives a higher weight in the calculation results, thereby enhancing the system's response to potential strong coupling states.

[0086] The above steps enhance the influence of highly sensitive parameter combinations within the critical frequency range on the output results, highlight potential high-risk resonant frequency bands, and improve the system's sensitivity to identifying dangerous frequencies.

[0087] When constructing multivariable combinations and identifying input samples within the critical frequency range, the system first logically integrates the three normalized parameters—frequency offset, amplitude amplification, and coupling strength score—to form a unified risk assessment benchmark. This integration is accomplished through a nonlinear mapping function. Essentially, it assigns different response weights to each parameter based on its impact on resonant excitation, then combines these weights into a comprehensive assessment result. During this assessment, if the system finds a parameter combination with minimal frequency offset and high amplitude and coupling strength, it determines that this combination is within the critical frequency range, potentially close to the conditions for parasitic resonance excitation. The purpose of the nonlinear mapping function is to highlight the coupling relationship between parameters, rather than simply weighting them, thereby improving the system's accuracy in identifying risk conditions within the boundary frequency range. This function typically uses a functional model with good gradient capability and response sensitivity demonstrated in engineering to more realistically reflect complex frequency domain behavior.

[0088] After identifying highly sensitive input combinations within the critical frequency range, the system further amplifies the response of these parameter combinations using an excitation function. The excitation function nonlinearly enhances input combinations that are within the parameter threshold but very close to the high-risk boundary, giving them a higher weight in the final output. The significance of this processing approach lies in the fact that the system not only identifies "whether it is close to risk" but also emphasizes "how close" it is. This introduces gradient perception into the risk level determination, allowing the control system to make more proactive frequency adjustments in advance. The excitation function was chosen because it exhibits a rapid growth trend in output values, making it ideal for simulating the system's sensitive response to boundary conditions. It is widely used in neural networks, signal amplification, and control strategies, and is also highly adaptable in power supply control systems.

[0089] The results after nonlinear combination and excitation transformation are reduced and mapped to output a response index value limited to 0 to 1, forming a nonlinear resonance response index that reflects the excitation risk level of the peak frequency segment, which serves as the core reference indicator for dynamic modulation frequency adjustment and controller decision triggering.

[0090] The above steps compress the complex multi-parameter input results into a quantitative index between 0 and 1, which is used to clearly represent the intensity of the resonant excitation caused by the current peak frequency band, providing a direct criterion for subsequent dynamic control.

[0091] The frequency offset, frequency amplitude magnification, and resonant coupling strength score are input into a nonlinear mapping function. After normalization and amplification mapping using an excitation function, the final output is a nonlinear resonant response index, limited to 0 and 1. Its core function is to convert multiple frequency-domain risk indicators into a quantitative indicator with clear physical meaning, enabling the system to quickly determine whether the current frequency state has a high risk of triggering resonance within the control strategy. Specifically, the frequency offset reflects the proximity between the current frequency and the parasitic resonant frequency, the amplitude magnification represents the local energy excitation intensity in this frequency range, and the resonant coupling strength score reflects the system structure's sensitivity to energy accumulation in this frequency range. Combining these three factors through a nonlinear mapping function effectively simulates the superposition effect of nonlinear frequency responses in actual power systems. Furthermore, the normalization function ensures comparability of the input parameters across different amplitude ranges, preventing the output from being dominated by a single parameter value that is too large. Furthermore, the excitation function is introduced to amplify high-risk combinations within the critical frequency range, further enhancing the system's ability to identify near-resonant states. The resulting nonlinear resonance response index ranges from 0 to 1, with higher values ​​indicating a greater excitation risk. This index can be directly used as a basis for decision-making in the frequency modulation control logic. This mechanism enables the system to quickly, accurately, and quantitatively identify potential resonance risks in the face of complex, multi-source, and multi-variable interference, enabling a shift from traditional static control to dynamic predictive control, significantly improving the safety, responsiveness, and robustness of the power system.

[0092] When the nonlinear resonance response index is greater than the reference threshold, the main switching frequency of the pulse width modulation and its higher harmonic distribution are dynamically adjusted. The modulation frequency is controlled by a digital controller to drift to a frequency band that does not overlap with the parasitic resonance frequency, avoiding the resonance excitation frequency point, suppressing the high-frequency energy superposition effect, and avoiding the resonance penetration phenomenon.

[0093] When the nonlinear resonance response index is greater than the reference threshold, spectrum dynamic control is performed. Based on the nonlinear resonance response index and the nonlinear resonance response index reference threshold, a frequency adjustment offset factor is constructed. The frequency adjustment offset factor determines the adjustment amplitude and direction of the pulse width modulation main frequency. The calculation expression is as follows:

[0094]

[0095] , where Δf a is the frequency adjustment offset factor, which is used to determine how much the main switching frequency should be offset. The larger the value, the higher the current risk and the more intense the frequency adjustment behavior. α is the frequency adjustment sensitivity coefficient, which is used to adjust the global amplification strength of the offset. The larger α is, the faster the frequency response is; the smaller α is, the smoother the adjustment is, which is suitable for high-precision scenarios. β is the nonlinear amplification order, which is used to adjust the χ close to 1. rAchieve exponential response amplification, so that when the resonance index approaches the critical state, the output adjustment offset is significantly enhanced, highlighting the high-risk response. The value is 2 or 3. If you want the controller to be more sensitive to high-risk conditions, you can increase this value. r It is the nonlinear resonance response index, which indicates the resonance excitation intensity of the current peak frequency segment. It is a risk indicator that maps the frequency offset, amplitude magnification and coupling strength score through a nonlinear function. The value range is 0-1. It is used as a quantitative criterion for the resonance risk of the current frequency point and is an important basis for the start-up and adjustment of the frequency modulation control amplitude. th is the reference threshold of the nonlinear resonance response index, (χ r -x th ) + is the positive difference after the index exceeds the threshold, and the symbol indicates that when χ r > th Output χ r -χ th , otherwise it is 0, ensuring that the frequency offset takes effect only when the resonance risk exceeds the set threshold, thereby avoiding unnecessary fluctuations in the system frequency and enhancing the target and stability of the frequency modulation behavior;

[0096] The purpose of this step is to quantify the resonance excitation risk detected by the system into a frequency offset instruction with dynamic response characteristics, thereby providing a precise, risk-driven adjustment basis for the frequency modulation control behavior.

[0097] Adjust the offset factor Δf based on the generated frequency a , dynamically reconstruct the main switching frequency of pulse width modulation and its harmonic energy distribution, the calculation expression is as follows:

[0098]

[0099] , where f new is the adjusted main switching frequency, f base is the originally set PWM main switching frequency, γ(t) is the frequency drift direction factor based on time change, with a value range of -1-1, which is used to control the direction of frequency offset, whether to increase or decrease it, so as to avoid the resonant response or high-order coupling peak of the current frequency upward; adaptive avoidance is achieved by combining the dynamic state of the system, H new (n) is the energy amplitude corresponding to the adjusted nth-order high-order harmonic frequency, which is the harmonic intensity that the system hopes to retain after frequency adjustment. After exponential decay processing, this value is significantly reduced, thereby reducing the risk of resonance penetration caused by high-frequency harmonics. base(n) is the original amplitude of the nth-order higher harmonic frequency before adjustment, e is the natural base, and λ is the harmonic suppression coefficient, which is used to control the sensitivity of the system to the attenuation of the higher harmonic amplitude when the frequency shifts. The larger the value, the more sensitive the system is to the shift and the stronger the harmonic energy suppression. It is usually set according to the system filtering capability and electromagnetic compatibility design requirements.

[0100] The purpose of this step is to move the modulation frequency away from the high-risk frequency point as a whole through mathematical modeling, and synchronously adjust the amplitude of the higher harmonics to avoid the superposition of energy in the critical frequency band, and ultimately achieve active avoidance of the resonant excitation area, thereby improving the stability and anti-interference ability of the system.

[0101] Through the above-mentioned high-speed power conversion control method, a full-process closed-loop control of real-time identification, quantitative evaluation and dynamic avoidance of parasitic resonance excitation risks is achieved, effectively solving the "resonance penetration" phenomenon caused by the coupling of high-frequency PWM modulation and system parasitic parameters in the existing technology. This method establishes a complete quantitative mechanism from waveform perception to risk warning by performing fine-grained monitoring of high-frequency energy accumulation behavior in the frequency domain dimension and using the spectrum response eigenvector combined with a nonlinear mapping function to output the resonance response index; further, the digital controller dynamically adjusts the PWM main frequency and its harmonic energy distribution to achieve coordinated control of frequency modulation avoidance and harmonic suppression. This solution not only improves the system's adaptive adjustment capability and robustness in the face of frequency domain resonance interference, but also significantly enhances the stability, electromagnetic compatibility and control accuracy of the power conversion process, and has significant engineering practical value and innovation.

[0102] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0103] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0104] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0105] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0106] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0107] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0108] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0109] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0110] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0111] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A high-speed power conversion control method, characterized in that: The following steps are involved: Through the analog-to-digital conversion module, the transient voltage waveform and transient current waveform at both ends of the power device in each pulse width modulation cycle are synchronously collected, and a high-frequency waveform data frame containing multiple modulation cycles is constructed based on a fixed time sliding window mechanism; Perform fast Fourier transform on high-frequency waveform data frames to extract the spectral amplitude distribution characteristics of the main frequency component and multiple high-order harmonic frequency components in the frequency domain, construct an energy spectrum density curve corresponding to the frequency, identify the peak frequency segments with abnormally amplified amplitudes in the energy spectrum density curve, extract the frequency value and corresponding amplitude of each peak frequency segment, and form a spectrum response feature vector; A frequency matching operation is performed on the peak frequency segment in each spectral response eigenvector. The peak frequency value is compared item by item with the pre-modeled parasitic resonance frequency list. The frequency offset and frequency amplitude magnification are calculated. The offset and magnification are combined to calculate the frequency proximity. The frequency coupling factor is introduced to construct the resonance coupling strength score, which quantifies the energy excitation intensity between the peak frequency segment and the parasitic resonance path. The frequency offset, frequency amplitude magnification and resonance coupling strength score are input into a nonlinear mapping function, the input parameters are processed by a normalization function, an excitation function is introduced to amplify and map the high-order response characteristics in the critical frequency range, and the output is a nonlinear resonance response index limited to between 0 and 1; When the nonlinear resonance response index is greater than the reference threshold, the main switching frequency of the pulse width modulation and its higher harmonic distribution are dynamically adjusted. The modulation frequency is controlled by a digital controller to drift to a frequency band that does not overlap with the parasitic resonance frequency, avoiding the resonance excitation frequency point and suppressing the high-frequency energy superposition effect.

2. The high-speed power conversion control method according to claim 1, wherein: The specific steps for constructing a high-frequency waveform data frame containing multiple modulation cycles are as follows: Based on the pulse width modulation cycle clock signal, the analog-to-digital conversion module is synchronously triggered so that the sampling starting points of the voltage sampling channel and the current sampling channel are strictly aligned within each modulation cycle, ensuring that the sampling starting points of the voltage waveform and the current waveform on the time axis are consistent; Within each modulation cycle, multiple equally spaced sampling points are evenly divided to fully cover the rising edge stage, stable conduction stage, falling edge stage and complete shutdown stage of the power device conduction process, and construct single-cycle transient voltage waveform units and transient current waveform units with structural consistency; A plurality of continuous single-cycle waveform units are sequentially spliced ​​according to a fixed sliding interval to generate a high-frequency waveform data frame containing a plurality of modulation cycles, and the data frame is cached in a high-speed cache.

3. The high-speed power conversion control method according to claim 1, wherein: The steps for generating the spectral response feature vector are as follows: Perform fast Fourier transform operations on high-frequency waveform data frames to obtain spectrum amplitude data within the complete frequency range, and extract the amplitude distribution information of the main frequency component and multiple high-order harmonic frequency components in the frequency domain; The frequency components and corresponding amplitude data points are mapped into a one-dimensional energy spectrum density curve in ascending order of frequency. The peak response area with abnormally amplified amplitude is identified in the curve. The central frequency value and maximum amplitude are obtained by calculating the local maximum value of each peak area to form a peak frequency segment with response characteristics. All peak frequency segments are arranged in order according to the frequency dimension, and the center frequency value, maximum amplitude and relative position of each peak frequency segment in the entire spectrum are combined to construct a three-dimensional vector structure to generate a spectrum response feature vector.

4. The high-speed power conversion control method according to claim 3, wherein: Perform a fast Fourier transform operation on the high-frequency waveform data frame to obtain the spectrum amplitude data within the complete frequency range, and extract the amplitude distribution information of the main frequency component and multiple high-order harmonic frequency components in the frequency domain. The specific steps are as follows: Based on the constructed high-frequency waveform data frame, the input parameters of the fast Fourier transform are initialized, and the transient voltage value and transient current value of each sampling point are respectively sorted into two groups of real time domain signal sequences of equal length; The two sets of time domain signal sequences were subjected to window function weighting processing respectively, and the Blackman window function was used to smooth the transition of the signal boundary and reduce the spectrum leakage effect caused by edge discontinuity. Call the high-precision fast Fourier transform operation module to convert the weighted real signal sequence into complex spectrum data arranged in ascending frequency, calculate the amplitude corresponding to each frequency point, and output a set of frequency domain amplitude arrays; The frequency points where the main frequency components are located and their corresponding energy amplitudes are identified from the frequency domain amplitude array, and multiple physically meaningful higher harmonic frequency points and their amplitudes are located to construct a complete frequency domain amplitude distribution set.

5. The high-speed power conversion control method according to claim 1, wherein: Perform frequency matching on the peak frequency segments in each spectral response eigenvector to construct a resonant coupling strength score. The specific steps are as follows: For each peak frequency segment in the spectral response eigenvector, extract its center frequency value and corresponding amplitude, perform frequency matching comparisons with each natural frequency value in the pre-modeled parasitic resonance frequency list, and calculate the frequency offset between each peak frequency value and the corresponding natural frequency value; Combine the amplitude of the peak frequency segment with the average amplitude level of the frequency region where it is located to calculate the corresponding frequency amplitude magnification factor. Combine the frequency offset and the frequency amplitude magnification factor as input factors and calculate the frequency proximity index through ratio normalization. The frequency proximity index is weighted and fused with the pre-set frequency coupling factor to construct a resonance coupling strength score that reflects the energy coupling relationship between the peak frequency segment and the parasitic resonance path.

6. The high-speed power conversion control method according to claim 1, wherein: The normalization function is used to process the input parameters, and the excitation function is introduced to amplify and map the high-order response characteristics in the critical frequency range. The specific steps for outputting the nonlinear resonance response index are as follows: The frequency offset, frequency amplitude magnification, and resonance coupling strength score corresponding to each peak frequency band are used as three input parameters and uniformly input into the nonlinear mapping function model, and each input parameter is normalized. Based on the normalized input, a multivariate combination relationship is constructed through a mapping function to identify input samples in the critical frequency range. For parameter combinations in highly sensitive areas, an excitation function is introduced to perform nonlinear amplification processing, so that the excitation signal in the critical frequency band receives a higher weight in the calculation result. The results after nonlinear combination and excitation transformation are reduced and mapped to form a nonlinear resonance response index that reflects the excitation risk level of the peak frequency band.

7. The high-speed power conversion control method according to claim 1, wherein: When the nonlinear resonance response index is greater than the reference threshold, spectrum dynamic control is performed. Based on the nonlinear resonance response index and the nonlinear resonance response index reference threshold, a frequency adjustment offset factor is constructed. The frequency adjustment offset factor determines the adjustment amplitude and direction of the pulse width modulation main frequency. The calculation expression is as follows: Where Δf a is the frequency adjustment offset factor, α is the frequency adjustment sensitivity coefficient, β is the nonlinear amplification order, χ r is the nonlinear resonant response index, χ th is the reference threshold of the nonlinear resonance response index, (χ r -χ th ) + is the positive difference after the index exceeds the threshold; Adjust the offset factor Δf based on the generated frequency a , dynamically reconstruct the main switching frequency of pulse width modulation and its harmonic energy distribution, the calculation expression is as follows: Where, f new is the adjusted main switching frequency, f base is the original set PWM main switching frequency, γ(t) is the frequency drift direction factor based on time change, H new (n) is the energy amplitude corresponding to the adjusted nth-order higher harmonic frequency, H base (n) is the original amplitude of the nth-order higher harmonic frequency before adjustment, e is the natural base, and λ is the harmonic suppression coefficient.

Images