Fast phase-locking method for inverter based on cordic algorithm

Abstract

The present invention discloses a fast phase-locking method for an inverter based on a cordic algorithm, which relates to the technical field of fast phase-locking of inverters and includes the following steps: balancing the dynamic response capability and the high-frequency noise suppression capability through a preset bandwidth range to achieve accurate initial tracking of the frequency and phase of the input signal. The present invention achieves accurate initial tracking of the input signal through a preset bandwidth, and combines real-time feature extraction with deep learning model prediction to intelligently identify the signal change state. When the signal is stable, the system maintains the preset bandwidth to balance the dynamic response and noise suppression; when a perturbation frequency change is detected, the bandwidth range is dynamically expanded to quickly track the frequency and phase changes. The envelope fluctuation and frequency band energy distribution characteristics are extracted, a reference value is generated and a signal change index is predicted, the bandwidth is dynamically adjusted, the phase offset accumulation is avoided, reactive power fluctuations and harmonic distortion are reduced, and the power quality and system reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of inverter fast phase locking, and in particular to an inverter fast phase locking method based on a cordic algorithm. Background Art

[0002] Fast inverter phase locking based on the CORDIC (COordinate Rotation DIgital Computer) algorithm refers to the use of the CORDIC algorithm's efficient trigonometric function calculation capabilities to achieve rapid estimation and synchronization of the input signal phase in the inverter system. Specifically, the CORDIC algorithm iteratively calculates the angle of the rotation vector and can calculate the sine and cosine values in real time with low hardware complexity to estimate the instantaneous phase and amplitude of the signal. In inverter control, this algorithm can be used to implement a fast phase-locked loop (PLL) to ensure that the inverter's output voltage or current remains synchronized with the grid signal, especially when the signal frequency changes or there is harmonic interference. Compared with traditional phase-locked loops, this method has the advantages of faster response speed and lower computing resource usage, thereby improving the dynamic performance of the inverter and system stability.

[0003] The existing technology has the following deficiencies:

[0004] The bandwidth of a phase-locked loop (PLL) refers to the frequency range its loop filter allows to pass, typically defined as the range from direct current (DC) to the cutoff frequency. As a key parameter, bandwidth determines the system's responsiveness to input signal frequency variations and its ability to suppress high-frequency noise. Traditionally, the bandwidth of a PLL is typically fixed, balancing dynamic responsiveness and noise suppression to suit specific application scenarios. However, when the frequency regulation of the input signal source (such as a wind farm or distributed generation system) is unstable and exhibits small periodic drifts, this fixed bandwidth may not meet practical requirements. In this case, the PLL may be unable to effectively track the frequency drift, causing the phase output to gradually deviate from the input signal's reference. This can result in an increased phase difference between the inverter output current and the grid signal, leading to increased reactive power and equipment detuning. Phase misalignment can also cause increased harmonics and degrade power quality. Furthermore, accumulated phase error can trigger frequency detuning alarms or grid disconnection protection, impacting system stability and even causing equipment shutdown or grid fluctuations.

[0005] 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

[0006] The purpose of the present invention is to provide a fast inverter phase-locking method based on the Cordic algorithm. This method achieves precise initial tracking of the input signal through a preset bandwidth, and intelligently identifies signal change states by combining real-time feature extraction with deep learning model prediction. When the signal is stable, the system maintains the preset bandwidth to balance dynamic response and noise suppression. When a perturbation frequency change is detected, the bandwidth range is dynamically expanded to quickly track frequency and phase changes. The envelope fluctuation and frequency band energy distribution characteristics are extracted, a reference value is generated, and a signal change index is predicted. The bandwidth is dynamically adjusted to avoid phase offset accumulation, reduce reactive power fluctuations and harmonic distortion, and improve power quality and system reliability to address the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a fast phase-locking method for an inverter based on a cordic algorithm, comprising the following steps:

[0008] Balance dynamic response and high-frequency noise suppression capabilities through a preset bandwidth range to achieve accurate initial tracking of input signal frequency and phase;

[0009] During the signal tracking process, parameter information of the input signal is obtained in real time to provide basic data for subsequent analysis and decision-making;

[0010] Preprocess the signal parameters acquired in real time and extract key features from the preprocessed signal parameters that reflect the unstable frequency regulation of the signal source and the occurrence of small periodic drift trends. Analyze the extracted features within the detection window to quantify the change trend and amplitude of the input signal.

[0011] The analyzed features are input into a pre-learned deep learning model, which then makes intelligent predictions on the analyzed features and identifies the signal change status.

[0012] According to the prediction results of the deep learning model, the signal change state is divided into perturbation frequency change and reference frequency stability;

[0013] For reference frequency stability, the preset bandwidth range is continued to be used to maintain a balance between dynamic response capability and noise suppression performance, ensuring high-precision tracking and low noise interference under stable signals;

[0014] In response to the perturbation frequency changes, the bandwidth range is dynamically expanded, the response speed of the phase-locked loop is enhanced, the frequency and phase changes of the input signal are quickly tracked, and the accumulation of phase offset is avoided.

[0015] Preferably, key features reflecting the unstable frequency regulation of the signal source and the trend of small periodic drift are extracted from the preprocessed signal parameters. The extracted features include the amplitude change characteristics of the input signal envelope fluctuation over time and the distribution change trend of the signal energy in the frequency band over time. Under the detection window, after analyzing the extracted amplitude change characteristics of the input signal envelope fluctuation over time and the distribution change trend of the signal energy in the frequency band over time, an envelope fluctuation reference value and a frequency band energy distribution reference value are generated respectively, and the change trend and amplitude of the input signal are quantified by the envelope fluctuation reference value and the frequency band energy distribution reference value.

[0016] Preferably, the specific steps of analyzing the amplitude variation characteristics of the input signal envelope fluctuation over time in the detection window to generate the envelope fluctuation reference value are as follows:

[0017] First, the instantaneous envelope of the input signal is constructed using its instantaneous information. The instantaneous envelope is a key characteristic that reflects the dynamic changes in the input signal amplitude. The formula is as follows:

[0018]

[0019] Where x(n) is the input signal, H[x(n)] is the Hilbert transform of the input signal, and E(n) is the instantaneous envelope;

[0020] Extract the fluctuation rate of the signal envelope and calculate the intensity and trend of the envelope over time by taking the derivative of the instantaneous envelope E(n). The fluctuation rate calculation expression is as follows:

[0021]

[0022] Where, is the first derivative of the instantaneous envelope, is the second derivative of the instantaneous envelope, γ is the adjustment factor, and R(n) is the envelope fluctuation rate;

[0023] The fluctuation intensity factor is calculated by the envelope fluctuation rate R(n) and the instantaneous envelope E(n). The calculation expression is as follows:

[0024]

[0025] Where sin(α·n) is the sinusoidal modulation term, α is the modulation frequency factor, n is the sequence index of the discrete signal, N is the total number of detection window time points, and F is the fluctuation intensity factor;

[0026] The instantaneous envelope E(n), envelope fluctuation rate R(n) and fluctuation intensity factor F are combined to generate the envelope fluctuation reference value. The generation formula is as follows:

[0027]

[0028] Where η is the weight factor, β is the adjustment coefficient, and EFI is the envelope fluctuation reference value.

[0029] Preferably, the specific steps of analyzing the distribution change trend of the signal energy in the frequency band over time in the detection window to generate the frequency band energy distribution reference value are as follows:

[0030] First, the input signal is decomposed into multiple frequency bands and the instantaneous energy distribution of each frequency band is calculated. By accurately decomposing the frequency domain characteristics of the signal, the energy concentration or diffusion of the signal in different frequency ranges is captured. The calculation expression is as follows:

[0031]

[0032] Where, E k represents the energy value of the kth frequency band, f k，1 is the starting frequency of the kth frequency band, f k，2 is the end frequency of the kth frequency band, S(f) is the power spectral density of the signal, f is the frequency, W k (f) is the weighting function of the kth frequency band;

[0033] In each frequency band, the dynamic change characteristics of the signal energy over time are extracted to generate a characteristic vector for quantifying the energy fluctuation within the frequency band. The calculation expression is as follows:

[0034]

[0035] Where, E k，i is the energy value of the kth frequency band at the i-th detection point, E k，i+1 is the energy value of the kth frequency band at the i+1th detection point, that is, the energy value of the previous detection point, N k is the total number of detection points in the kth frequency band, g k (i) is the nonlinear weighting factor of the kth frequency band, D k is the dynamic change characteristic of the kth frequency band;

[0036] Perform pattern analysis on the dynamic change characteristics within each frequency band, integrate the energy distribution and change trend, and generate a pattern value that characterizes the energy distribution characteristics of the frequency band. The generation formula is as follows:

[0037]

[0038] Where, P k is the distribution mode value, indicating the distribution mode value of the kth frequency band, F k (E k , D k ) is the characteristic function, which is the energy value of the combined frequency band E k and dynamic change characteristics Dk The characteristic function, H k (f) is the intra-band weighting function;

[0039] The distribution mode values of all frequency bands are combined to generate the overall frequency band energy distribution reference value. The generation formula is as follows:

[0040]

[0041] Where K is the total number of frequency bands, ω k is the balance factor for the kth frequency band, is the nonlinear weight factor for the kth frequency band, γ k is the normalized weight factor of the kth frequency band, and R is the reference value of the frequency band energy distribution.

[0042] Preferably, the envelope fluctuation reference value and the frequency band energy distribution reference value generated after analysis are input into a pre-learned deep learning model, and a signal change index is generated by the deep learning model, and the signal change state is identified by the signal change index.

[0043] Preferably, the analyzed features are intelligently predicted using a pre-learned deep learning model, and the signal change index generated when the signal change state is identified is compared and analyzed with a pre-set signal change index reference threshold to divide the signal change state. The division steps are as follows:

[0044] If the signal change index is greater than a preset signal change index reference threshold, the signal state is classified as a perturbation frequency change;

[0045] If the signal variation index is less than or equal to a preset signal variation index reference threshold, the signal state is classified as reference frequency stability.

[0046] Preferably, the specific steps for dynamically expanding the bandwidth range, enhancing the response speed of the phase-locked loop, quickly tracking the frequency and phase changes of the input signal, and avoiding the accumulation of phase offset in response to the perturbation frequency change are as follows:

[0047] When the signal is in a perturbation frequency change, the phase-locked loop needs to calculate the dynamic bandwidth adjustment coefficient based on the signal change index deviation to guide the dynamic expansion of the bandwidth. The calculation expression is as follows:

[0048] α dyn =k1·ln(1+Δ sVDI )+k2

[0049] Where k1 is the proportional constant of the adjustment coefficient, k2 is the basic bandwidth adjustment offset, and α dyn is the dynamic bandwidth adjustment coefficient, Δ SVDI is the signal change index deviation, and the calculation expression is as follows: ΔSVDI =SVDI-SVDI ref , where SVDI is the signal variation index, SVDI ref is the signal change index reference threshold;

[0050] According to the preset bandwidth range and dynamic adjustment coefficient α dyn , calculate the new dynamic bandwidth range to improve the dynamic response capability of the phase-locked loop. The calculation expression is as follows:

[0051] BW dyn =BW pre ·(1+α dyn )

[0052] Where BW dyn is the adjusted bandwidth range, BW pre is the preset bandwidth range;

[0053] After calculating the adjusted bandwidth range BW dyn Afterwards, the phase-locked loop enhances its ability to track the input signal frequency and phase through a dynamic bandwidth adjustment mechanism. At this time, the phase tracking error is recalibrated to ensure the accuracy of phase synchronization. The calculation expression is as follows:

[0054]

[0055] Where θ err，new is the adjusted phase tracking error, θ err，pre is the phase tracking error before adjustment.

[0056] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0057] The present invention achieves accurate initial tracking of the input signal through a preset bandwidth range, and combines real-time signal feature extraction and deep learning model prediction to intelligently identify the signal change state. When the signal is in a reference frequency stability state, the system maintains the preset bandwidth range to balance the dynamic response capability and noise suppression performance; when a perturbation frequency change is detected, the system dynamically expands the bandwidth range to quickly track frequency and phase changes. This adaptive adjustment mechanism enables the phase-locked loop to cope with frequency regulation instability and small periodic drift problems that occur at the signal source (such as a wind farm or distributed power source), avoids phase offset and tracking lag caused by a fixed bandwidth design, and thus improves the tracking accuracy and response speed of the inverter.

[0058] The present invention extracts the key features of the input signal envelope fluctuation and frequency band energy distribution, and generates envelope fluctuation reference values and frequency band energy distribution reference values. The phase-locked loop can accurately identify the signal change trend and predict the signal change index through a deep learning model. When it is detected that the signal frequency regulation is unstable, the system can quickly and dynamically adjust the bandwidth to avoid the accumulation of phase offsets, reduce the phase difference between the inverter output current and the grid signal, and reduce reactive power fluctuations. At the same time, this dynamic adjustment mechanism effectively reduces harmonic distortion, improves the quality of power, and avoids the occurrence of system protection misoperation, thereby improving the operating reliability of the inverter and the power supply stability of the power grid, and reducing the risk of large-scale power outages or equipment failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] 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.

[0060] Figure 1 The flowchart of the method for fast phase locking of inverter based on Cordic algorithm of the present invention is shown. DETAILED DESCRIPTION

[0061] 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.

[0062] The present invention provides Figure 1 The inverter fast phase locking method based on the cordic algorithm shown includes the following steps:

[0063] Balance dynamic response and high-frequency noise suppression capabilities through a preset bandwidth range to achieve accurate initial tracking of input signal frequency and phase;

[0064] The preset bandwidth range refers to the fixed frequency range set for the loop filter of a phase-locked loop (PLL) during initial configuration. It is used to balance the system's dynamic response capability and high-frequency noise suppression performance. This range, typically from direct current (DC) to a specific cutoff frequency, is set based on the signal characteristics and performance requirements of the target application scenario. A wider preset bandwidth allows the PLL to quickly respond to dynamic changes in the input signal's frequency and phase, improving the system's tracking speed; while a narrower bandwidth helps suppress high-frequency noise and harmonic interference, enhancing signal stability and accuracy. The preset bandwidth range provides the PLL with an initial operating state, enabling it to achieve accurate initial tracking of the input signal under specific conditions and providing a foundation for subsequent dynamic adjustment or optimization.

[0065] The bandwidth of the phase-locked loop is a fixed parameter set by the loop filter, which determines the system's response speed to changes in the input signal and its ability to suppress high-frequency noise. In traditional designs, the bandwidth is optimized to adapt to specific application scenarios, such as stabilizing the grid frequency or specific load conditions. By presetting the bandwidth, the phase-locked loop can effectively suppress noise when the signal is stable and improve the accuracy of phase locking; when the signal changes, the bandwidth provides sufficient response speed to maintain signal synchronization and control accuracy, thereby ensuring efficient grid-connected operation of the inverter and the grid. Continuously sample and monitor the input signal to capture possible changes in the signal.

[0066] During the signal tracking process, parameter information of the input signal is obtained in real time to provide basic data for subsequent analysis and decision-making;

[0067] To accurately monitor and respond to changes in input signals, the system needs to continuously collect various parameters of the input signal, such as frequency, phase, and amplitude. These parameters are acquired in real time by sensors and data acquisition modules and updated at a high frequency. This step ensures the system's comprehensive understanding of the input signal, providing the necessary data foundation for subsequent signal analysis and feature extraction, ensuring the system can maintain efficient operation in a dynamically changing environment.

[0068] Preprocess the signal parameters acquired in real time and extract key features from the preprocessed signal parameters that reflect the unstable frequency regulation of the signal source and the occurrence of small periodic drift trends. Analyze the extracted features within the detection window to quantify the change trend and amplitude of the input signal.

[0069] Preprocessing is a key step in the data processing pipeline, involving operations such as filtering, denoising, and normalization. By applying digital filters or other signal processing techniques, the system can remove high-frequency noise and random interference from the input signal, enhancing signal clarity and stability. Preprocessing also involves synchronizing and normalizing the signal to ensure that feature extraction and analysis in subsequent steps can be performed on a consistent basis, thereby improving the accuracy and reliability of the overall system.

[0070] A signal source refers to the core device or system that generates the input signal. It is the direct source of signal variations and plays a key, leading role in the entire signal transmission and processing process. The state and operating characteristics of the signal source directly determine the stability and variation trends of the input signal's parameters, such as frequency, phase, and amplitude. In power systems, typical signal sources include dynamic power equipment such as wind farms, photovoltaic power plants, and distributed power sources. These devices, influenced by environmental factors (such as fluctuating wind speed and light intensity) or load fluctuations, may experience unstable frequency regulation, resulting in small periodic drifts in the signal source's output signal. This drift may manifest as phase shifts, amplitude fluctuations, or changes in harmonic components, impacting subsequent signal processing and control systems. Therefore, in a phase-locked loop (PLL) system, accurately monitoring and analyzing the dynamic characteristics of the signal source is crucial for achieving precise tracking and stable system operation. Signal source instability is the root cause of signal variations. Identifying and extracting the unstable characteristics of the signal source's frequency regulation facilitates more effective identification and processing of signal variations.

[0071] Key features reflecting the unstable frequency regulation of the signal source and the trend of small periodic drift are extracted from the preprocessed signal parameters. The extracted features include the amplitude variation characteristics of the input signal envelope fluctuation over time and the distribution variation trend of the signal energy in the frequency band over time. Within the detection window, after analyzing the extracted amplitude variation characteristics of the input signal envelope fluctuation over time and the distribution variation trend of the signal energy in the frequency band over time, an envelope fluctuation reference value and a frequency band energy distribution reference value are generated respectively. The variation trend and amplitude of the input signal are quantified by the envelope fluctuation reference value and the frequency band energy distribution reference value.

[0072] When the input signal's envelope fluctuation exhibits periodic amplitude variations over time, it typically indicates unstable frequency regulation and a tendency toward small periodic drift. This is because the periodic amplitude variations of the envelope fluctuation are often directly related to the frequency regulation dynamics of the signal source device. When frequency regulation is unstable, the signal frequency will exhibit small periodic drifts around the main frequency. This drift modulates the signal amplitude and energy distribution, causing periodic variations in the envelope fluctuation amplitude. This phenomenon is often caused by periodic disturbances in the signal source (such as wind speed variations in a wind farm or dynamic regulation of distributed power loads), reflecting the combined nonlinear characteristics of the signal in frequency, phase, and amplitude. The periodic envelope fluctuations further indicate that the signal's modulation characteristics are consistent with the periodic disturbances in the input signal frequency, revealing that the signal is experiencing unstable frequency regulation, accompanied by small periodic drifts. This envelope fluctuation characteristic provides an important observation dimension for analyzing input signal dynamics and is a key indicator for detecting frequency regulation anomalies.

[0073] The specific steps for analyzing the amplitude variation characteristics of the input signal envelope fluctuation over time in the detection window to generate the envelope fluctuation reference value are as follows:

[0074] First, the instantaneous envelope of the input signal is constructed using its instantaneous information. The instantaneous envelope is a key characteristic that reflects the dynamic changes in the input signal amplitude. The formula is as follows:

[0075]

[0076] Where x(n) is the input signal, which represents the discrete time series of the input signal, is the sampled value of the input signal at each time point, H[x(n)] is the Hilbert transform of the input signal, and E(n) is the instantaneous envelope, which is the instantaneous envelope of the input signal x(n), indicating the dynamic change of the signal amplitude over time.

[0077] Extract the fluctuation rate of the signal envelope and calculate the intensity and trend of the envelope over time by taking the derivative of the instantaneous envelope E(n). The fluctuation rate calculation expression is as follows:

[0078]

[0079] Where, is the first derivative of the instantaneous envelope, which calculates the rate of change of the envelope at each point, is the second-order derivative of the instantaneous envelope, which represents the rate of change of the envelope change rate, γ is the adjustment factor used to adjust the weight of the second-order derivative in the fluctuation rate calculation, and R(n) is the envelope fluctuation rate;

[0080] This step not only quantifies the changing trend of the envelope, but is also highly sensitive to nonlinear changes, providing core parameters for further analysis.

[0081] The fluctuation intensity factor is calculated by the envelope fluctuation rate R(n) and the instantaneous envelope E(n). The calculation expression is as follows:

[0082]

[0083] Where sin(α·n) is a sinusoidal modulation term used to enhance sensitivity to periodic changes, α is the modulation frequency factor, n is the sequence index of the discrete signal, representing the discrete time points within the detection window, N is the total number of time points in the detection window, and F is the fluctuation intensity factor.

[0084] By accumulating the product of these characteristics, the fluctuation intensity factor F describes the overall intensity of the envelope fluctuation within the detection window and is an important quantitative indicator of the signal's periodic drift trend.

[0085] The instantaneous envelope E(n), envelope fluctuation rate R(n) and fluctuation intensity factor F are combined to generate the envelope fluctuation reference value. The generation formula is as follows:

[0086]

[0087] Where η is the weight factor, β is the adjustment coefficient used to adjust the weight factor of the fluctuation intensity factor F, and EFI is the envelope fluctuation reference value.

[0088] The weight factor η is a factor used to adjust the signal envelope energy E(n) 2 and envelope fluctuation rate energy R(n) 2 Parameters of relative contribution. When analyzing the dynamic characteristics of the input signal, the envelope energy E(n) 2 It reflects the overall change characteristics of the signal amplitude, while the fluctuation rate energy R(n) 2 The weight factor η is used to balance the importance of these two parts of energy in the total energy calculation. Specifically, a larger value of η will increase the fluctuation rate R(n). 2 The weighting in the analysis makes the system more sensitive to dynamic changes in the envelope; conversely, a smaller value of η emphasizes the overall changes in the envelope amplitude and ignores smaller fluctuations. By flexibly adjusting η, the system can adapt to different signal environments, ensuring that the generated indicators more accurately reflect the true state of the signal. This is particularly crucial for analyzing frequency regulation instability and periodic drift trends.

[0089] The Envelope Fluctuation Reference Value (RFV) is generated by analyzing the amplitude variation of the input signal's envelope fluctuation over time within the detection window. A larger RFV value indicates more significant fluctuations and frequency of the signal's envelope, often reflecting instability in the signal source's frequency regulation and the presence of periodic drift. This increase in RFV value is highly correlated with the amplitude and regularity of frequency drift. Conversely, a smaller RFV value, or even close to zero, indicates low and stable fluctuations, indicating stable frequency regulation and no significant periodic drift.

[0090] When the distribution of signal energy within a frequency band shows a periodic increase in the energy percentage of the frequency band over time, this typically indicates that the current input signal is experiencing unstable frequency regulation and is exhibiting a trend of small periodic drift. The root cause of this phenomenon is that the instability of frequency regulation causes a slight shift in the signal's main frequency, resulting in a periodic redistribution of energy between the main frequency band and adjacent frequency bands. Specifically, when the frequency regulation of a signal source (such as a wind farm or distributed power source) experiences periodic perturbations, the energy in the spectrum shifts periodically with the regulation. For example, energy concentrated in the main frequency band diffuses upward or downward to adjacent frequency bands and returns to the main frequency within a certain period. This periodic increase or decrease in the energy percentage of the frequency band reflects the significant influence of the regulation perturbation on the signal spectrum and is a typical characteristic of frequency instability. Furthermore, the dynamic fluctuations in the energy distribution of the frequency bands also indicate that the regulation process of the signal source has a certain regularity and periodicity, consistent with the trend of small periodic drift.

[0091] The specific steps for analyzing the distribution trend of signal energy within the frequency band over time in the detection window to generate the frequency band energy distribution reference value are as follows:

[0092] First, the input signal is decomposed into multiple frequency bands and the instantaneous energy distribution of each frequency band is calculated. By accurately decomposing the signal's frequency domain characteristics, the energy concentration or diffusion of the signal in different frequency ranges is captured, laying the foundation for subsequent dynamic change characteristic analysis. The calculation expression is as follows:

[0093]

[0094] Where, E k represents the energy value of the kth frequency band, which represents the sum of the signal power in the frequency band, f k，1 is the starting frequency of the kth frequency band, f k，2 is the end frequency of the kth frequency band, S(f) is the power spectral density of the signal, which is the power spectral density function of the signal in the frequency domain, describing the energy distribution of the signal at different frequencies f, where f is the frequency, and W k (f) is a weighting function for the kth frequency band, which is a weighting function set for the frequency components within the kth frequency band and is used to adjust the contribution of the components of different frequencies f within the frequency band;

[0095] By calculating E k , it can quantify the energy distribution characteristics of the signal in each frequency band and provide basic data for the next step of analyzing its dynamic change trend.

[0096] In each frequency band, the dynamic change characteristics of the signal energy over time are extracted to generate a characteristic vector for quantifying the energy fluctuation within the frequency band. This vector reflects the periodic changes in energy distribution that may be caused by unstable signal frequency regulation. The calculation expression is as follows:

[0097]

[0098] Where, E k，i is the energy value of the kth frequency band at the i-th detection point, which represents the energy distribution of the frequency band over time. k，i+1 is the energy value of the kth frequency band at the i+1th detection point, that is, the energy value of the previous detection point, N k is the total number of detection points in the kth frequency band, g k (i) is the nonlinear weighting factor of the kth frequency band, which is used to emphasize the contribution of the energy change at a specific time point in the frequency band to the overall dynamic characteristics. D is the dynamic change characteristic of the kth frequency band, which describes the energy change trend and fluctuation amplitude in the kth frequency band.

[0099] By calculating D k , it is possible to quantify the energy variation trend of the signal within the frequency band and identify the dynamic instability of the signal during the regulation process.

[0100] The dynamic change characteristics within each frequency band are analyzed in mode, and the energy distribution and change trend are integrated to generate a mode value that characterizes the energy distribution characteristics of the frequency band. The purpose is to capture the overall law of energy transfer between frequency bands. The generation formula is as follows:

[0101]

[0102] Where, P k is the distribution mode value, which represents the distribution mode value of the kth frequency band and is used to quantify the distribution characteristics of the energy in the frequency band and its relationship with dynamic changes. k (E k , D k ) is the characteristic function, which is the energy value of the combined frequency band E k and dynamic change characteristics D k Characteristic function, used to integrate the static and dynamic characteristics within the frequency band, H k (f) is the intra-band weighting function, which is used to emphasize the contribution of specific frequencies within the band;

[0103] By P kThe calculation can quantify the energy distribution pattern of each frequency band and its correlation with the dynamic change characteristics, providing a basis for generating an overall reference value.

[0104] The distribution mode values of all frequency bands are combined to generate the overall frequency band energy distribution reference value, which is used to characterize the global dynamic characteristics of the signal energy distribution. This step provides an important quantitative indicator of the signal frequency domain stability. The generation formula is as follows:

[0105]

[0106] Where K is the total number of frequency bands, ω k is the balance factor of the kth frequency band, and is the positive correction parameter used to adjust the offset of the reference value of the kth frequency band. is the nonlinear weight factor of the kth frequency band, which is an exponential factor that controls the nonlinear amplification of the contribution of the kth frequency band, γ k is the normalized weight factor of the kth frequency band, which is the proportional factor that controls the contribution of the kth frequency band in the denominator and is used to normalize the contribution of different frequency bands. R is the reference value of the frequency band energy distribution.

[0107] By calculating the reference value R, we can integrate the energy distribution pattern of each frequency band and its importance to form an integrated indicator that characterizes the frequency domain characteristics of the signal, providing a reliable benchmark for dynamic analysis and state judgment of the signal.

[0108] The frequency band energy distribution reference value, generated by analyzing the temporal distribution of signal energy within the frequency band within the detection window, is generated. A larger performance value indicates that the current input signal is more likely to be experiencing frequency regulation instability and exhibiting a small periodic drift trend. Conversely, a smaller performance value indicates that the signal is relatively stable and lacks significant frequency regulation instability. The magnitude of the performance value reflects the intensity and regularity of the temporal distribution of signal energy within the frequency band. When the input signal's frequency regulation is unstable, its primary frequency energy diffuses into adjacent frequency bands and periodically redistributes between them, resulting in an increased reference value. Conversely, when the signal is stable, its energy is primarily concentrated in the primary frequency band and evenly distributed over time, resulting in a smaller performance value.

[0109] The analyzed features are input into a pre-learned deep learning model, which then makes intelligent predictions on the analyzed features and identifies the signal change status.

[0110] The envelope fluctuation reference value and frequency band energy distribution reference value generated after analysis are input into a pre-learned deep learning model, and a signal change index is generated by the deep learning model. The signal change state is identified by the signal change index.

[0111] A pre-learned deep learning model refers to a model that, through sufficient training and validation with historical data, has the ability to extract signal change patterns from input features and perform state classification or prediction. This model typically undergoes multiple rounds of data processing, including data cleaning, feature engineering, model training, and tuning, with the goal of accurately identifying the dynamic changes in the input signal. The model training process typically involves large-scale historical signal data that covers different signal change scenarios (such as frequency stability, perturbation frequency drift, unstable frequency fluctuations, etc.) and contains relevant envelope fluctuation reference values and frequency band energy distribution reference values. Through supervised learning or semi-supervised learning, the model learns the mapping relationship between the characteristics of signal changes and their corresponding states. For example, through training, the model understands the correlation between abnormal increases in envelope fluctuation values and periodic signal drift, as well as the impact of dynamic changes in frequency band energy distribution on signal instability.

[0112] These deep learning models can use architectures such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), or hybrid models (such as LSTM-CNN) to extract the spatiotemporal information and pattern changes of signal features. Pre-learned models also undergo cross-validation, test set validation, and model accuracy assessment to ensure good generalization capabilities for unseen data. Therefore, when the envelope fluctuation reference values and frequency band energy distribution reference values generated after analysis are input into the model, the model can efficiently process these features and generate a signal change index associated with the input pattern for further identification of signal change states.

[0113] Pre-learned deep learning models automatically and intelligently process complex input features, providing a reliable means for identifying signal variation. In practical applications, the input envelope fluctuation reference values and frequency band energy distribution reference values can reflect the dynamic change trend of the signal within a specific time window. The deep learning model uses these reference values as feature inputs and, through its deep network structure, identifies the signal's nonlinear variation patterns and underlying periodicity, thereby generating a quantitative signal variation index. The signal variation index is the result of the model's comprehensive analysis of the input features, reflecting key attributes of the signal variation, such as amplitude, frequency, and stability, and has strong descriptive and discriminative capabilities.

[0114] Based on the signal change index, the model can further classify the current signal status into categories such as stable frequency or unstable frequency regulation. This classification process relies on the classification boundaries and prediction rules established by the model during the training phase. Combined with real-time input features, it can quickly and efficiently identify signal states. This approach not only reduces the need for manual intervention but also significantly improves the accuracy and real-time performance of state identification in complex signal environments. It is particularly suitable for monitoring and control scenarios of dynamic signal source equipment such as wind farms and distributed power sources. Through pre-learned deep learning models, the characteristics of signal changes can be quickly quantified and interpreted, providing key support for dynamic system adjustments and optimization decisions.

[0115] The deep learning model is not limited here. Any deep learning model that can perform a comprehensive analysis of the envelope fluctuation reference value EFI and the frequency band energy distribution reference value R to generate the signal change index SVDI is acceptable. To implement the technical solution of the present invention, the present invention provides a specific implementation method.

[0116] The signal variation index SVDI generation formula is as follows:

[0117]

[0118] Where s a 、s b are the preset proportional coefficients of the envelope fluctuation reference value EFI and the frequency band energy distribution reference value R, and s a 、s b Both are greater than 0.

[0119] It can be seen from the signal change index that the larger the envelope fluctuation reference value performance value generated after analyzing the amplitude change characteristics of the input signal envelope fluctuation over time in the detection window, and the larger the frequency band energy distribution reference value performance value generated after analyzing the distribution change trend of the signal energy in the frequency band over time under the detection window, it indicates that the larger the signal change index performance value generated when the analyzed features are intelligently predicted by the pre-learned deep learning model and the signal change state is identified, the greater the probability that the signal source is developing towards unstable frequency regulation and small periodic drift. Conversely, it indicates that the signal source is in a stable state.

[0120] According to the prediction results of the deep learning model, the signal change state is divided into perturbation frequency change and reference frequency stability;

[0121] The pre-learned deep learning model is used to make intelligent predictions on the analyzed features. The signal change index generated when the signal change state is identified is compared with the pre-set signal change index reference threshold to classify the signal change state. The classification steps are as follows:

[0122] If the signal change index is greater than a preset signal change index reference threshold, the signal state is classified as a perturbation frequency change;

[0123] If the signal variation index is less than or equal to a preset signal variation index reference threshold, the signal state is classified as reference frequency stability.

[0124] Perturbation frequency variation refers to the phenomenon that the frequency adjustment process of the input signal exhibits small, periodic and nonlinear fluctuations; reference frequency stability refers to the frequency of the input signal remaining near a relatively constant value during the adjustment process, and the signal change amplitude is extremely small, showing a steady-state characteristic.

[0125] For reference frequency stability, the preset bandwidth range is continued to be used to maintain a balance between dynamic response capability and noise suppression performance, ensuring high-precision tracking and low noise interference under stable signals;

[0126] Regarding reference frequency stability, the purpose of continuing to use a preset bandwidth range is to ensure that, when the signal remains stable, the phase-locked loop (PLL) can maintain high-precision tracking of the input signal's frequency and phase without introducing additional complexity, while effectively suppressing high-frequency noise and interference. The use of this fixed bandwidth range ensures that the system has sufficient dynamic response capability under steady-state signal conditions to quickly adapt to small, normal fluctuations and minimize the impact of high-frequency harmonics and random noise when the signal remains unchanged. This strategy reduces unnecessary bandwidth adjustments, lowering the system's computational burden and stability risks, thereby improving the long-term tracking accuracy of the phase-locked loop and the reliability and efficiency of system operation under stable input signal conditions.

[0127] In response to perturbation frequency changes, the bandwidth range is dynamically expanded, the response speed of the phase-locked loop is enhanced, and the frequency and phase changes of the input signal are quickly tracked to avoid the accumulation of phase offset;

[0128] The specific steps to dynamically expand the bandwidth range in response to perturbation frequency changes, enhance the response speed of the phase-locked loop, quickly track the frequency and phase changes of the input signal, and avoid the accumulation of phase offset are as follows:

[0129] When the signal is in a perturbation frequency change, the phase-locked loop needs to calculate the dynamic bandwidth adjustment coefficient based on the signal change exponential deviation to guide the dynamic expansion of the bandwidth. The adjustment coefficient is proportional to the amplitude of the signal change. That is, the more drastic the signal change, the greater the bandwidth expansion. The calculation expression is as follows:

[0130] α dyn =k1·ln(1+Δ sVDI )+k2

[0131] Where k1 is the proportional constant of the adjustment coefficient, which is used to control the sensitivity of the bandwidth extension, k2 is the basic bandwidth adjustment offset, which ensures fine-tuning even at low deviations, and α dyn is the dynamic bandwidth adjustment coefficient, which represents the proportional factor of bandwidth expansion, Δ SVDI Is the signal change index deviation, which indicates the degree of deviation of the current signal change relative to the reference threshold. The calculation expression is as follows: Δ SVDI =SVDI-SVDI ref , where SVDI is the signal variation index, SVDI ref is the signal change index reference threshold;

[0132] The dynamic bandwidth adjustment coefficient is processed using a logarithmic function, avoiding overly drastic bandwidth adjustments while quickly responding to signal amplitude changes. This coefficient is used to calculate the new bandwidth range, ensuring that the phase-locked loop can follow the frequency and phase changes of the input signal.

[0133] According to the preset bandwidth range and dynamic adjustment coefficient α dyn , calculate the new dynamic bandwidth range, which is expanded on the basis of the original bandwidth to improve the dynamic response capability of the phase-locked loop. The calculation expression is as follows:

[0134] BW dyn =BW pre ·(1+α dyn )

[0135] Where BW dyn is the adjusted bandwidth range, BW pre is the preset bandwidth range;

[0136] When the signal exponentially exceeds a reference threshold, the bandwidth is expanded proportionally to the dynamic adjustment factor. This dynamic bandwidth adjustment ensures that the phase-locked loop can quickly track small periodic drifts in the input signal, preventing the accumulation of phase errors.

[0137] After calculating the adjusted bandwidth range BW dyn Afterwards, the phase-locked loop enhances its ability to track the input signal frequency and phase through a dynamic bandwidth adjustment mechanism. At this time, the phase tracking error is recalibrated to ensure the accuracy of phase synchronization. The calculation expression is as follows:

[0138]

[0139] Where θ err，new is the adjusted phase tracking error, θ err，pre is the phase tracking error before adjustment.

[0140] This expansion of the dynamic bandwidth reduces the phase-tracking error of the phase-locked loop, ensuring precise frequency and phase synchronization despite variations in the perturbation frequency. The exponential decay model of the phase tracking error shows that the greater the bandwidth expansion, the faster the phase error decays, effectively preventing the accumulation of phase offsets.

[0141] Dynamically expanding the phase-locked loop (PLL) bandwidth in response to perturbation frequency fluctuations primarily improves the system's response to changes in the input signal's frequency and phase. This ensures that even when the signal exhibits small periodic drifts, the PLL can quickly track the dynamic changes in the input signal, preventing accumulation of phase offsets and thus maintaining stable system operation. In practical applications, perturbation frequency fluctuations are often caused by load fluctuations and environmental changes in wind farms, photovoltaic power plants, or distributed power sources. These variations are small but periodic. If the PLL's bandwidth is fixed, it may not be able to adapt to these frequency variations in a timely manner, causing the output phase to gradually deviate from the input signal's baseline. This can lead to increased reactive power, increased harmonic distortion, and malfunctioning system protection systems. Therefore, the purpose of dynamically expanding the bandwidth is to provide faster dynamic response capabilities when the signal is unstable, enabling the PLL to promptly adjust the output signal's frequency and phase to maintain synchronization with the input signal.

[0142] The core of dynamic bandwidth adjustment lies in flexibly adjusting the phase-locked loop's filtering range based on the degree of input signal variation. When the signal changes rapidly and fluctuates significantly, expanding the bandwidth allows the phase-locked loop to capture more details of the input signal changes, allowing it to quickly lock onto the new phase reference and avoid cumulative phase errors. Dynamic bandwidth adjustment also effectively reduces phase tracking lag, mitigates system risks caused by frequency offsets, and improves the robustness and adaptability of the phase-locked loop in complex dynamic environments.

[0143] This process enables the phase-locked loop to not only quickly respond to subtle signal changes but also achieve high-precision frequency tracking and low-noise interference suppression in dynamically changing environments, thereby improving the overall performance and system stability of grid-connected equipment. This intelligent dynamic bandwidth adjustment mechanism can better meet the high-dynamic frequency tracking requirements of distributed power and renewable energy systems, effectively avoiding the limitations of traditional fixed-bandwidth designs and enabling reliable system operation even in the face of large frequency fluctuations.

[0144] The present invention achieves accurate initial tracking of the input signal through a preset bandwidth range, and combines real-time signal feature extraction and deep learning model prediction to intelligently identify the signal change state. When the signal is in a reference frequency stability state, the system maintains the preset bandwidth range to balance the dynamic response capability and noise suppression performance; when a perturbation frequency change is detected, the system dynamically expands the bandwidth range to quickly track frequency and phase changes. This adaptive adjustment mechanism enables the phase-locked loop to cope with frequency regulation instability and small periodic drift problems that occur at the signal source (such as a wind farm or distributed power source), avoids phase offset and tracking lag caused by a fixed bandwidth design, and thus improves the tracking accuracy and response speed of the inverter.

[0145] The present invention extracts the key features of the input signal envelope fluctuation and frequency band energy distribution, and generates envelope fluctuation reference values and frequency band energy distribution reference values. The phase-locked loop can accurately identify the signal change trend and predict the signal change index through a deep learning model. When it is detected that the signal frequency regulation is unstable, the system can quickly and dynamically adjust the bandwidth to avoid the accumulation of phase offsets, reduce the phase difference between the inverter output current and the grid signal, and reduce reactive power fluctuations. At the same time, this dynamic adjustment mechanism effectively reduces harmonic distortion, improves the quality of power, and avoids the occurrence of system protection misoperation, thereby improving the operating reliability of the inverter and the power supply stability of the power grid, and reducing the risk of large-scale power outages or equipment failures.

[0146] 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.

[0147] 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.

[0148] 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.

[0149] 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.

[0150] 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.

[0151] 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.

[0152] 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.

[0153] 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.

[0154] 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.

[0155] 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. The inverter fast phase locking method based on cordic algorithm is characterized by: The following steps are involved: Balance dynamic response and high-frequency noise suppression capabilities through a preset bandwidth range to achieve accurate initial tracking of input signal frequency and phase; During the signal tracking process, parameter information of the input signal is obtained in real time to provide basic data for subsequent analysis and decision-making; Preprocess the signal parameters acquired in real time and extract key features from the preprocessed signal parameters that reflect the unstable frequency regulation of the signal source and the occurrence of small periodic drift trends. Analyze the extracted features within the detection window to quantify the change trend and amplitude of the input signal. The analyzed features are input into a pre-learned deep learning model, which then makes intelligent predictions on the analyzed features and identifies the signal change status. According to the prediction results of the deep learning model, the signal change state is divided into perturbation frequency change and reference frequency stability; For reference frequency stability, the preset bandwidth range is continued to be used to maintain a balance between dynamic response capability and noise suppression performance, ensuring high-precision tracking and low noise interference under stable signals; In response to the perturbation frequency changes, the bandwidth range is dynamically expanded, the response speed of the phase-locked loop is enhanced, the frequency and phase changes of the input signal are quickly tracked, and the accumulation of phase offset is avoided.

2. The inverter fast phase locking method based on the cordic algorithm according to claim 1, characterized in that: Key features reflecting the unstable frequency regulation of the signal source and the trend of small periodic drift are extracted from the preprocessed signal parameters. The extracted features include the amplitude variation characteristics of the input signal envelope fluctuation over time and the distribution variation trend of the signal energy in the frequency band over time. Within the detection window, after analyzing the extracted amplitude variation characteristics of the input signal envelope fluctuation over time and the distribution variation trend of the signal energy in the frequency band over time, an envelope fluctuation reference value and a frequency band energy distribution reference value are generated respectively. The variation trend and amplitude of the input signal are quantified by the envelope fluctuation reference value and the frequency band energy distribution reference value.

3. The inverter fast phase locking method based on the cordic algorithm according to claim 2, characterized in that: The specific steps for analyzing the amplitude variation characteristics of the input signal envelope fluctuation over time in the detection window to generate the envelope fluctuation reference value are as follows: First, the instantaneous envelope of the input signal is constructed using its instantaneous information. The instantaneous envelope is a key characteristic that reflects the dynamic changes in the input signal amplitude. The formula is as follows: Where x(n) is the input signal, H[x(n)] is the Hilbert transform of the input signal, and E(n) is the instantaneous envelope; Extract the fluctuation rate of the signal envelope and calculate the intensity and trend of the envelope over time by taking the derivative of the instantaneous envelope E(n). The fluctuation rate calculation expression is as follows: Where, is the first derivative of the instantaneous envelope, is the second derivative of the instantaneous envelope, γ is the adjustment factor, and R(n) is the envelope fluctuation rate; The fluctuation intensity factor is calculated by the envelope fluctuation rate R(n) and the instantaneous envelope E(n). The calculation expression is as follows: Where sin(α·n) is the sinusoidal modulation term, α is the modulation frequency factor, n is the sequence index of the discrete signal, N is the total number of detection window time points, and F is the fluctuation intensity factor; The instantaneous envelope E(n), envelope fluctuation rate R(n) and fluctuation intensity factor F are combined to generate the envelope fluctuation reference value. The generation formula is as follows: Where η is the weight factor, β is the adjustment coefficient, and EFI is the envelope fluctuation reference value.

4. The inverter fast phase locking method based on the cordic algorithm according to claim 2, characterized in that: The specific steps for analyzing the distribution trend of signal energy within the frequency band over time in the detection window to generate the frequency band energy distribution reference value are as follows: First, the input signal is decomposed into multiple frequency bands and the instantaneous energy distribution of each frequency band is calculated. By accurately decomposing the frequency domain characteristics of the signal, the energy concentration or diffusion of the signal in different frequency ranges is captured. The calculation expression is as follows: Where, E k represents the energy value of the kth frequency band, f k，1 is the starting frequency of the kth frequency band, f k，2 is the end frequency of the kth frequency band, S(f) is the power spectral density of the signal, f is the frequency, W k (f) is the weighting function of the kth frequency band; In each frequency band, the dynamic change characteristics of the signal energy over time are extracted to generate a characteristic vector for quantifying the energy fluctuation within the frequency band. The calculation expression is as follows: Where, E k，i is the energy value of the kth frequency band at the i-th detection point, E k，i+1 is the energy value of the kth frequency band at the i+1th detection point, that is, the energy value of the previous detection point, N k is the total number of detection points in the kth frequency band, g k (i) is the nonlinear weighting factor of the kth frequency band, D k is the dynamic change characteristic of the kth frequency band; Perform pattern analysis on the dynamic change characteristics within each frequency band, integrate the energy distribution and change trend, and generate a pattern value that characterizes the energy distribution characteristics of the frequency band. The generation formula is as follows: Where, P k is the distribution mode value, indicating the distribution mode value of the kth frequency band, F k (E k , D k ) is the characteristic function, which is the energy value of the combined frequency band E k and dynamic change characteristics D k The characteristic function, H k (f) is the intra-band weighting function; The distribution mode values of all frequency bands are combined to generate the overall frequency band energy distribution reference value. The generation formula is as follows: Where K is the total number of frequency bands, ω k is the balance factor for the kth frequency band, is the nonlinear weight factor for the kth frequency band, γ k is the normalized weight factor of the kth frequency band, and R is the reference value of the frequency band energy distribution.

5. The inverter fast phase locking method based on the cordic algorithm according to claim 2, characterized in that: The envelope fluctuation reference value and frequency band energy distribution reference value generated after analysis are input into a pre-learned deep learning model, and a signal change index is generated by the deep learning model. The signal change state is identified by the signal change index.

6. The inverter fast phase locking method based on the cordic algorithm according to claim 5, characterized in that: The pre-learned deep learning model is used to make intelligent predictions on the analyzed features. The signal change index generated when the signal change state is identified is compared with the pre-set signal change index reference threshold to classify the signal change state. The classification steps are as follows: If the signal change index is greater than a preset signal change index reference threshold, the signal state is classified as a perturbation frequency change; If the signal variation index is less than or equal to a preset signal variation index reference threshold, the signal state is classified as reference frequency stability.

7. The inverter fast phase locking method based on the cordic algorithm according to claim 6, characterized in that: The specific steps to dynamically expand the bandwidth range in response to perturbation frequency changes, enhance the response speed of the phase-locked loop, quickly track the frequency and phase changes of the input signal, and avoid the accumulation of phase offset are as follows: When the signal is in a perturbation frequency change, the phase-locked loop needs to calculate the dynamic bandwidth adjustment coefficient based on the signal change index deviation to guide the dynamic expansion of the bandwidth. The calculation expression is as follows: a dyn =k1·ln(1+Δ SVDI )+k2 Where k1 is the proportional constant of the adjustment coefficient, k2 is the basic bandwidth adjustment offset, and α dyn is the dynamic bandwidth adjustment coefficient, Δ SVDI is the signal change index deviation, and the calculation expression is as follows: Δ SVDI =SVDI-SVDI ref , where SVDI is the signal variation index, SVDI ref is the signal change index reference threshold; According to the preset bandwidth range and dynamic adjustment coefficient α dyn , calculate the new dynamic bandwidth range to improve the dynamic response capability of the phase-locked loop. The calculation expression is as follows: BW dyn =BW pre ·(1+a dyn ) Where BW dyn is the adjusted bandwidth range, BW pre is the preset bandwidth range; After calculating the adjusted bandwidth range BW dyn Afterwards, the phase-locked loop enhances its ability to track the input signal frequency and phase through a dynamic bandwidth adjustment mechanism. At this time, the phase tracking error is recalibrated to ensure the accuracy of phase synchronization. The calculation expression is as follows: Where θ err，new is the adjusted phase tracking error, θ err，pre is the phase tracking error before adjustment.

Images