Power quality disturbance identification method and system based on time-frequency joint analysis

Abstract

This invention discloses a method and system for identifying power quality disturbances using time-frequency joint analysis, belonging to the field of power system power quality monitoring and analysis technology. The method includes the following steps: synchronously sampling voltage or current signals in the power system under test to obtain raw time-domain sampled signals; performing time-frequency joint analysis on the preprocessed signals to obtain corresponding instantaneous frequency information; calculating the instantaneous frequency trajectory of the disturbance signal and performing second-order change analysis on the instantaneous frequency trajectory to generate instantaneous frequency change curvature parameters; calculating the energy temporal change curvature based on the evolution information of the disturbance energy over time to determine the start and end times of the power quality disturbance event; constructing a feature set within the time range defined by the start and end times to identify the power quality disturbance and output the disturbance type and its corresponding time location parameters. This invention can achieve high-precision identification and location of complex power quality disturbances.

Description

Technical Field

[0001] This invention relates to the field of power system power quality monitoring and analysis technology, and in particular to a power quality disturbance identification method and system based on time-frequency joint analysis. Background Technology

[0002] With the large-scale grid connection of renewable energy and the widespread integration of power electronic equipment, power quality disturbances in modern power systems are becoming increasingly frequent and complex. These disturbances, such as voltage sags, voltage rises, harmonics, interharmonics, and oscillations, typically exhibit transient, non-stationary, and multi-component superposition characteristics. Accurate and rapid automatic identification and location of these disturbances are a crucial technical foundation for ensuring the safe and stable operation of the power grid and providing users with high-quality power supply.

[0003] Currently, most methods for identifying power quality disturbances are based on time-frequency analysis techniques, such as short-time Fourier transform and wavelet transform. These methods typically decompose the signal into time-frequency components, extracting energy, amplitude, or statistical features within a specific frequency band, and then combining this with threshold judgment or statistical feature-based classification methods to identify the disturbance type. Some schemes also introduce data-driven classification models for auxiliary discrimination. However, these methods still have certain limitations in practical applications.

[0004] On the one hand, due to the inherent trade-off between time-domain and frequency-domain resolution, traditional time-frequency analysis methods often result in blurred energy diffusion and indistinct contours in engineering applications, especially when the disturbance frequency changes rapidly or multiple components are closely superimposed, making it difficult to accurately characterize the instantaneous frequency trajectory of each disturbance component. On the other hand, the features extracted by existing methods are mostly concentrated on low-order static features such as amplitude, energy, or fixed-band power, lacking effective quantitative descriptions of the dynamic process of energy and frequency changes over time during disturbance occurrence, and failing to reflect the rate of change and trend inflection points in the disturbance evolution process. This can easily lead to decreased recognition accuracy in scenarios with significant dynamic frequency changes, complex superposition of disturbances, or blurred disturbance boundaries, resulting in problems such as misjudgment of disturbance type, inaccurate location of start and end times, and difficulty in effectively separating complex disturbances. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a power quality disturbance identification method based on time-frequency joint analysis, comprising the following steps:

[0007] The voltage or current signal in the power system under test is synchronously sampled to obtain the original time-domain sampled signal;

[0008] The original time-domain sampled signal is preprocessed, and the preprocessing includes performing adaptive fundamental notch filtering on the fundamental component of the signal;

[0009] Perform time-frequency joint analysis on the preprocessed signal to obtain the energy distribution of the preprocessed signal in the time-frequency plane and the evolution information of the disturbance energy derived from the energy distribution over time, and simultaneously obtain the corresponding instantaneous frequency information.

[0010] Based on the instantaneous frequency information, the instantaneous frequency trajectory of the disturbance signal is calculated, and a second-order change analysis is performed on the instantaneous frequency trajectory to generate an instantaneous frequency change curvature parameter that characterizes the evolution of the disturbance frequency.

[0011] Based on the evolution information of the disturbance energy over time, the curvature of the energy time series change is calculated, and the start and end times of the power quality disturbance event are determined based on the abrupt change characteristics of the curvature of the energy time series change.

[0012] Within the time range defined by the start and end times, a feature set for characterizing the disturbance pattern is constructed based on the instantaneous frequency change curvature parameter, and the power quality disturbance is identified according to the feature set, outputting the disturbance type and its corresponding time location parameter.

[0013] As a preferred embodiment of the power quality disturbance identification method of time-frequency joint analysis described in this invention, wherein: the time-frequency joint analysis adopts a synchronous compression-type time-frequency transformation method;

[0014] The synchronous compression process redistributes energy in the initial time-frequency representation, concentrating the energy along the instantaneous frequency trajectory.

[0015] Furthermore, during synchronous compression, the instantaneous frequency is calculated based on the phase change rate of the time-frequency coefficient.

[0016] As a preferred embodiment of the power quality disturbance identification method of time-frequency joint analysis described in this invention, the instantaneous frequency change curvature parameter is obtained by calculating at least the second derivative of the instantaneous frequency trajectory, wherein the first derivative is used to characterize the frequency drift rate and the second derivative is used to characterize the degree of bending in the direction of frequency change.

[0017] When generating the instantaneous frequency change curvature parameter, the instantaneous frequency trajectories corresponding to different frequency components are first time-aligned, and the amplitude of each component is normalized according to the frequency change amplitude during the disturbance duration. Then, the corresponding change curvature parameter is calculated based on the aligned and normalized instantaneous frequency trajectory.

[0018] Furthermore, statistical analysis is performed on the sequence of instantaneous frequency change curvature parameters during the duration of the disturbance, extracting at least one or more of the following: mean, peak value, range of variation, and standard deviation.

[0019] As a preferred embodiment of the power quality disturbance identification method based on time-frequency joint analysis of the present invention, the determination of the start and end times based on the abrupt change characteristics of the energy time-series change curvature includes:

[0020] Based on the evolution information of the disturbance energy over time, an energy time series is generated;

[0021] The energy time series is smoothed.

[0022] Numerical rate of change calculation is performed on the smoothed energy time series to obtain an energy change curvature sequence for characterizing the temporal curvature of the perturbation energy;

[0023] Local extreme points in the energy change curvature sequence that exceed an adaptive threshold are detected and used as boundary candidate points; wherein, the adaptive threshold is generated based on the energy fluctuation level of the background steady-state signal before the disturbance occurs;

[0024] Temporal continuity and persistence constraints are applied to the candidate boundary points to determine the start and end times.

[0025] As a preferred embodiment of the power quality disturbance identification method based on time-frequency joint analysis described in this invention, wherein: a feature set for characterizing the disturbance morphology is constructed based on the instantaneous frequency change curvature parameter, including:

[0026] An initial feature vector is generated based on the instantaneous frequency change curvature parameter corresponding to each frequency component.

[0027] The instantaneous frequency change curvature parameters corresponding to each frequency component during the duration of the disturbance are processed to optimize the initial feature vector;

[0028] Furthermore, the optimized initial eigenvectors are weighted and fused based on the perturbation energy ratio or frequency stability index obtained from the time-frequency joint analysis of each frequency component.

[0029] As a preferred embodiment of the power quality disturbance identification method based on time-frequency joint analysis described in this invention, the method further includes:

[0030] The three-phase voltage or current signals acquired synchronously with the voltage or current signals are preprocessed respectively. The preprocessing includes performing adaptive fundamental notch filtering on the fundamental component of the signal to weaken the fundamental component.

[0031] Perform time-frequency joint analysis on the preprocessed three-phase signals respectively, and calculate the instantaneous phase difference between each phase at the corresponding time-frequency position;

[0032] Based on the instantaneous phase difference, a deviation index reflecting the consistency of the three-phase time-frequency phase is calculated to characterize the degree of consistency of the disturbance characteristics between different phases.

[0033] When the deviation of the phase consistency is lower than a preset threshold, the identified power quality disturbance is determined to be a systematic disturbance; when the deviation of the phase consistency is higher than the preset threshold, it is determined to be a local or asymmetric disturbance.

[0034] The determination results of the systemic disturbance, local disturbance, or asymmetric disturbance are associated with the disturbance type and time location parameters and output.

[0035] As a preferred embodiment of the power quality disturbance identification method of the time-frequency joint analysis described in this invention, the deviation index reflecting the consistency of the three-phase time-frequency phase is obtained by statistical analysis of the instantaneous phase of each corresponding frequency component, and the statistical analysis includes at least one or more of the following: mean phase difference, phase difference fluctuation range, or phase difference stability index.

[0036] As a preferred embodiment of the power quality disturbance identification method based on time-frequency joint analysis described in this invention, the method further includes evaluating the measurement uncertainty of the disturbance identification result;

[0037] The measurement uncertainty is calculated comprehensively based at least on the sampling noise level, the transformation stability of the time-frequency joint analysis, the fluctuation of the instantaneous frequency change curvature parameter, and the consistency level of the instantaneous phase difference between each phase when identifying three-phase signals.

[0038] The measurement uncertainty is then output in association with the disturbance type and time location parameters.

[0039] As a preferred embodiment of the power quality disturbance identification method based on time-frequency joint analysis described in this invention, the method further includes:

[0040] The output disturbance type and time location parameters are mapped to the corresponding power quality evaluation parameters;

[0041] The power quality evaluation parameters include at least the voltage sag depth, disturbance duration, or harmonic impact level.

[0042] This invention also provides a power quality disturbance identification system based on time-frequency joint analysis. This system is applied to the aforementioned power quality disturbance identification method based on time-frequency joint analysis, and includes:

[0043] The synchronous sampling module is used to synchronously sample voltage or current signals in a power system.

[0044] The preprocessing module is used to perform adaptive fundamental notch filtering on the sampled signal;

[0045] The time-frequency analysis module is used to perform joint time-frequency analysis and calculate instantaneous frequency trajectories;

[0046] The feature generation module is used to calculate the instantaneous frequency change curvature parameter and the energy temporal change curvature, and to construct a feature set to characterize the perturbation morphology;

[0047] The disturbance identification module is used to identify the type of power quality disturbance based on the feature set, and output the disturbance type and its corresponding time and location parameters.

[0048] The beneficial effects of this invention are:

[0049] 1. This invention improves the characterization and discrimination of time-frequency perturbations with time-varying frequencies and multiple superimposed components on the time-frequency plane by employing synchronous compression-type time-frequency transform and extracting two types of high-order dynamic features: instantaneous frequency change curvature and energy temporal change curvature. Specifically, the synchronous compression process overcomes the energy diffusion problem of traditional time-frequency analysis, obtaining a high-resolution time-frequency representation and instantaneous frequency trajectory; furthermore, by taking the second derivative of the frequency trajectory to obtain the "change curvature" and taking the second rate of change of the energy evolution sequence to obtain the "temporal change curvature," the identification features are elevated from static amplitude and frequency points to high-order temporal features that can characterize the dynamic evolution rate of the perturbation and its inflection points. This allows different perturbations that may be similar in the spectrum (such as harmonics and interharmonics, sags and oscillations) to be effectively distinguished in terms of their dynamic evolution morphology.

[0050] 2. This invention achieves high-precision, adaptive positioning of the start and end times of disturbance events by detecting abrupt changes based on the curvature of energy temporal variations, combined with adaptive thresholds and physical constraint rules. This enhances robustness to complex operating conditions and noisy environments. Compared to traditional criteria that rely on fixed amplitude thresholds, this invention captures abrupt change inflection points by analyzing the curvature (second-order rate of change) of the energy evolution trend. It also dynamically generates a judgment threshold based on the fluctuation level of the background steady-state signal before the disturbance, effectively avoiding misjudgments and missed judgments caused by load fluctuations or noise interference. This provides a precise time window for subsequent feature extraction and recognition.

[0051] 3. This invention constructs a feature set that integrates the proportion of disturbance energy and the weight of frequency stability, and extends this to three-phase phase consistency analysis and measurement uncertainty assessment. This forms a complete and reliable power quality disturbance diagnosis system, encompassing signal processing, feature extraction, disturbance identification, and result evaluation. The system can not only output the type and precise time and location of the disturbance, but also determine whether it is a systematic or local disturbance. It provides quantitative indicators of measurement uncertainty, ultimately mapping these to standard power quality evaluation parameters. This provides power system operators with a highly reliable diagnostic report that offers comprehensive decision support information and can be directly used for industry assessments. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0053] Figure 1 This is a flowchart illustrating the overall process of the power quality disturbance identification method based on time-frequency joint analysis of the present invention.

[0054] Figure 2 This diagram illustrates the principle of synchronous compression transformation and the comparison of time and frequency planes in the power quality disturbance identification method of the present invention, which is based on time-frequency joint analysis.

[0055] Figure 3 This is a schematic diagram of the instantaneous frequency trajectory and its curvature calculation in the power quality disturbance identification method of the time-frequency joint analysis of the present invention (taking a sudden drop in frequency as an example).

[0056] Figure 4 This is a schematic diagram of the process for detecting the start and end times of disturbances based on the curvature of energy temporal changes in the power quality disturbance identification method of the present invention, which is based on time-frequency joint analysis.

[0057] Figure 5 This is a flowchart of the three-phase voltage signal disturbance identification and systematic / local determination in the power quality disturbance identification method of the time-frequency joint analysis of the present invention. Detailed Implementation

[0058] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0059] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0060] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0061] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0062] Example 1

[0063] Reference Figure 1-4 The first embodiment of the present invention provides a power quality disturbance identification method based on time-frequency joint analysis, comprising the following steps:

[0064] S1. Synchronously sample the voltage or current signal in the power system under test to obtain the original time-domain sampled signal.

[0065] Among them, synchronous sampling uses a sampling device (such as a synchronous acquisition card) that is synchronized with the power grid frequency clock to continuously sample the voltage or current signal in the power grid at equal intervals with a fixed sampling frequency to obtain the original discrete time domain sampling signal; wherein, the fixed sampling frequency is set to an integer multiple of the system power frequency, for example, 6.4kHz, which corresponds to 128 times the 50Hz power frequency) to ensure that a constant number of sample points are collected in each power frequency cycle.

[0066] It should be noted that the purpose of step S1 is to obtain the original time-domain signal that is strictly synchronized with the fundamental period of the power grid, so as to provide a high-fidelity, phase-distortion-free data foundation for all subsequent analysis and processing steps.

[0067] S2. Preprocess the original time-domain sampled signal. The preprocessing includes performing adaptive fundamental notch filtering on the fundamental component of the signal.

[0068] The purpose of this step is to adaptively filter out the dominant power frequency fundamental component in the original signal, thereby highlighting the weaker but crucial non-fundamental components related to power quality disturbances, providing a signal with a better signal-to-noise ratio for subsequent time-frequency analysis.

[0069] The process of performing adaptive fundamental notch filtering on the fundamental component of the original time-domain sampled signal includes the following sub-steps:

[0070] S21: Fundamental Frequency Tracking. This step provides a frequency reference for the entire filtering process. Specifically, a phase-locked loop or a prediction-based frequency estimation method is used to estimate and track the fundamental frequency of the input signal in real time and accurately. The estimation and tracking of the fundamental frequency reflects any slight deviations in the power frequency that may occur during power system operation (e.g., variations within the range of 49.8 Hz to 50.2 Hz), ensuring that the center frequency of the subsequent notch filter remains consistent with the actual system fundamental frequency. This avoids the problem of insufficient or false fundamental frequency suppression caused by frequency deviation when using a fixed nominal frequency (e.g., 50 Hz).

[0071] S22: Notch filter design based on tracking frequency. This sub-step is based on the fundamental frequency estimated in real time in step S21. A digital notch filter is dynamically generated. In practice, the filter can be of the infinite impulse response or finite impulse response type, and its transfer function is configured as follows: [The text abruptly ends here, likely due to an incomplete sentence or missing information.] A very narrow stopband centered on (e.g., Within this stopband, it exhibits a significant attenuation depth (e.g., -40 dB or lower); while outside this stopband, it maintains a flat amplitude-frequency response. This design ensures maximum suppression of the time-varying fundamental component while minimizing distortion or attenuation of nearby low-order harmonics or interharmonics that are closely related to power quality disturbances.

[0072] S23: Filtering Execution. This sub-step is the final execution stage of the filtering process. The time-domain sampled signal is passed through a dynamically generated digital notch filter, and convolution or recursive operations are performed to filter out the fundamental frequency component located in the stopband in real time, resulting in a preprocessed signal. In the preprocessed signal, the fundamental frequency energy is significantly weakened, while various non-fundamental frequency components related to power quality disturbances (such as harmonics, interharmonics, and transient oscillations) are clearly highlighted in the time-domain waveform, creating conditions for subsequent high-precision time-frequency joint analysis.

[0073] It should be noted that the purpose of adaptive fundamental notch filtering is to weaken, rather than completely eliminate, the fundamental component. By setting an appropriate notch depth (e.g., -40dB), the fundamental component retains a detectable energy level after preprocessing. The significance of this design is that for voltage sag / rise disturbances characterized by fundamental amplitude changes, the energy evolution information of the disturbance is mainly reflected in the fundamental frequency component. After this step, the fundamental frequency component, as a special frequency component in time-frequency analysis, is incorporated into the analysis framework of step S3, enabling the abrupt change feature detection based on the curvature of energy time-series changes in step S5 to also be applied to the energy changes of the fundamental frequency, thereby achieving accurate detection of the start and end times of voltage sag / rise disturbances.

[0074] Therefore, through the above operations, the preprocessing of the original time-domain sampled signal is completed, resulting in a preprocessed signal with effectively suppressed fundamental background and relatively prominent perturbation components. Based on this preprocessed signal, performing the time-frequency joint analysis in step S3 can accurately extract the dynamic energy distribution and instantaneous frequency information of the perturbation in the time-frequency plane.

[0075] S3. Perform time-frequency joint analysis on the preprocessed signal to obtain the energy distribution of the preprocessed signal in the time-frequency plane and the evolution information of the disturbance energy derived from the energy distribution over time, and simultaneously obtain the corresponding instantaneous frequency information.

[0076] This step is the core analytical step of the invention. Its purpose is to perform high-resolution time-frequency transformation on the preprocessed signal, taking into account the non-stationary and multi-component characteristics of power quality disturbance signals. This allows for the simultaneous acquisition of the fine distribution of disturbance energy in the time-frequency plane, its trajectory over time, and instantaneous frequency information reflecting local frequency changes in the signal. This information forms the basis for subsequent extraction of deep dynamic features such as "changing curvature" to achieve accurate disturbance identification.

[0077] Specifically, the time-frequency joint analysis adopts a synchronous compression-type time-frequency transform method;

[0078] Among them, the synchronous compression process redistributes the energy in the initial time-frequency representation, so that the energy is concentrated along the instantaneous frequency trajectory;

[0079] Furthermore, during synchronous compression, the instantaneous frequency is calculated based on the phase change rate of the time-frequency coefficient.

[0080] In one specific implementation, time-frequency analysis is performed using a synchronous compression-type time-frequency transform method, which specifically includes the following sub-steps:

[0081] S31: Calculate the initial time-frequency representation.

[0082] The purpose of this step is to obtain the initial time-frequency representation of the preprocessed signal as the input for synchronous compression transformation.

[0083] Specifically, the Short-Time Fourier Transform (STFT) is used as the basic method for generating the initial time-frequency representation. STFT obtains an initial time-frequency coefficient matrix by windowing the signal and calculating the Discrete Fourier Transform frame by frame. This is the initial time-frequency representation. Preferably, to balance the time positioning accuracy of transient disturbances (such as voltage drops) and the frequency resolution of harmonic components, this embodiment uses a Hamming window, with the window length set to an integer multiple of the power frequency period (such as 2 cycles) and the overlap rate set to 50%. This initial time-frequency coefficient matrix will be directly used for the instantaneous frequency estimation in step S32.

[0084] S32: Estimate instantaneous frequency.

[0085] The purpose of this step is to provide a precise redistribution direction for synchronous compression based on the initial time-frequency coefficient matrix.

[0086] Specifically, for the initial time-frequency coefficient matrix For each time-frequency point, the instantaneous frequency estimate is calculated based on the phase change rate of its time-frequency coefficients. A specific implementation calculation formula is as follows:

[0087] ;

[0088] in, This represents the phase angle of a complex number. In practical numerical calculations, the phase difference is often used to approximate this partial derivative. The calculated instantaneous frequency field will serve as the mapping basis for energy redistribution in step S33.

[0089] S33: Energy redistribution (compression).

[0090] The purpose of this step is to concentrate the diffused energy in the initial time-frequency representation onto the true instantaneous frequency trajectory, thereby obtaining a high-resolution final time-frequency representation, i.e., the synchronous compressed transform spectrum. .

[0091] Specifically, based on the instantaneous frequency obtained in step S32 For the initial time-frequency representation Perform energy redistribution. In one specific implementation, the following redistribution algorithm is executed: for each discrete frequency point... Retrieve all results that meet the conditions The original time and frequency points ,in, This is a preset frequency tolerance parameter. Then, the corresponding time-frequency points are... The values ​​are summed together, and the sum is assigned to the value. Through this algorithm, The energy diffusing along the frequency axis is strictly compressed and concentrated at its estimated true instantaneous frequency position, ultimately forming a highly concentrated energy with a clear time-frequency ridge. This provides a high-fidelity data foundation for the subsequent accurate extraction of perturbation features.

[0092] S34: Extract disturbance evolution information and instantaneous frequency information.

[0093] The purpose of this step is to obtain the synchronous compression transform spectrum. In this process, basic information is extracted for subsequent generation and identification of power quality disturbance features.

[0094] Specifically, from Extract the following two core pieces of information from the text:

[0095] Perturbation energy evolution information: By summarizing the energy in the frequency dimension of the high-resolution final time-frequency representation, an energy evolution sequence reflecting the temporal changes in perturbation intensity is obtained. It is used to characterize the overall energy change trend of a disturbance during its occurrence, development, and dissipation.

[0096] Instantaneous frequency information: for For each significant perturbation component in the medium energy range (obtainable by finding local energy peaks at each time t), record the frequency value corresponding to the peak value of each component, which constitutes the first... Instantaneous frequency trajectory of each disturbance component .

[0097] It should be noted that step S3, by employing a synchronous compression-type time-frequency transform, successfully overcomes the inherent trade-offs in time-frequency resolution inherent in traditional STFTs or wavelet transforms. Its output synchronous compression transform spectrum... It has a time-frequency ridge with highly concentrated energy, which enables the synchronous and precise extraction of two types of key information: one is the energy evolution sequence reflecting the overall intensity evolution of the disturbance. Secondly, it characterizes the instantaneous frequency trajectories of the subtle dynamics of each disturbance component. This lays a high-quality data foundation for subsequent steps to delve deeper into the higher-order dynamic characteristics (curvature variation) of the perturbation signal from the two dimensions of "energy" and "frequency".

[0098] S4. Based on instantaneous frequency information, calculate the instantaneous frequency trajectory of the disturbance signal, and perform second-order change analysis on the instantaneous frequency trajectory to generate instantaneous frequency change curvature parameters to characterize the evolution of the disturbance frequency.

[0099] The purpose of this step is to: extract the instantaneous frequency trajectories from step S3, which reflect the dynamic behavior of each power quality disturbance component. Further mathematical analysis is then conducted. By calculating its rate of change (first derivative) and acceleration (second derivative), the curve describing the frequency change over time is transformed into a series of high-order dynamic features that can quantify its "speed of change" and "curvature." These high-order features extracted from the frequency evolution process can distinguish different types of disturbances more precisely than the frequency value itself (for example, distinguishing between slow frequency changes caused by a smooth voltage drop and rapid frequency fluctuations caused by oscillating transients), providing a crucial basis for the accurate identification of subsequent disturbance types.

[0100] Specifically, the instantaneous frequency change curvature parameter is obtained by calculating at least the second derivative of the instantaneous frequency trajectory, where the first derivative is used to characterize the frequency drift rate and the second derivative is used to characterize the degree of bending in the direction of frequency change.

[0101] When generating instantaneous frequency change curvature parameters, the instantaneous frequency trajectories corresponding to different frequency components are first time-aligned, and the amplitude of each component is normalized according to the frequency change amplitude during the disturbance duration. Then, the corresponding change curvature parameters are calculated based on the aligned and normalized instantaneous frequency trajectories.

[0102] Furthermore, statistical analysis is performed on the sequence of instantaneous frequency change curvature parameters during the duration of the disturbance, extracting at least one or more of the following indices: mean, peak value, range of variation, and standard deviation.

[0103] It should be noted that, during time alignment, to eliminate the small time differences at the start of different disturbance components, the initial reference point is the moment when the energy of each frequency component first significantly exceeds the steady-state level. After step S5 determines the precise start and end times of the disturbance, only the sequence of changing curvature parameters within the time range defined by the start and end times is truncated and statistically analyzed.

[0104] In one specific implementation, the first derivative is calculated using the following formula: The sequence It characterizes the instantaneous drift rate (unit can be Hz / s) of the frequency component of the disturbance at each moment, and its absolute value directly reflects the severity of frequency change during the power disturbance.

[0105] In one specific implementation, the second derivative is calculated using the following formula: The sequence It characterizes the instantaneous change in frequency drift rate, i.e. the acceleration of frequency change, and can keenly capture the local bending, turning or oscillation pattern of frequency trajectory caused by specific types of power quality events (such as motor starting, arc fault).

[0106] In a preferred embodiment, to obtain the dimensionless curvature that is more directly related to the trajectory geometry, the following normalization formula can be used for calculation: The calculated sequence of changing curvature or its normalized form , which is the instantaneous frequency change curvature parameter used to characterize the evolution of the perturbation frequency.

[0107] In one specific implementation, the statistical analysis is calculated as follows:

[0108] Mean: Calculated Arithmetic mean over the duration of the disturbance This is used to reflect the average bending trend and direction of the frequency component throughout the entire disturbance event;

[0109] Peak value: Calculation Maximum absolute value It is used to characterize the most severe instantaneous bending or turning intensity experienced by the frequency trajectory of this component;

[0110] Range of variation: calculation Difference between maximum and minimum values This is used to describe the overall span of the curvature change of this component;

[0111] Standard deviation: Calculation Standard deviation It is used to measure the stationarity or volatility of frequency evolution.

[0112] In a preferred embodiment, multiple statistical features can be extracted simultaneously to form a feature combination. For example, for a dominant perturbation frequency component, the mean of its curvature can be extracted. Peak and standard deviation This constitutes a feature combination. These quantified features extracted from the frequency dynamic evolution unique to power quality disturbances will be combined with the features extracted from energy evolution in the subsequent step S5 to form a complete disturbance description, which will then be input into the final identification module.

[0113] It should be noted that the key to implementing step S4 lies in introducing the curvature of the instantaneous frequency trajectory as a perturbation characteristic parameter. Specifically, this is achieved by analyzing the instantaneous frequency trajectory obtained in step S3. By performing at least second-order derivative operations, the instantaneous frequency information, which originally only reflects the trend of frequency change over time, is transformed into a sequence of changing curvature parameters that can characterize the rate of frequency change and the degree of bending. Through the above processing, curvature features, including mean, peak value, range of variation, and stability indicators, are extracted. Compared to methods that only describe based on spectral distribution or instantaneous frequency values, these curvature features can reflect the rate and pattern of frequency changes during disturbances. This allows disturbances caused by different physical mechanisms but exhibiting similar spectral characteristics to show distinguishable differences in curvature features (for example, disturbances caused by motor startup typically show a slow frequency drop or rise, with a small and continuous curvature; while disturbances caused by the switching action of power electronic devices show rapid frequency oscillations, with a large and drastic curvature amplitude). This provides a more stable and physically meaningful feature basis for subsequent disturbance type identification.

[0114] S5. Based on the evolution information of the disturbance energy over time, calculate the curvature of the energy time series change, and determine the start and end times of the power quality disturbance event based on the abrupt change characteristics of the curvature of the energy time series change.

[0115] It should be noted that the purpose of this step is to adaptively and accurately locate the start and end times of the disturbance event by using the disturbance energy evolution information obtained in step S3 and analyzing the curvature (second-order rate of change) characteristics of its changing trend. Compared with traditional criteria that rely on fixed amplitude thresholds, the detection method based on trend inflection points in this invention has stronger robustness to background noise and load fluctuations, and can effectively avoid misjudgments caused by slow energy fluctuations or noise spikes. It is especially suitable for the clear separation of the boundaries of complex disturbance events.

[0116] Specifically, determining the start and end times based on the abrupt changes in the curvature of energy temporal variations includes:

[0117] Energy time series are generated based on the evolution information of perturbation energy over time.

[0118] Smoothing of energy time series;

[0119] Numerical rate of change calculation is performed on the smoothed energy time series to obtain an energy change curvature sequence for characterizing the temporal curvature of the perturbation energy;

[0120] Local extreme points in the energy change curvature sequence that exceed an adaptive threshold are detected and used as boundary candidate points; wherein, the adaptive threshold is generated based on the energy fluctuation level of the background steady-state signal before the disturbance occurs;

[0121] Temporal continuity and persistence constraints are applied to the candidate boundary points to determine the start and end times.

[0122] The energy time series is the perturbation energy evolution sequence extracted in step S34. The sequence In terms of time, it is represented by discrete data points distributed according to sampling intervals, thus completely recording the discrete process of the evolution of the total disturbance energy over time.

[0123] In one specific implementation, the energy time series is smoothed to suppress the interference of measurement noise and random fluctuations on trend judgment. A Savitzky-Golay filter can be used for smoothing, which effectively suppresses high-frequency noise while preserving the trend characteristics and key inflection point information in the energy change curve to the maximum extent, avoiding the weakening of inflection point characteristics at the start and end times of disturbances due to excessive smoothing. The filter window length can be adaptively set according to the system power frequency cycle; for example, the number of sampling points covering 1 to 2 power frequency cycles can be selected as the window length.

[0124] In one specific implementation, the rate of change of the smoothed energy time series is calculated numerically to obtain an energy curvature sequence for characterizing the curvature of the perturbation energy time series. This includes calculating the second derivative of the smoothed energy time series with respect to time. Specifically, the central difference method can be used for numerical calculations. ;in, The sampling time interval, This is the smoothed energy time series; the second derivative sequence In a physical sense, it represents the "acceleration" of energy changes; when energy begins to rise sharply (the start of a disturbance) or decrease sharply (the end of a disturbance), The response will produce significant positive or negative extreme points, which are potential inflection points that mark the time boundary.

[0125] In one specific implementation, generating an adaptive threshold includes the following steps: First, in the smoothed energy time series... In this process, a period of known steady-state operation prior to the disturbance is selected as the background segment. Then, the energy change curvature sequence corresponding to this background segment is calculated. Statistical characteristics of the absolute value sequence. Adaptive threshold. Generate according to the following formula: ,in, and These are the absolute values ​​of the curvature of the background segment. The mean and standard deviation, This is a constant factor, typically ranging from 3 to 5. This threshold generation method allows the judgment criteria to adapt to the current noise background and fluctuation level of the system. Finally, the entire... The absolute value in the sequence exceeds the adaptive threshold The local extreme points are identified, and these points are determined as candidate boundary points indicating the start or end of the disturbance.

[0126] In one specific implementation, temporal continuity and persistence constraints are applied to the candidate boundary points to determine the start and end times. The selection and decision-making are specifically performed using the following constraint rules:

[0127] Persistence constraint: For an initial candidate point (usually corresponding to...) The positive extreme value needs to be verified, and the energy sequence over a subsequent period (e.g., not less than 1 / 4 of a power frequency cycle) needs to be verified. Whether it remains consistently above the steady-state background level; for a termination candidate point (usually corresponding to...) If the negative extreme value is found, then it is verified whether the energy subsequently recovers to the steady-state level. Candidate points that do not meet the sustainability requirement will be eliminated.

[0128] Continuity constraint: Pair a valid starting candidate point with the next most recent valid ending candidate point to form a candidate disturbance event window. If there is no qualified ending point between two starting points, it is necessary to check whether it is a composite disturbance or a detection error.

[0129] Final ruling: After passing the above constraint screening, the remaining pairs of start and end points with clear physical meaning are ultimately determined as the start time of this power quality disturbance event. and the end time .

[0130] In summary, step S5, by focusing on the second-order curvature change in the energy evolution trend and employing an adaptive threshold based on background statistics and physical constraints that include persistence, achieves accurate and robust detection of the perturbation time boundary. This provides a reliable guarantee for step S6 to strictly limit feature extraction to the actual perturbation duration, which is a prerequisite for the high-precision identification of this method.

[0131] S6. Within the time range defined by the start and end times, construct a feature set to characterize the disturbance pattern based on the instantaneous frequency change curvature parameter, identify the power quality disturbance based on the feature set, and output the disturbance type and its corresponding time location parameter.

[0132] Specifically, a feature set for characterizing the perturbation morphology is constructed based on the instantaneous frequency change curvature parameter, including:

[0133] An initial feature vector is generated based on the instantaneous frequency change curvature parameters corresponding to each frequency component.

[0134] The instantaneous frequency change curvature parameters corresponding to each frequency component during the duration of the disturbance are processed to optimize the initial feature vector;

[0135] Furthermore, based on the disturbance energy ratio or frequency stability index obtained from the time-frequency joint analysis of each frequency component, the optimized initial feature vector is weighted and fused to form a feature set for characterizing the disturbance morphology.

[0136] In one specific implementation, a multidimensional feature vector is generated based on the instantaneous frequency change curvature parameter, including: obtaining the feature vector within the perturbation window from step S4. Within this, the curvature statistical features extracted for the M main perturbation frequency components are combined. For example, for each component... Extract the mean of its curvature. Peak and standard deviation Arrange the 3M eigenvalues ​​of these M frequency components in ascending order of frequency to form a 3M-dimensional initial eigenvector required for this step: This lays the data foundation for subsequent accurate classification.

[0137] In one specific implementation, the instantaneous frequency change curvature parameters corresponding to each frequency component during the disturbance duration are processed to optimize the initial feature vector. This is achieved through the following two steps:

[0138] Feature time window alignment: This process unifies the curvature feature values ​​from each frequency component to the precise perturbation time base determined in step S5. Specifically, it is based on the unified start time finally determined in step S5. and termination time It extracts only the portion of the instantaneous frequency change curvature parameter sequence of each frequency component that falls within the window, and recalculates its statistical characteristics (such as mean, peak value, etc.) based on these extracted sequences. This ensures that each value used to construct the feature vector strictly corresponds to the same physical disturbance event.

[0139] Eigenvalue standardization: This process aims to eliminate the scaling effects on subsequent classifiers caused by the different dimensions and numerical ranges of various types of features (such as the mean, peak, and standard deviation of curvature). In a preferred embodiment, Z-score standardization is used. That is, for the initial feature vector... Each feature dimension Calculate its standardized value ;in, and These are the population mean and population standard deviation, respectively, obtained from a large number of pre-constructed training samples. This process yields a standardized feature vector that is uniformly scaled and suitable for comparison and computation.

[0140] In one specific implementation, based on the disturbance energy proportion or frequency stability index obtained from time-frequency joint analysis of each frequency component, the optimized initial feature vector is weighted and fused. This includes the following steps: First, the energy proportion weight and frequency stability weight are calculated. Then, the two weights are combined by weighted summation to obtain the final weight of each frequency component. Finally, the weighted fusion is performed, i.e., the weight of the first frequency component is... All features corresponding to each component (such as...) , , Each component is multiplied by its final weight. This yields the weighted feature vector. This operation highlights the contribution of the dominant perturbation component features and enhances the representational power of the feature set.

[0141] In a preferred embodiment, the formula for calculating the energy percentage weight is: ;in, For the first Each frequency component is generated from the initial time. and termination time The total perturbation energy within the defined perturbation time window (as measured by the synchronous compression transform spectrum) (The energy of the time-frequency ridge corresponding to this component is obtained by integrating). This represents the total number of frequency components. Indicates all Disturbance energy of each frequency component Summation is performed. This energy percentage weighting... This reflects the relative energy intensity of each component in this disturbance event.

[0142] In a preferred embodiment, the formula for calculating the frequency stability weight is: ,in, Indicates the first Each frequency component in its instantaneous frequency estimation The standard deviation within the disturbance time window. Higher stability ( The smaller the value, the greater the weight.

[0143] In a more preferred embodiment, the final weight of the i-th frequency component By weighting the proportion of energy With frequency stability weight The linear combination yields the result, and the calculation formula is as follows: ;in, The preset balance coefficient, and By adjusting The value of can be flexibly adjusted to determine the relative importance of energy dominance and frequency stability in feature fusion.

[0144] In one specific implementation, power quality disturbances are identified based on a feature set, and the disturbance type and its corresponding time and location parameters are output, including:

[0145] The weighted and fused feature set is input into a pre-trained power quality disturbance classification model to obtain disturbance type labels. The classification model (such as a support vector machine, random forest, or deep neural network) is trained using a large number of historical power quality disturbance data samples and their corresponding feature sets extracted in steps S1 to S6 of this invention. The label output by the model is the identification result of the current event, and its type includes, but is not limited to, voltage sag, voltage swell, harmonics, interharmonics, and oscillations.

[0146] Subsequently, the complete identification result is output, which includes at least the perturbation type label and the time location parameter of the event determined in step S5, i.e., the start time. and the end time Based on this, the duration of the disturbance can be calculated. In a preferred embodiment, the identification result is output in a structured data format, such as: {“Event Type”: “Voltage Sag”, “Start Time”: “2025-10-23 14:05:30.125”, “End Time”: “2025-10-23 14:05:30.356”, “Duration”: “0.231s”}.

[0147] It should be noted that step S6, as the final integration and intelligent decision-making stage of the entire method, transforms the deep time-frequency features (curvature of energy and frequency changes) characterizing the dynamic evolution of power quality disturbances mined in the preceding steps (S3-S5) into intuitive, reliable, and physically meaningful engineering diagnostic results through systematic feature vector construction, time window alignment, standardization, weighted fusion, and classification identification. This signifies that the present invention has achieved a complete technical closed loop from "high-precision time-frequency signal analysis" to "intelligent disturbance identification and location," providing core methodological support for online monitoring and precise governance of power systems.

[0148] In summary, this embodiment 1, by employing synchronous compression-type time-frequency transform and extracting two types of high-order dynamic features—instantaneous frequency change curvature and energy temporal change curvature—improves the characterization and discrimination capabilities of composite power quality disturbances with time-varying frequencies and multiple superimposed components on the time-frequency plane. Specifically, the synchronous compression process overcomes the energy diffusion problem of traditional time-frequency analysis, obtaining a high-resolution time-frequency representation and instantaneous frequency trajectory; furthermore, by taking the second derivative of the frequency trajectory to obtain the "change curvature" and taking the second rate of change of the energy evolution sequence to obtain the "temporal change curvature," the identification features are elevated from static amplitude and frequency points to high-order temporal features capable of characterizing the dynamic evolution rate of disturbances and their inflection points. This allows different disturbances that may be similar in the spectrum (such as harmonics and interharmonics, sags and oscillations) to be effectively distinguished in terms of their dynamic evolution morphology. This invention achieves high-precision, adaptive localization of the start and end times of disturbance events by detecting abrupt changes based on the curvature of energy temporal variations, combined with adaptive thresholds and physical constraint rules. This enhances robustness to complex operating conditions and noisy environments. Compared to traditional criteria that rely on fixed amplitude thresholds, this invention captures abrupt change inflection points by analyzing the curvature (second-order rate of change) of the energy evolution trend. It also dynamically generates a judgment threshold based on the fluctuation level of the background steady-state signal before the disturbance, effectively avoiding misjudgments and missed judgments caused by load fluctuations or noise interference. This provides a precise time window for subsequent feature extraction and recognition.

[0149] Example 2, as Figure 5 This is the second embodiment of the present invention. Based on Embodiment 1, this embodiment further provides a method for systematically or locally determining disturbance events in a three-phase power system. The purpose of this method is to distinguish between global disturbances affecting the entire system (such as voltage dips on the grid side) and local disturbances affecting only a portion of the phase lines (such as single-phase grounding faults or unbalanced load switching) by utilizing the phase consistency characteristics exhibited by the three-phase voltage or current signals when a disturbance occurs. The specific operation steps are as follows:

[0150] First, the three-phase voltage or current signals that are synchronously acquired with the voltage or current signals in step S1 are preprocessed. The preprocessing includes performing adaptive fundamental notch filtering on the fundamental component of the signal as mentioned in step S2 of Example 1, so as to weaken the fundamental component and highlight the disturbance component of each phase, so as to obtain the preprocessed three-phase signal.

[0151] Next, the preprocessed three-phase signals are subjected to joint time-frequency analysis (preferably synchronous compression transform) as mentioned in step S3 of the embodiment to obtain the high-resolution time spectrum of each phase. Subsequently, based on the obtained high-resolution time spectrum of each phase, the main disturbance frequency components of interest (e.g., fundamental frequency or dominant harmonic frequency) are analyzed at the same time-frequency position in the spectrum. The corresponding instantaneous phase value is extracted from the above. That is, for the frequency component... The instantaneous phase sequences of the three phases were obtained respectively: , , .

[0152] Then, based on the extracted instantaneous phases, the instantaneous phase difference between each pair of phases is calculated. For example:

[0153] ;

[0154] ;

[0155] ;

[0156] Next, based on these instantaneous phase difference sequences, a comprehensive three-phase time-frequency phase consistency deviation index D is calculated to quantify the synchronization degree of the three-phase disturbance characteristics or to characterize the consistency degree of disturbance characteristics between different phases. In a preferred embodiment, the deviation index D is obtained through statistical analysis of all phase difference values ​​at all time points. The statistical analysis includes at least one or more of the following: mean phase difference, phase difference fluctuation range, or phase difference stability index. The smaller the D value, the smaller the three-phase phase difference fluctuation and the higher the phase consistency; the larger the D value, the greater the three-phase phase difference and the worse the consistency.

[0157] Then, the calculated consistency deviation index D is compared with a preset judgment threshold. Comparison:

[0158] If D ≤ If so, the current power quality disturbance is determined to be a systematic disturbance (symmetric disturbance).

[0159] If D > If so, the current power quality disturbance is determined to be a local or asymmetric disturbance.

[0160] Among them, the preset threshold The setting method is as follows: Select multiple historical event samples known as systematic disturbances (such as three-phase short-circuit faults), calculate the statistical distribution of their phase consistency deviation index D, and take the 95th quantile of this distribution as... The initial value; if it needs to adapt to different voltage levels or system structures, it can be set. The value ranges from 0.1 radians to 0.3 radians. In practical applications, the value can be adjusted based on field operating experience. Fine-tuning was performed to balance the sensitivity of systemic disturbance detection with the accuracy of local disturbance determination.

[0161] Finally, the result of the disturbance nature determination (systematic / local) is correlated with the disturbance type and time location parameters output in step S6 of Example 1 to form a more comprehensive diagnostic report, which is output together.

[0162] In summary, this embodiment 2 performs high-resolution time-frequency analysis on the preprocessed three-phase signals, extracts and compares their instantaneous phases at the main disturbance frequencies, and then calculates the phase consistency deviation index. Finally, based on the comparison of this index with a preset threshold, it realizes the automatic and objective determination of whether the power quality disturbance is a "systemic disturbance" affecting the entire power system or a "local or asymmetrical disturbance" involving only some phase lines. The determination result is then output in association with the type and time location of the disturbance, thereby significantly improving the comprehensiveness of disturbance diagnosis and its engineering guidance value.

[0163] Example 3 is the third embodiment of the present invention. Based on Example 1 or Example 2, this embodiment further adds confidence assessment and standardized output functions for the identification results, enhancing the practicality and engineering value of the method. Specifically, it includes the following operational steps:

[0164] First, the confidence level of the identification results output by the aforementioned embodiments (including the type of disturbance, start and end times, and the nature of the disturbance in Embodiment 2) is evaluated quantitatively, and a comprehensive measurement uncertainty is calculated. In one specific implementation, the measurement uncertainty A comprehensive calculation is performed based on one or more of the following input factors:

[0165] Sampling noise level: ,in This is the estimated signal-to-noise ratio of the original signal.

[0166] Time-frequency transformation stability: H ,in H Time spectrum The higher the entropy value, the more dispersed the energy and the lower the stability of the transformation.

[0167] Fluctuations in curvature parameters due to instantaneous frequency changes: , which is the average of the standard deviations of the curvature sequences of each major frequency component, reflects the volatility of the feature itself.

[0168] Three-phase phase coherence level (where applicable): D, which is the phase consistency deviation index calculated in Example 2.

[0169] In another specific implementation, the synthesis is performed using the weighted sum of square roots method:

[0170] ;

[0171] Where is the normalized weight coefficient of each factor, which can be calibrated experimentally. The lower the value, the higher the reliability of the identification result.

[0172] Next, to ensure the identification results directly conform to industry standards and are used for evaluation, the disturbance type and time location parameters output from Example 1 are mapped to standardized power quality evaluation parameters. Power quality evaluation parameters include, but are not limited to, the following examples:

[0173] Disturbance duration: For any type of disturbance event, the duration is determined according to the precise start time defined in step S5 of Example 1. and the end time Calculate its duration .

[0174] Voltage sag / surge depth: Calculate the duration of the disturbance from the original voltage signal. The percentage decrease or increase of the root mean square value of the internal voltage relative to the rated voltage.

[0175] Harmonic Influence Level: From the Time Spectrum In the process, the average amplitude or energy percentage of each harmonic (such as the 3rd, 5th, and 7th harmonics) at the frequency of each harmonic within the disturbance window is extracted, and the total harmonic distortion rate or the content rate of each harmonic is calculated.

[0176] Finally, all the above information is integrated to form a structured high-level diagnostic report. The report includes: disturbance type, start and end times and duration, disturbance nature (systematic / local, from Example 2), and measurement uncertainty. And the corresponding power quality evaluation parameters. For example: {"Event Type": "Voltage Sag", "Nature": "Systemic", "Start Time":} "Duration": “Descent depth”: 85%; “Measurement uncertainty”: 0.05.

[0177] In summary, Example 3 integrates a comprehensive assessment of measurement uncertainty based on sampling noise level, time-frequency transformation stability, curvature parameter fluctuation, and phase consistency into the identification process, providing a quantitative reliability index for the identification results and enhancing the reliability and decision-making reference value of the output conclusions. In addition, it maps the identified disturbance type and time parameter to standard evaluation parameters such as duration, sag depth, and harmonic impact degree, thereby outputting a power quality diagnostic report that includes confidence assessment and standardized indicators, thus improving the engineering output system of the method.

[0178] Example 4 is the fourth embodiment of the present invention. This embodiment provides a power quality disturbance identification system based on time-frequency joint analysis. This system is applied to the above method and includes:

[0179] The synchronous sampling module is used to synchronously sample voltage or current signals in a power system.

[0180] The preprocessing module is used to perform adaptive fundamental notch filtering on the sampled signal;

[0181] The time-frequency analysis module is used to perform joint time-frequency analysis and calculate instantaneous frequency trajectories;

[0182] The feature generation module is used to calculate the instantaneous frequency change curvature parameter and the energy temporal change curvature, and to construct a feature set to characterize the perturbation morphology;

[0183] The disturbance identification module is used to identify the type of power quality disturbance based on the feature set, and output the disturbance type and its corresponding time and location parameters.

[0184] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for power quality disturbance identification by time-frequency joint analysis, characterized in that, Includes the following steps: The voltage or current signal in the power system under test is synchronously sampled to obtain the original time-domain sampled signal; The original time-domain sampled signal is preprocessed, and the preprocessing includes performing adaptive fundamental notch filtering on the fundamental component of the signal; Perform time-frequency joint analysis on the preprocessed signal to obtain the energy distribution of the preprocessed signal in the time-frequency plane and the evolution information of the disturbance energy derived from the energy distribution over time, and simultaneously obtain the corresponding instantaneous frequency information. Based on the instantaneous frequency information, the instantaneous frequency trajectory of the disturbance signal is calculated, and a second-order change analysis is performed on the instantaneous frequency trajectory to generate an instantaneous frequency change curvature parameter that characterizes the evolution of the disturbance frequency. Based on the evolution information of the disturbance energy over time, the curvature of the energy time series change is calculated, and the start and end times of the power quality disturbance event are determined based on the abrupt change characteristics of the curvature of the energy time series change. Within the time range defined by the start and end times, a feature set for characterizing the disturbance pattern is constructed based on the instantaneous frequency change curvature parameter, and the power quality disturbance is identified according to the feature set, and the disturbance type and its corresponding time location parameter are output. The method further includes: The three-phase voltage or current signals acquired synchronously with the voltage or current signals are preprocessed respectively. The preprocessing includes performing adaptive fundamental notch filtering on the fundamental component of the signal to weaken the fundamental component. Perform time-frequency joint analysis on the preprocessed three-phase signals respectively, and calculate the instantaneous phase difference between each phase at the corresponding time-frequency position; Based on the instantaneous phase difference, a deviation index reflecting the consistency of the three-phase time-frequency phase is calculated to characterize the degree of consistency of the disturbance characteristics between different phases. When the deviation of the phase consistency is lower than or equal to a preset threshold, the identified power quality disturbance is determined to be a systematic disturbance; when the deviation of the phase consistency is higher than the preset threshold, it is determined to be a local or asymmetric disturbance. The determination results of the systemic disturbance, local disturbance, or asymmetric disturbance are associated with the disturbance type and time location parameters and output.

2. The method for power quality disturbance identification of time-frequency joint analysis according to claim 1, characterized in that: The time-frequency joint analysis employs a synchronous compressed time-frequency transform method. The synchronous compression process redistributes energy in the initial time-frequency representation, concentrating the energy along the instantaneous frequency trajectory. Furthermore, during synchronous compression, the instantaneous frequency is calculated based on the phase change rate of the time-frequency coefficient.

3. The method for power quality disturbance identification of time-frequency joint analysis according to claim 1, characterized in that: The instantaneous frequency change curvature parameter is obtained by calculating at least the second derivative of the instantaneous frequency trajectory, where the first derivative is used to characterize the frequency drift rate and the second derivative is used to characterize the degree of bending in the direction of frequency change. When generating the instantaneous frequency change curvature parameter, the instantaneous frequency trajectories corresponding to different frequency components are first time-aligned, and the amplitude of each component is normalized according to the frequency change amplitude during the disturbance duration. Then, the corresponding change curvature parameter is calculated based on the aligned and normalized instantaneous frequency trajectory. Furthermore, statistical analysis is performed on the sequence of instantaneous frequency change curvature parameters during the duration of the disturbance, extracting at least one or more of the following: mean, peak value, range of variation, and standard deviation.

4. The method for power quality disturbance identification of time-frequency joint analysis according to claim 1, characterized in that: Determining the start and end times based on the abrupt changes in the curvature of the energy temporal variation includes: Based on the evolution information of the disturbance energy over time, an energy time series is generated; The energy time series is smoothed. Numerical rate of change calculation is performed on the smoothed energy time series to obtain an energy change curvature sequence for characterizing the temporal curvature of the perturbation energy; Local extreme points in the energy change curvature sequence that exceed an adaptive threshold are detected and used as boundary candidate points; wherein, the adaptive threshold is generated based on the energy fluctuation level of the background steady-state signal before the disturbance occurs; Temporal continuity and persistence constraints are applied to the candidate boundary points to determine the start and end times.

5. The method for power quality disturbance identification of time-frequency joint analysis according to claim 1, characterized in that: Based on the instantaneous frequency change curvature parameter, a feature set for characterizing the perturbation pattern is constructed, including: An initial feature vector is generated based on the instantaneous frequency change curvature parameter corresponding to each frequency component. The instantaneous frequency change curvature parameters corresponding to each frequency component during the duration of the disturbance are processed to optimize the initial feature vector; Furthermore, the optimized initial eigenvectors are weighted and fused based on the perturbation energy ratio or frequency stability index obtained from the time-frequency joint analysis of each frequency component.

6. The method for power quality disturbance identification of time-frequency joint analysis according to claim 1, characterized in that: The deviation index reflecting the consistency of the three-phase time-frequency phase is obtained by statistical analysis of the instantaneous phase of each corresponding frequency component. The statistical analysis includes at least one or more of the following: mean phase difference, phase difference fluctuation range, or phase difference stability index.

7. The method for power quality disturbance identification of time-frequency joint analysis according to claim 1, characterized in that: The method also includes evaluating the measurement uncertainty of the disturbance identification results; The measurement uncertainty is calculated comprehensively based at least on the sampling noise level, the transformation stability of the time-frequency joint analysis, the fluctuation of the instantaneous frequency change curvature parameter, and the consistency level of the instantaneous phase difference between each phase when identifying three-phase signals. The measurement uncertainty is then output in association with the disturbance type and time location parameters.

8. The method for power quality disturbance identification of time-frequency joint analysis according to claim 1, characterized in that: The method further includes: The output disturbance type and time location parameters are mapped to the corresponding power quality evaluation parameters; The power quality evaluation parameters include at least the voltage sag depth, disturbance duration, or harmonic impact level.

9. A power quality disturbance identification system using time-frequency joint analysis, applied to the power quality disturbance identification method using time-frequency joint analysis according to any one of claims 1 to 8, characterized in that, include: The synchronous sampling module is used to synchronously sample voltage or current signals in the power system, as well as three-phase voltage or current signals that are synchronously acquired with the voltage or current signals. The preprocessing module is used to perform adaptive fundamental notch filtering on the sampled signal; The time-frequency analysis module is used to perform joint time-frequency analysis and calculate the instantaneous frequency trajectory, as well as to perform joint time-frequency analysis on the preprocessed three-phase signals and calculate the instantaneous phase difference between each phase; The feature generation module is used to calculate the instantaneous frequency change curvature parameter and the energy temporal change curvature, and to construct a feature set to characterize the perturbation morphology; The disturbance identification module is used to identify the type of power quality disturbance based on the feature set, calculate the deviation index reflecting the consistency of the three-phase time-frequency phase based on the instantaneous phase difference, obtain the judgment result based on the comparison result of the deviation index and the preset threshold, and output the judgment result, the disturbance type and its corresponding time position parameters.

Images