A power grid broadband voltage fluctuation type identification method based on multi-feature fusion

Abstract

This invention discloses a method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion, relating to the field of voltage fluctuation technology. It addresses the problem of fluctuation type confusion under conditions of multiple disturbances and non-stationary operation. The method involves acquiring three-phase voltage data from voltage transformers and calculating the voltage deviation rate, while simultaneously retrieving the short-time flicker index. A fluctuation identification mechanism is jointly triggered based on these two factors. After triggering, a broadband voltage signal is acquired and modal decomposition is performed. The instantaneous frequencies of each mode are extracted and classified. Historical voltage data is retrieved to construct the voltage envelope trajectory, and the envelope asymmetry is calculated. Disturbances are introduced to different modes to enhance energy, and the modal energy spectra are calculated and fused to form modal modulation feature vectors. A pre-trained support vector machine model is used for classification mapping, achieving accurate identification of sinusoidal modulation, rectangular modulation, and interharmonic modulation voltage fluctuations, thus improving identification accuracy under complex operating conditions.

Description

Technical Field

[0001] This invention relates to the field of voltage fluctuation technology, and more specifically, to a method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion. Background Technology

[0002] With the large-scale integration of new energy sources and the widespread application of power electronic equipment, the power grid operation is gradually evolving from a traditional stable operating condition to a complex dynamic operating condition with strong fluctuations and strong coupling. The voltage signal at the grid-connected substation exhibits obvious broadband characteristics, including not only low-frequency voltage offsets but also flicker, interharmonics, and various modulation disturbance components. These voltage fluctuations have significant non-stationarity and multi-modal characteristics on both the time and frequency scales, making the manifestations of voltage quality problems more diverse and concealed.

[0003] The existing technology has the following shortcomings: Currently, existing technologies mainly rely on single voltage indicators or fixed frequency band analysis methods to determine voltage signals. They lack the ability to adaptively decompose the multi-modal structure of broadband signals and collaboratively model multi-source features. This makes it difficult to accurately characterize the differences in instantaneous frequency evolution, energy distribution, and envelope shape of voltage fluctuations of different modulation types. As a result, fluctuation type confusion and misjudgment are prone to occur under multiple disturbance superposition and non-stationary operating conditions. This cannot meet the requirements for high-precision identification and stable determination of voltage fluctuation types in complex power grid environments. Therefore, a broadband voltage fluctuation type identification method based on multi-feature fusion is proposed.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion. This method addresses the problems mentioned in the background art by employing a multi-source triggering determination mechanism, a broadband signal adaptive mode decomposition method, a voltage envelope asymmetric feature extraction mechanism, a mode energy spectrum and disturbance enhancement energy fusion modeling method, and a feature mapping classification mechanism based on support vector machines.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion, comprising the following steps: Step S1: Monitor the grid-connected substation under test, collect the three-phase voltage data of the grid-connected substation under test, calculate the voltage deviation rate based on the three-phase voltage data, access the power grid operation platform to retrieve the short-time flicker index, and determine whether the fluctuation identification mechanism is triggered based on the voltage deviation rate. Step S2: In the fluctuation identification mechanism, the wideband signal of the grid-connected substation under test is acquired, the wideband signal is processed by mode decomposition, the instantaneous frequency of each decomposed mode is monitored, and the mode category of the decomposed mode is classified. Step S3: Obtain historical voltage data of the grid-connected substation under test, generate voltage envelope trajectory based on historical voltage data and calculate voltage envelope asymmetry, set disturbance enhancement energy for each mode category based on voltage envelope asymmetry, and detect modal energy spectrum for each mode category; Step S4: Generate the mode modulation characteristics of the grid-connected substation under test by integrating the mode energy spectrum and the disturbance enhancement energy, call the machine learning model to perform feature mapping processing on the mode modulation characteristics, and identify the voltage fluctuation type of the grid-connected substation under test based on the processing results.

[0007] In a preferred embodiment, in step S1, a voltage transformer is installed at the grid connection point of the substation to be tested, and the three-phase voltage signal is synchronously acquired at a preset sampling frequency to obtain three-phase voltage data. The three-phase voltage data consists of real-time measurements of the voltage to ground for each phase, including the instantaneous sampled values ​​of phase A voltage, phase B voltage, and phase C voltage. Based on the three-phase voltage data, the equivalent root mean square value of the three-phase voltage is calculated to obtain the equivalent voltage value; Access the power grid operation platform to retrieve the rated voltage value corresponding to the grid-connected substation under test. The rated voltage value is the standard operating voltage specified in the power grid plan.

[0008] In a preferred embodiment, in step S1, the voltage deviation rate is calculated based on the equivalent voltage value and the rated voltage value; The short-time flicker index of the grid-connected substation under test is retrieved from the power grid operation platform within the current time window. The short-time flicker index is a voltage fluctuation sensing index obtained based on a preset time scale. The voltage deviation rate and short-time flicker index are combined for judgment. When the voltage deviation rate is greater than the preset deviation threshold and the short-time flicker index is greater than the preset flicker threshold, the current grid-connected substation under test is determined to be in a state of abnormal voltage fluctuation, triggering the fluctuation identification mechanism. Otherwise, the fluctuation identification mechanism will not be triggered, and the normal monitoring status will be maintained.

[0009] In a preferred embodiment, in step S2, after the fluctuation identification mechanism is triggered, the equivalent voltage value calculated from the sequence of instantaneous three-phase voltage samples output by the voltage transformer at the grid connection point is used as the three-phase composite voltage signal. The three-phase composite voltage signal is subjected to mean removal processing to obtain a wide frequency domain signal; The empirical mode decomposition method is used to perform mode decomposition processing on the broadband signal to obtain several decomposed modes. Each decomposed mode represents the oscillation component of the broadband signal at a specific frequency scale. The analytic signal of each decomposed mode is constructed using Hilbert transform, the instantaneous phase is calculated based on the analytic signal, and the instantaneous frequency of the decomposed mode is calculated based on the instantaneous phase.

[0010] In a preferred embodiment, in step S2, the average instantaneous frequency of the decomposed mode within a preset time window is calculated, and the average instantaneous frequency characterizes the dominant frequency characteristics of the decomposed mode. When the average instantaneous frequency is less than the preset first division threshold, the decomposed mode is determined to be a low-frequency mode; When the average instantaneous frequency is greater than or equal to the preset first division threshold and less than the preset second division threshold, the decomposed mode is determined to be a mid-frequency mode. When the average instantaneous frequency is greater than or equal to the preset second division threshold, the decomposed mode is determined to be a high-frequency mode.

[0011] In a preferred embodiment, in step S3, a historical analysis window is preset, and historical voltage data of the grid-connected substation under test, including positive sequence voltage amplitude sequence, is obtained through the substation monitoring database within the historical analysis window. For each three-phase voltage effective value in the positive sequence voltage amplitude sequence, if the current positive sequence voltage amplitude is greater than both the previous and the next positive sequence voltage amplitude, then the current positive sequence voltage amplitude is marked as the upper envelope feature point. Similarly, if the current positive sequence voltage amplitude is smaller than both the previous and the next positive sequence voltage amplitudes, then the current positive sequence voltage amplitude is marked as the lower envelope feature point. The voltage envelope trajectory includes an upper envelope trajectory and a lower envelope trajectory. The upper envelope trajectory is generated by interpolating and connecting the feature points of the upper envelope in chronological order, and the lower envelope trajectory is generated by interpolating and connecting the feature points of the lower envelope in chronological order.

[0012] In a preferred embodiment, in step S3, the average value of the positive sequence voltage amplitude sequence is used as the voltage center reference value, and the positive offset amplitude between the upper envelope trajectory and the voltage center reference value and the negative offset amplitude between the lower envelope trajectory and the voltage center reference value are calculated respectively. The average values ​​of the positive and negative offset amplitudes are calculated to obtain the average offset of the upper envelope and the average offset of the lower envelope. The average envelope span is obtained by comprehensively calculating the values ​​of the upper and lower envelope trajectories. The voltage envelope asymmetry is calculated by combining the average offset of the upper envelope, the average offset of the lower envelope, and the average envelope span. The voltage envelope asymmetry is standardized to obtain the envelope adjustment coefficient, and a preset category weight coefficient is applied to different mode categories. The product of the preset category weight coefficient and the envelope adjustment coefficient is used as the perturbation enhancement energy for each modality.

[0013] In a preferred embodiment, in step S3, the decomposed modes of each modal category are subjected to frequency domain transformation to obtain a spectral amplitude sequence. In the spectral amplitude sequence, the spectral amplitude is divided according to a preset frequency range. The modal energy values ​​of the frequency range are obtained by squaring and summing the spectral amplitudes within the preset frequency range. When a modal category contains multiple decomposed modes, the modal energy values ​​within the same frequency range are summed to obtain the total modal energy value of the modal category in each frequency range; The total modal energy values ​​of each frequency range are arranged in order of increasing frequency to form the modal energy spectrum of the modal category.

[0014] In a preferred embodiment, in step S4, the mode categories are arranged in the order of low-frequency mode, mid-frequency mode, and high-frequency mode to obtain the mode category division order; The total modal energy spectrum is multiplied by the perturbation enhancement energy of the corresponding modal category to obtain the corrected total modal energy value; The modal energy spectra of the modal categories are arranged according to the classification order, and the corrected total modal energy values ​​in the modal energy spectra are arranged in order to form modal modulation characteristics; The machine learning model is invoked through the model call interface. The modal modulation features of the grid-connected substation under test are input into the machine learning model. The machine learning model maps the modal modulation features according to the classification decision boundary formed during the training phase and outputs the corresponding voltage fluctuation type identification result. The voltage fluctuation type identification results include sinusoidal modulation fluctuation, rectangular modulation fluctuation, and interharmonic modulation fluctuation.

[0015] The technical effects and advantages of this invention are as follows: This invention acquires three-phase voltage data and calculates the voltage deviation rate using voltage transformers. Simultaneously, it retrieves the short-time flicker index from the power grid operation platform. These two factors serve as a joint trigger condition to determine whether to activate the fluctuation identification mechanism. Upon triggering, a wide-frequency voltage signal is acquired and modal decomposition is performed. The instantaneous frequency of each modal component is extracted, and modal categories are classified. Historical voltage data is retrieved to construct the voltage envelope trajectory. The voltage envelope asymmetry is calculated based on the difference between the upper and lower edges of the envelope. Disturbance enhancement energy is introduced for different modal categories, and the modal energy spectrum is calculated for each category. This spectrum is then fused with the disturbance enhancement energy to construct a modal modulation feature vector. Finally, a pre-trained support vector machine model is used to classify and map the feature vectors, achieving accurate identification of sinusoidal modulation fluctuations, rectangular modulation fluctuations, and interharmonic modulation fluctuations. By fusing multi-source features, the accuracy of voltage fluctuation type identification under complex operating conditions is improved. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the implementation of a method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion, as described in this invention.

[0017] Figure 2 This is a schematic diagram illustrating the steps of a method for identifying broadband voltage fluctuation types in a power grid based on multi-feature fusion, according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention acquires three-phase voltage data and calculates the voltage deviation rate using voltage transformers. Simultaneously, it retrieves the short-time flicker index from the power grid operation platform. These two factors are used as a joint trigger condition to determine whether to activate the fluctuation identification mechanism. After triggering, it acquires a wide-frequency voltage signal and performs mode decomposition processing, extracts the instantaneous frequency of each mode component, and completes the mode category classification. It retrieves historical voltage data to construct a voltage envelope trajectory, calculates the voltage envelope asymmetry based on the difference between the upper and lower edges of the envelope, introduces disturbance enhancement energy into different mode categories, calculates the mode energy spectrum for each mode category, and fuses it with the disturbance enhancement energy to construct a mode modulation feature vector. It then calls a pre-trained support vector machine model to classify and map the feature vector, achieving accurate identification of sinusoidal modulation fluctuations, rectangular modulation fluctuations, and interharmonic modulation fluctuations.

[0020] Example 1, such as Figures 1 to 2As shown, a method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion includes the following steps: Step S1: Monitor the grid-connected substation under test, collect the three-phase voltage data of the grid-connected substation under test, calculate the voltage deviation rate based on the three-phase voltage data, access the power grid operation platform to retrieve the short-time flicker index, and determine whether the fluctuation identification mechanism is triggered based on the voltage deviation rate. Step S2: In the fluctuation identification mechanism, the wideband signal of the grid-connected substation under test is acquired, the wideband signal is processed by mode decomposition, the instantaneous frequency of each decomposed mode is monitored, and the mode category of the decomposed mode is classified. Step S3: Obtain historical voltage data of the grid-connected substation under test, generate voltage envelope trajectory based on historical voltage data and calculate voltage envelope asymmetry, set disturbance enhancement energy for each mode category based on voltage envelope asymmetry, and detect modal energy spectrum for each mode category; Step S4: Generate the mode modulation characteristics of the grid-connected substation under test by integrating the mode energy spectrum and the disturbance enhancement energy, call the machine learning model to perform feature mapping processing on the mode modulation characteristics, and identify the voltage fluctuation type of the grid-connected substation under test based on the processing results.

[0021] The specific implementation is as follows: In step S1, continuous operation status monitoring is performed on the grid-connected substation under test. Specifically, a voltage transformer is installed at the grid connection point of the grid-connected substation under test, and the three-phase voltage signals are synchronously acquired at a preset sampling frequency to obtain three-phase voltage data. The three-phase voltage data are the real-time measured values ​​of the voltage to ground of each phase, including the instantaneous sampling value sequence of phase A voltage, the instantaneous sampling value sequence of phase B voltage, and the instantaneous sampling value sequence of phase C voltage.

[0022] It should be noted that the grid connection point refers to the common connection node where the substation under test is electrically connected to the external power grid and exchanges electrical energy, corresponding to the interface location where the substation's busbar side connects to the power grid; a voltage transformer is an electrical measuring device installed at the grid connection point to convert the primary side high voltage signal into a secondary side low voltage signal proportionally. Its primary side is connected to the high voltage line under test, and its secondary side outputs a low voltage signal of standard amplitude for use by measurement and protection equipment. The transformation ratio of the voltage transformer is a fixed value, and its output voltage satisfies a linear proportional relationship with the primary side voltage.

[0023] Based on the three-phase voltage data, the equivalent root mean square value of the three-phase voltage is calculated to obtain the equivalent voltage value. The specific calculation method is as follows: ; in, This is the equivalent voltage value. This is a sequence of instantaneous sampled values ​​of phase A voltage. This is a sequence of instantaneous sampled values ​​of phase B voltage. This is a sequence of instantaneous sampled values ​​of the C-phase voltage. The sampling time.

[0024] The equivalent voltage value represents the overall voltage amplitude at the grid connection point at the sampling time. The larger the value, the higher the overall voltage level at the corresponding sampling time.

[0025] Furthermore, the rated voltage value corresponding to the grid-connected substation under test is retrieved from the power grid operation platform. The power grid operation platform refers to a comprehensive information management and control system constructed by the power dispatching agency, used for centralized collection, storage, analysis, and dispatch control of the status parameters of each operating node of the power grid. In this embodiment, the power grid operation platform pre-stores the basic operating parameter information of each grid-connected substation, including the rated voltage value. The rated voltage value is the standard operating voltage specified in the power grid plan, and its value is a fixed reference quantity used to characterize the voltage level under normal operating conditions.

[0026] The voltage deviation rate is calculated based on the equivalent voltage value and the rated voltage value. The specific calculation method is as follows: ; in, Voltage deviation rate, This is the equivalent voltage value. Sampling time, This is the rated voltage value.

[0027] The voltage deviation rate characterizes the degree of deviation of the current voltage from the rated voltage. The larger the value, the more serious the deviation of the voltage from the standard operating condition.

[0028] Meanwhile, the short-time flicker index of the grid-connected substation under test is retrieved through the power grid operation platform within the current time window. The short-time flicker index is a voltage fluctuation perception index obtained based on a preset time scale. Its value reflects the degree of flickering effect of voltage fluctuation on electrical equipment (such as lighting equipment). The larger the value, the more obvious the visual flicker caused by voltage fluctuation.

[0029] Finally, the voltage deviation rate and short-time flicker index are jointly judged. When the absolute value of the voltage deviation rate is greater than the preset deviation threshold and the short-time flicker index is greater than the preset flicker threshold, the current grid-connected substation under test is determined to be in a state of abnormal voltage fluctuation, and the fluctuation identification mechanism is triggered; otherwise, the fluctuation identification mechanism is not triggered, and the normal monitoring state is maintained.

[0030] It should be noted that the preset deviation threshold is used to determine the allowable deviation range of the voltage amplitude relative to the rated voltage, and its setting process is based on a combination of historical statistical characteristics. Specifically, historical data on the voltage deviation rate of the substation under test during its long-term operation period are retrieved to construct a voltage deviation rate sample set, and its mean and standard deviation are calculated. The sum of the mean and standard deviation is used as the preset deviation threshold. The preset flicker threshold is used to determine the acceptable degree of the visual impact of voltage fluctuations on electrical equipment, and its setting is based on a comprehensive determination of flicker perception standards and equipment sensitivity characteristics. Specifically, firstly, according to the limit requirements for short-time flicker index in the power quality standards (such as the flicker tolerance level specified in the standards), then the operating characteristic parameters of typical loads (such as lighting equipment) in the power supply area of ​​the target substation are retrieved to construct a correction factor. The limit requirements are multiplied by the correction factor for adjustment to obtain the preset flicker threshold applicable to the current application scenario.

[0031] Through the above processing, dual constraint triggering based on voltage amplitude deviation and flicker sensing intensity is achieved, which improves the accuracy and reliability of fluctuation identification start determination.

[0032] In step S2, after the fluctuation identification mechanism is triggered, the wideband signal of the grid-connected substation under test is acquired.

[0033] Specifically, the equivalent voltage value calculated from the sequence of instantaneous three-phase voltage samples output by the voltage transformer at the grid connection point is used as the three-phase composite voltage signal.

[0034] Furthermore, the three-phase composite voltage signal is subjected to mean removal processing to obtain a wideband signal, the specific expression of which is as follows: ; in, It is a wideband signal. This is the equivalent voltage value. Let be the equivalent voltage value at the i-th sampling time, which is used here as the three-phase composite voltage signal. Sampling time, This represents the number of sampling points within the current preset analysis time window. This is the index value of the sampling point.

[0035] Wideband signals are used to characterize the fluctuation component of voltage relative to its average level. The larger the value, the greater the instantaneous deviation from the average voltage, thus highlighting the wideband fluctuation characteristics.

[0036] Empirical mode decomposition (EMD) is applied to a wideband signal to perform mode decomposition, resulting in several decomposed modes, the specific expressions of which are as follows: ; in, It is a wideband signal. To decompose the total number of modes, To decompose the modal index values, Indicates the first One decomposition mode, Sampling time, This is the residual component.

[0037] Each decomposed mode represents the oscillation component of a broadband signal at a specific frequency scale. The higher the frequency, the faster the voltage fluctuation change reflected by the corresponding decomposed mode.

[0038] For each decomposed mode, construct its analytical signal: ; in, To analyze the signal, Indicates the first One decomposition mode, Sampling time, To decompose the modal index values, The imaginary unit, Represents the Hilbert transform operator. The signal obtained after performing a Hilbert transform on the decomposed modes.

[0039] It should be noted that the purpose of introducing the imaginary unit is to expand the real signal into a complex signal, thereby simultaneously representing the amplitude and phase information of the signal. In the above expression, the imaginary unit is used to construct orthogonal components, making the original signal and its transformation result form a complex plane representation. The introduction of the imaginary part enables the signal to have phase analyzability. The Hilbert transform operator is a class of linear time-invariant operators used in the field of signal processing to construct analytic signals and extract instantaneous amplitude and instantaneous phase. Its essence is a singular integral transform of the original real signal with a 90-degree phase shift.

[0040] Calculation of instantaneous phase based on analytic signal: ; in, For instantaneous phase, Sampling time, Indicates the first One decomposition mode, To decompose the modal index values, Represents the Hilbert transform operator. The signal obtained after performing a Hilbert transform on the decomposed modes.

[0041] Further calculation of the instantaneous frequencies of the decomposed modes based on the instantaneous phase: ; in, Let be the instantaneous frequency of the k-th decomposition mode. The sampling time interval, For instantaneous phase, The instantaneous phase at the previous sampling time. To decompose the modal index value.

[0042] It should be noted that instantaneous phase may become coiled, i.e., from Jump to This can lead to negative or extreme values ​​in the difference. Therefore, when the instantaneous phase difference is negative, phase dewinding is performed to map the instantaneous phase difference to... The interval is used to avoid the instantaneous frequency being negative.

[0043] Instantaneous frequency represents the oscillation frequency of the decomposed mode at the current moment. The larger the value, the faster the mode changes, corresponding to a higher frequency voltage fluctuation component.

[0044] Based on the instantaneous frequency of each decomposition mode, the decomposition modes are classified into mode categories.

[0045] Specifically, the average instantaneous frequency of the decomposed mode within a preset time window is calculated. The average instantaneous frequency characterizes the dominant frequency characteristics of the decomposed mode, and its magnitude reflects the frequency range to which the decomposed mode belongs.

[0046] Modal categories are classified according to preset frequency ranges: When the average instantaneous frequency is less than the first division threshold, the decomposed mode is determined to be a low-frequency mode. The low-frequency mode reflects the slow voltage change caused by the fluctuation of new energy output. When the average instantaneous frequency is greater than or equal to the first division threshold and less than the second division threshold, the decomposed mode is determined to be an intermediate frequency mode, which reflects the flicker-like voltage modulation process. When the average instantaneous frequency is greater than or equal to the second dividing threshold, the decomposed mode is determined to be a high-frequency mode. The high-frequency mode reflects interharmonics or rapid voltage disturbances caused by power electronic devices.

[0047] It should be noted that the first and second dividing thresholds are used to segment the frequency scale of the decomposed modes, and their settings are based on a combination of voltage fluctuation mechanisms and spectral statistical characteristics. Specifically, firstly, spectral analysis is performed on the voltage signal under typical operating conditions to statistically analyze the energy distribution in different frequency ranges and identify the boundary ranges of the dominant frequency band of new energy power fluctuations, the flicker modulation frequency band, and the high-frequency disturbance frequency band. Based on this, the boundary frequency between low and medium frequencies is used as the first dividing threshold, and the boundary frequency between medium and high frequencies is used as the second dividing threshold.

[0048] Through the above processing, multi-scale decomposition of wideband signals was achieved, and each decomposed mode was quantitatively classified based on instantaneous frequency, providing basic data support for subsequent modal energy spectrum analysis and voltage fluctuation type identification.

[0049] In step S3, a historical analysis window is preset. The historical voltage data of the grid-connected substation under test is obtained through the substation monitoring database in the historical analysis window. This data is used to reflect the voltage change trajectory of the grid-connected substation under test over a period of time, including the positive sequence voltage amplitude sequence obtained by transforming the three-phase voltage effective value sequence with symmetrical components. The three-phase voltage RMS sequence is a time series composed of the RMS values ​​of phase A, phase B, and phase C voltages collected in chronological order. After performing symmetrical component transformation on the three-phase voltage RMS sequence, the positive sequence voltage amplitude is extracted and combined to obtain the positive sequence voltage amplitude sequence, which is used to eliminate the interference of interphase imbalance on subsequent analysis. For each positive sequence voltage amplitude in the positive sequence voltage amplitude sequence, if the current positive sequence voltage amplitude is greater than both the previous and the next positive sequence voltage amplitude, then the current positive sequence voltage amplitude is marked as the upper envelope feature point. Similarly, if the current positive sequence voltage amplitude is smaller than both the previous and the next positive sequence voltage amplitudes, then the current positive sequence voltage amplitude is marked as the lower envelope feature point. The upper envelope feature points are interpolated and connected in chronological order to generate the upper envelope trajectory, and the lower envelope feature points are interpolated and connected in chronological order to generate the lower envelope trajectory. Among them, the interpolation connection processing is based on the cubic spline interpolation continuous curve reconstruction method, which achieves smooth connection of envelope trajectory by constructing piecewise cubic polynomials between adjacent feature points.

[0050] It should be explained that the preset historical analysis window can be set according to the needs of voltage fluctuation analysis; the substation monitoring database is a data management system used to store substation operation monitoring data.

[0051] The upper and lower envelope trajectories are combined to form the voltage envelope trajectory, which reflects the amplitude variation boundary range of the voltage waveform within the historical analysis window. The upper envelope trajectory reflects the upper boundary variation trend of the voltage waveform, and the lower envelope trajectory reflects the lower boundary variation trend of the voltage waveform.

[0052] The average value of the positive sequence voltage amplitude is taken as the voltage center reference value. The values ​​in the upper envelope trajectory are successively subtracted from the voltage center reference value to obtain the positive offset amplitude. The voltage center reference value is successively subtracted from the values ​​in the lower envelope trajectory to obtain the negative offset amplitude. The average values ​​of the positive and negative offset amplitudes are taken respectively to obtain the average offset of the upper envelope and the average offset of the lower envelope. Using each sampling time in the historical analysis window as a time index, the upper envelope trajectory value and the lower envelope trajectory value of the corresponding time are extracted one by one. For the same sampling time, the difference operation is performed between the upper envelope trajectory value and the lower envelope trajectory value to obtain the envelope span. The average value of each envelope span is taken to obtain the average envelope span. The voltage envelope asymmetry is calculated based on the average offset of the upper and lower envelopes, using the following formula: ,in, This represents the average offset of the upper envelope. This represents the average offset of the lower envelope. For the average envelope span, Voltage envelope asymmetry; Voltage envelope asymmetry reflects the degree of deviation of the upper and lower envelope distribution of the voltage waveform within the historical analysis window. The larger the value, the more significant the difference in the offset between the upper and lower envelopes relative to the center voltage.

[0053] The voltage envelope asymmetry is standardized to obtain the envelope regulation coefficient, which reflects the modulation intensity of the asymmetric disturbance on the mode energy distribution. Preset class weight coefficients for different modal categories, and use the product of the preset class weight coefficients and the envelope adjustment coefficients as the perturbation enhancement energy for each modal category; The perturbation enhancement energy is used to reflect the enhancement effect of voltage envelope asymmetry on different mode categories. The larger the value, the more obvious the influence of asymmetric perturbation on the corresponding mode category during the current voltage fluctuation.

[0054] The decomposed modes of each modal category are subjected to frequency domain transformation to obtain a spectral amplitude sequence. In the spectral amplitude sequence, the spectral amplitude is divided according to a preset frequency interval, and the spectral amplitude in each frequency interval is squared and accumulated to obtain the modal energy value of the corresponding frequency interval. Among them, frequency domain transformation processing refers to performing discrete Fourier transform on the time domain signal of the decomposed mode; When a mode category contains multiple decomposed modes, the modal energy values ​​within the same frequency range are summed to obtain the total modal energy value of that mode category within that frequency range. The total modal energy values ​​corresponding to each frequency range are arranged in order of frequency from low to high to form the modal energy spectrum of the modal category; It should be noted that the standardization process includes, but is not limited to, standard linear transformation based on interval scaling, Z-Score standardization based on statistics, or normalization based on nonlinear mapping functions. The application methods of standardization will not be elaborated here. The preset category weight coefficients can be set according to the sensitivity of different modal categories to voltage fluctuations or the contribution of each modal category to the classification results in historical samples. The preset frequency range can be divided according to the frequency distribution range of the voltage fluctuation signal.

[0055] In step S4, the modal categories are arranged in the order of low-frequency modal category, mid-frequency modal category, and high-frequency modal category to obtain the modal category division order; The total modal energy value in the modal energy spectrum is multiplied by the perturbation enhancement energy of the corresponding modal category to obtain the corrected total modal energy value; The modal energy spectra of the modal categories are arranged according to the classification order, and the corrected total modal energy values ​​in the modal energy spectra are arranged in order to form modal modulation characteristics; The machine learning model is retrieved through the model call interface. The machine learning model is a pre-trained support vector machine model. It's important to explain that the machine learning model is a classification model trained offline based on historical sample data. During the training phase, historical mode modulation features are first extracted from historical operational data. The voltage fluctuation types corresponding to these historical mode modulation features are then manually labeled to form a training sample set. This training sample set is input into the machine learning model, and the classification decision boundary is determined by iteratively optimizing the classification margin, thereby obtaining a classification model capable of distinguishing different voltage fluctuation types.

[0056] The modal modulation features of the current grid-connected substation under test are input into the machine learning model. The machine learning model maps the modal modulation features according to the classification decision boundary formed during the training phase and outputs the corresponding voltage fluctuation type identification result.

[0057] The voltage fluctuation type identification results include sinusoidal modulation fluctuations, rectangular modulation fluctuations, and interharmonic modulation fluctuations. The voltage fluctuation type identification results are correlated with the corresponding modal modulation characteristics and stored in the substation monitoring database.

[0058] It should be noted that the model call interface is a data interaction interface used to realize the input of feature data and the output of classification results. Its input is modal modulation features, and its output is the corresponding voltage fluctuation type identification result.

[0059] Finally, it should be noted that in this paper, relational terms such as first and second are used only 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.

[0060] Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0061] In this document, the singular forms “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that terms such as “comprising / including” or “having” specify the presence of the stated features, integrals, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, integrals, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0062] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0063] The above description of the disclosed embodiments will enable those skilled in the art to make or use various modifications to these embodiments. It will be readily apparent to those skilled in the art that the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion, characterized in that: Includes the following steps: Step S1: Monitor the grid-connected substation under test, collect the three-phase voltage data of the grid-connected substation under test, calculate the voltage deviation rate based on the three-phase voltage data, access the power grid operation platform to retrieve the short-time flicker index, and determine whether the fluctuation identification mechanism is triggered based on the voltage deviation rate. Step S2: In the fluctuation identification mechanism, the wideband signal of the grid-connected substation under test is acquired, the wideband signal is processed by mode decomposition, the instantaneous frequency of each decomposed mode is monitored, and the mode category of the decomposed mode is classified. Step S3: Obtain historical voltage data of the grid-connected substation under test, generate voltage envelope trajectory based on historical voltage data and calculate voltage envelope asymmetry, set disturbance enhancement energy for each mode category based on voltage envelope asymmetry, and detect modal energy spectrum for each mode category; Step S4: Generate the mode modulation characteristics of the grid-connected substation under test by integrating the mode energy spectrum and the disturbance enhancement energy, call the machine learning model to perform feature mapping processing on the mode modulation characteristics, and identify the voltage fluctuation type of the grid-connected substation under test based on the processing results.

2. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 1, characterized in that: In step S1, a voltage transformer is installed at the grid connection point of the substation to be tested, and the three-phase voltage signal is synchronously acquired at a preset sampling frequency to obtain three-phase voltage data. The three-phase voltage data consists of real-time measurements of the voltage to ground for each phase, including the instantaneous sampled values ​​of phase A voltage, phase B voltage, and phase C voltage. Based on the three-phase voltage data, the equivalent root mean square value of the three-phase voltage is calculated to obtain the equivalent voltage value; Access the power grid operation platform to retrieve the rated voltage value corresponding to the grid-connected substation under test. The rated voltage value is the standard operating voltage specified in the power grid plan.

3. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 2, characterized in that: In step S1, the voltage deviation rate is calculated based on the equivalent voltage value and the rated voltage value; The short-time flicker index of the grid-connected substation under test is retrieved from the power grid operation platform within the current time window. The short-time flicker index is a voltage fluctuation sensing index obtained based on a preset time scale. The voltage deviation rate and short-time flicker index are combined for judgment. When the voltage deviation rate is greater than the preset deviation threshold and the short-time flicker index is greater than the preset flicker threshold, the current grid-connected substation under test is determined to be in a state of abnormal voltage fluctuation, triggering the fluctuation identification mechanism. Otherwise, the fluctuation identification mechanism will not be triggered, and the normal monitoring status will be maintained.

4. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 1, characterized in that: In step S2, after the fluctuation identification mechanism is triggered, the equivalent voltage value calculated from the sequence of instantaneous three-phase voltage samples output by the voltage transformer at the grid connection point is used as the three-phase composite voltage signal. The three-phase composite voltage signal is subjected to mean removal processing to obtain a wide frequency domain signal; The empirical mode decomposition method is used to perform mode decomposition processing on the broadband signal to obtain several decomposed modes. Each decomposed mode represents the oscillation component of the broadband signal at a specific frequency scale. The analytic signal of each decomposed mode is constructed using Hilbert transform, the instantaneous phase is calculated based on the analytic signal, and the instantaneous frequency of the decomposed mode is calculated based on the instantaneous phase.

5. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 4, characterized in that: In step S2, the average instantaneous frequency of the decomposed mode within a preset time window is calculated. The average instantaneous frequency characterizes the dominant frequency characteristics of the decomposed mode. When the average instantaneous frequency is less than the preset first division threshold, the decomposed mode is determined to be a low-frequency mode; When the average instantaneous frequency is greater than or equal to the preset first division threshold and less than the preset second division threshold, the decomposed mode is determined to be a mid-frequency mode. When the average instantaneous frequency is greater than or equal to the preset second division threshold, the decomposed mode is determined to be a high-frequency mode.

6. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 1, characterized in that: In step S3, a historical analysis window is preset, and historical voltage data of the grid-connected substation under test is obtained through the substation monitoring database within the historical analysis window, including the positive sequence voltage amplitude sequence. For each three-phase voltage effective value in the positive sequence voltage amplitude sequence, if the current positive sequence voltage amplitude is greater than both the previous and the next positive sequence voltage amplitude, then the current positive sequence voltage amplitude is marked as the upper envelope feature point. Similarly, if the current positive sequence voltage amplitude is smaller than both the previous and the next positive sequence voltage amplitudes, then the current positive sequence voltage amplitude is marked as the lower envelope feature point. The voltage envelope trajectory includes an upper envelope trajectory and a lower envelope trajectory. The upper envelope trajectory is generated by interpolating and connecting the feature points of the upper envelope in chronological order, and the lower envelope trajectory is generated by interpolating and connecting the feature points of the lower envelope in chronological order.

7. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 6, characterized in that: In step S3, the average value of the positive sequence voltage amplitude is used as the voltage center reference value, and the positive offset amplitude between the upper envelope trajectory and the voltage center reference value and the negative offset amplitude between the lower envelope trajectory and the voltage center reference value are calculated respectively. The average values ​​of the positive and negative offset amplitudes are calculated to obtain the average offset of the upper envelope and the average offset of the lower envelope. The average envelope span is obtained by comprehensively calculating the values ​​of the upper and lower envelope trajectories. The voltage envelope asymmetry is calculated by combining the average offset of the upper envelope, the average offset of the lower envelope, and the average envelope span. The voltage envelope asymmetry is standardized to obtain the envelope adjustment coefficient, and a preset category weight coefficient is applied to different mode categories. The product of the preset category weight coefficient and the envelope adjustment coefficient is used as the perturbation enhancement energy for each modality.

8. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 4, characterized in that: In step S3, the decomposed modes of each mode category are transformed in the frequency domain to obtain a spectrum amplitude sequence. In the spectrum amplitude sequence, the spectrum amplitude is divided according to a preset frequency range. The modal energy values ​​of the frequency range are obtained by squaring and summing the spectral amplitudes within the preset frequency range. When a modal category contains multiple decomposed modes, the modal energy values ​​within the same frequency range are summed to obtain the total modal energy value of the modal category in each frequency range; The total modal energy values ​​of each frequency range are arranged in order of increasing frequency to form the modal energy spectrum of the modal category.

9. The method for identifying broadband voltage fluctuation types in power grids based on multi-feature fusion according to claim 5, characterized in that: In step S4, the mode categories are arranged in the order of low-frequency mode, mid-frequency mode, and high-frequency mode to obtain the mode category classification order; The total modal energy spectrum is multiplied by the perturbation enhancement energy of the corresponding modal category to obtain the corrected total modal energy value; The modal energy spectra of the modal categories are arranged according to the classification order, and the corrected total modal energy values ​​in the modal energy spectra are arranged in order to form modal modulation characteristics; The machine learning model is invoked through the model call interface. The modal modulation features of the grid-connected substation under test are input into the machine learning model. The machine learning model maps the modal modulation features according to the classification decision boundary formed during the training phase and outputs the corresponding voltage fluctuation type identification result. The voltage fluctuation type identification results include sinusoidal modulation fluctuation, rectangular modulation fluctuation, and interharmonic modulation fluctuation.

Images