A novel power system wideband harmonic event detection method

Abstract

The application discloses a novel power system broadband harmonic event detection method, according to the disturbance mechanism of new energy power electronic equipment, a physical heuristic harmonic event atom library containing steady-state harmonic atoms, dynamic inter-harmonic atoms, oscillation-type harmonic atoms and transient harmonic atoms is constructed. A double-sparse space joint decomposition model is established, and abnormal pollution and real physical disturbance in the monitoring data are mapped to different sparse spaces for synchronous separation, realizing integrated processing of data quality enhancement and disturbance component extraction. According to the activated physical atom type, the dynamic characteristic parameters are adaptively extracted and integrated into a structured broadband harmonic event fingerprint, realizing broadband harmonic event detection. The application can realize credible detection of broadband harmonic disturbance and event-level semantic description under the condition of low-quality monitoring data, and is suitable for engineering detection scenes of new energy stations, power electronic power grids and intelligent power quality monitoring systems.

Description

Technical Field

[0001] This invention relates to the technical field of power quality analysis and signal processing in new power systems, and in particular to a novel power system broadband harmonic event detection method, system, storage medium and computing device based on structured sparse representation. It is designed for scenarios with a high proportion of new energy access and addresses the broadband harmonic dynamic disturbances generated by power electronic devices by using physical heuristic structured sparse representation to achieve data quality enhancement, signal decomposition and event-level detection. Background Technology

[0002] With the large-scale integration of new energy sources and power electronic equipment, a large number of broadband harmonic disturbance signals with frequency drift, amplitude variation, and transient characteristics have emerged in new power systems. These disturbances include not only traditional integer harmonics, but also interharmonics, sub / supersynchronous oscillations, and PWM switching sideband harmonics. The dynamic interaction of control loops in grid-connected converters, energy storage PCS, and other equipment makes the disturbances exhibit significant characteristics of non-stationarity, broadband, and multiple types. At the same time, field monitoring data generally suffers from low-quality problems such as noise pollution, missing data, asynchronous sampling, and abnormal spikes.

[0003] Existing harmonic detection methods face the following technical limitations in field applications:

[0004] 1. Fourier analysis methods are only applicable to steady-state signals. The Fast Fourier Transform and its improved windowed interpolation algorithm assume that the signal has stable frequency characteristics within the analysis window, which cannot describe non-stationary dynamic processes such as interharmonic frequency drift and oscillation mode evolution. Furthermore, it suffers from spectral leakage and picket fence effects under asynchronous sampling conditions.

[0005] 2. Time-frequency analysis methods lack physical interpretability. The selection of basis functions for wavelet transform relies on human experience, and different basis functions have a significant impact on the detection results; empirical mode decomposition and its variants suffer from mode aliasing and parameter sensitivity issues, making it difficult to map the decomposition results to a clear physical perturbation type; the number of modes and penalty factor in variational mode decomposition need to be preset, and the optimal parameters vary greatly under different operating conditions.

[0006] 3. Compressed sensing methods lack power system semantics. Existing sparse representation methods use general mathematical dictionaries such as Fourier basis, DCT basis, or wavelet basis. Although they can achieve sparse signal reconstruction, the dictionary atoms do not have the physical meaning of the power system and cannot distinguish the physical source of different types of disturbances. The detection result is only a set of mathematical coefficients rather than a description of physical events.

[0007] 4. The results of deep learning methods are not interpretable. Data-driven classification models rely on a large number of labeled samples, and their generalization ability under extreme conditions of new energy systems is questionable. Furthermore, their black-box nature makes the detection results unreliable as a basis for grid connection testing and fault diagnosis.

[0008] The common drawback of the above methods is that the output is essentially still a list of "frequency-amplitude" parameters, which cannot answer the core cognitive question of "what physical event is happening in the system at this moment"; and data cleaning and harmonic analysis are usually performed as separate independent steps, which can easily lead to the accidental filtering out of weak disturbances in the preprocessing stage, resulting in the irreversible loss of key dynamic information.

[0009] The core problem to be solved is: under the condition of low-quality monitoring data, how to decompose broadband harmonic disturbances into independent event components with clear physical meaning, and to provide a structured description of the dynamic evolution process of each component, so as to achieve the improvement from "spectrum detection" to "event cognition".

[0010] Therefore, there is an urgent need for a technical solution that can decompose broadband harmonic disturbances into independent event components with clear physical meaning under low-quality monitoring data conditions, and provide a structured description of the dynamic evolution process of each component. Summary of the Invention

[0011] The primary objective of this invention is to provide a novel broadband harmonic event detection method for power systems based on structured sparse representation. This method replaces the general mathematical basis of traditional signal processing with a physically heuristic harmonic event atom library that maps the mechanisms of power electronic disturbances. Through joint decomposition of dual sparse spaces and physical atom type-driven parsing, it achieves a cognitive closed loop from low-quality data to interpretable events. This invention enables reliable detection and event-level semantic description of broadband harmonic disturbances under low-quality monitoring data conditions, and is applicable to engineering detection scenarios in new energy power plants, power electronic grids, and smart power quality monitoring systems.

[0012] The second objective of this invention is to provide a novel broadband harmonic event detection system for power systems based on structured sparse representation.

[0013] A third objective of this invention is to provide a storage medium.

[0014] A fourth objective of this invention is to provide a computing device.

[0015] The first objective of this invention is achieved through the following technical solution: a novel method for detecting broadband harmonic events in a power system, comprising:

[0016] S1. Obtain broadband harmonic monitoring data containing broadband harmonic disturbances;

[0017] S2. Based on the broadband harmonic disturbance mechanism of new energy power electronic equipment, establish a structured overcomplete dictionary with clear physical mapping relationships that includes multiple physical atom sub-dictionaries;

[0018] S3. Establish a dual sparse space joint decomposition model that includes physical disturbance sparse space and data anomaly sparse space to decompose the broadband harmonic monitoring data and decompose the data anomaly component and the physical disturbance component belonging to the structured overcomplete dictionary.

[0019] S4. Based on the physical atom type matched by the physical perturbation component and its corresponding physical atom sub-dictionary, perform adaptive parameter parsing to obtain the corresponding dynamic feature parameters;

[0020] S5. Integrate the dynamic feature parameters to generate a structured broadband harmonic event fingerprint containing physical semantics. The structured broadband harmonic event fingerprint is a unified event description vector constructed based on the dynamic feature parameters obtained from the analysis.

[0021] S6. Based on the structured broadband harmonic event fingerprint, determine and output the broadband harmonic event detection result.

[0022] Furthermore, the broadband harmonic monitoring data, which includes broadband harmonic disturbances, is obtained as follows:

[0023] At new energy power plants or grid connection points, broadband voltage and current signals are collected at a set sampling frequency as observation signals to be processed. The observation signals contain one or more of the following broadband harmonic disturbances: a) random noise introduced by sensor thermal noise or electromagnetic interference; b) data loss caused by communication packet loss or storage failure; c) time-domain sparse abnormal spikes introduced by transient interference or equipment switching.

[0024] Furthermore, the structured overcomplete dictionary consists of the following four sets of atoms generated based on a perturbation evolution model, each set of atoms corresponding to a different perturbation mechanism:

[0025] ;

[0026] In the formula, This is a structured, overcomplete dictionary, referred to as the physical heuristic harmonic event atom library; The set of steady-state harmonic atoms that describes a fixed-frequency harmonic is called the steady-state harmonic atom sub-dictionary; The set of dynamic interharmonic atoms that describes the perturbation of frequency over time is called the dynamic interharmonic atom sub-dictionary; The set of oscillating harmonic atoms that describes oscillations with damped or negatively damped characteristics is called the oscillating harmonic atom sub-dictionary; The set of transient harmonic atoms used to describe the mutation process is called the transient harmonic atom sub-dictionary.

[0027] Furthermore, the physical descriptions and parameterization models of each sub-dictionary are as follows:

[0028] Steady-state harmonic atomic dictionary : Used to describe the periodic perturbation components of power electronic devices, whose atoms are constructed using a cosine parameterized model:

[0029] ;

[0030] In the formula, For steady-state harmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; The initial amplitude, These are harmonic frequencies, corresponding to integer multiples of the fundamental frequency. This is the initial phase;

[0031] Dynamic Interharmonic Atomic Dictionary This is used to map the time-varying frequency disturbance components caused by phase-locked loop dynamics, weak grid interactions, or maximum power point tracking disturbances. Its atoms are constructed using a time-varying spectral parameterization model.

[0032] ;

[0033] In the formula, For the dynamic interharmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; For the time-varying trajectory of amplitude, For the center frequency, The rate of change of frequency characterizes the speed and direction of frequency drift.

[0034] Oscillating Harmonic Atomic Dictionary : Used to map the damped evolution component of control instability in grid-type converters, or subsynchronous or supersynchronous oscillations in new energy power plants. Its atoms are constructed using a decaying sinusoidal parameterized model.

[0035] ;

[0036] In the formula, For oscillatory harmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; The attenuation coefficient is... Corresponding to damped oscillation, Corresponding to divergent oscillation, Corresponding to constant amplitude oscillation, The oscillation frequency;

[0037] Transient Harmonic Atomic Dictionary Used for mapping short-term burst disturbance components caused by switching operations, fault clearing, and control mode switching; its atoms are constructed using a decay-modulated wavelet parameterization model.

[0038] ;

[0039] In the formula, Transient Harmonic Atomic Sub-Dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; For decay rate, These are wavelet basis functions with compact support properties;

[0040] Each sub-dictionary generates a set of redundant atoms by discretizing its own parameter space, forming an overcomplete representation basis covering typical broadband harmonic disturbance types; the discretization range and step size of each sub-dictionary parameter are determined according to the frequency domain resolution and time domain positioning accuracy requirements of the corresponding physical disturbance type.

[0041] Furthermore, the dual sparse space joint decomposition model represents the observed signal as a superposition of physical perturbation components, data anomaly components, and random noise:

[0042] ;

[0043] In the formula, For observation signals; For a structured overcomplete dictionary; It is a sparse coefficient vector; For the structured overcomplete dictionary The physical disturbance component; These are time-domain sparse data anomalous components, including spikes and missing data segments; It is a random noise component; This represents the number of sampling points; Represents the set of real numbers; introduces outlier components by explicitly specifying them. This incorporates data quality issues into a unified mathematical framework, making data augmentation an integral part of the perturbation decomposition process.

[0044] Furthermore, the joint decomposition model of the dual sparse space solves for the following joint optimization objective function:

[0045] ;

[0046] In the formula, To constrain the sparsity of the physical perturbation components, the first norm term is used. To constrain the sparsity of outlier components in the data, the second norm term is used. For signal reconstruction fidelity; This is an abnormal sparsity regularization parameter used to control the sparsity of abnormal data components; This is a data fidelity regularization parameter used to balance the signal reconstruction accuracy with the sparsity of the solution; where, with norm Constrained sparse coefficient vector This reflects the prior knowledge that actual physical events have sparse representations in the physical heuristic harmonic event atom library, with norm Constraint data outliers This reflects the prior that the spikes and gaps in the observed signal have a sparse distribution in the time domain.

[0047] Furthermore, the dual sparse space joint decomposition model is solved iteratively using either the alternating direction multiplier method or the block coordinate descent method, with the sparse coefficient vector being updated alternately in each iteration. Data anomaly components Using Lagrange multipliers, the solution is obtained by solving the problem until the convergence condition is met, thus obtaining the purified physical perturbation components. and the separated data anomaly components The abnormal component They are explicitly separated into independent sparse spaces and no longer mixed with physical perturbation components. This process allows for data purification while preserving complete broadband harmonic dynamic information.

[0048] Furthermore, the adaptive parameter parsing is performed based on the sparse coefficient vector. The positions of non-zero elements are determined, the activated atoms and their corresponding sub-dictionaries are located, and an adaptive mapping rule for the "physical atom type - parameter resolution strategy" is established:

[0049] When the activated atom belongs to the dynamic interharmonic atom sub-dictionary At that time, a time-varying frequency parameter estimation method is adopted, which uses Taylor series to model the time-varying amplitude and phase of the physical disturbance component, and extracts the instantaneous frequency trajectory through weighted least squares estimation. Rate of change of frequency and amplitude time-varying trajectory The time-varying frequency parameter estimation method is implemented using the Taylor-Fourier expansion method, the synchronous compression transform method, or the Hilbert transform method.

[0050] When the activated atom belongs to the oscillating harmonic atom sub-dictionary At that time, the exponential decay mode parameter estimation method is used to construct the Hankel matrix for the physical disturbance components and perform singular value decomposition to extract the damping factor. and modal frequencies The exponential decay mode parameter estimation method is implemented using a matrix bundle method, the Prony method, or a rotation-invariant technique signal parameter estimation method.

[0051] When the activated atom belongs to the steady-state harmonic atom sub-dictionary In this process, a high-resolution spectrum estimation method is used to extract the steady-state amplitude and phase parameters; wherein, the high-resolution spectrum estimation method is implemented using a rotation-invariant signal parameter estimation method or a multiple signal classification method;

[0052] When the activated atom belongs to the transient harmonic atom sub-dictionary At that time, the transient duration was extracted using time-frequency analysis. and energy distribution characteristics.

[0053] Furthermore, the structured broadband harmonic event fingerprint adopts the following data model:

[0054] ;

[0055] In the formula, For structured broadband harmonic event fingerprinting; The event physical category includes at least steady-state harmonic events, dynamic interharmonic events, oscillating harmonic events, and transient harmonic events; The dominant frequency; The frequency drift trajectory over time; The amplitude is a time-varying trajectory; It is a set of damping characteristic parameters, which at least includes the damping factor and the damping ratio; Duration of the event; To test the confidence level.

[0056] Furthermore, the detection confidence level Based on reconstruction residual calculation:

[0057] ;

[0058] In the formula, To reconstruct the residuals Norm, representing the residual energy between the observed signal and the physical disturbance components and data anomaly components; For the observation signal Norm, representing the total energy of the observed signal; where, The range of values ​​is The closer the value is to 1, the more reliable the detection result; when the detection confidence level... When the result is below a preset threshold, the detection result is marked as low confidence, indicating that higher quality data or manual review is needed.

[0059] Finally, based on the physical category and dynamic characteristic parameters of the event in the structured broadband harmonic event fingerprint, the broadband harmonic event detection result, i.e., the broadband harmonic disturbance event type, is determined and output.

[0060] The second objective of this invention is achieved through the following technical solution: a novel power system broadband harmonic event detection system, used to implement the aforementioned novel power system broadband harmonic event detection method, comprising:

[0061] The data acquisition module is used to acquire broadband harmonic monitoring data that includes broadband harmonic disturbances;

[0062] The dictionary construction module is used to build a structured, overcomplete dictionary with clear physical mapping relationships, containing multiple physical atom sub-dictionaries, based on the broadband harmonic disturbance mechanism of new energy power electronic equipment.

[0063] The joint decomposition module is used to establish a dual sparse space joint decomposition model that includes a physical disturbance sparse space and a data anomaly sparse space, and decompose the broadband harmonic monitoring data to decompose the data anomaly component and the physical disturbance component belonging to the structured overcomplete dictionary.

[0064] The parameter parsing module is used to perform adaptive parameter parsing to obtain the corresponding dynamic feature parameters based on the physical atom type matched by the physical perturbation component and its corresponding physical atom sub-dictionary.

[0065] The fingerprint generation module is used to integrate the dynamic feature parameters to generate a structured broadband harmonic event fingerprint containing physical semantics;

[0066] The event recognition module is used to determine and output the broadband harmonic event detection result based on the structured broadband harmonic event fingerprint.

[0067] The third objective of this invention is achieved through the following technical solution: a storage medium storing a program, which, when executed by a processor, implements the novel power system broadband harmonic event detection method described above.

[0068] The fourth objective of this invention is achieved through the following technical solution: a computing device, including a processor and a memory for storing processor-executable programs, wherein when the processor executes the program stored in the memory, it implements the above-mentioned novel power system broadband harmonic event detection method.

[0069] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0070] 1. By constructing a physical heuristic harmonic event atom library with clear physical mapping relationships, the signal representation is elevated from a general mathematical base to an event description language with physical semantics, enabling the detection results to be transformed from a spectrum list into a structured event report containing event types and dynamic features.

[0071] 2. By using a dual sparse space joint decomposition model, data anomalies and physical disturbances are mapped to different sparse spaces for simultaneous solution. Data cleaning is no longer an independent preprocessing step. While separating data anomaly components, the complete broadband harmonic dynamic information is preserved, reducing the risk of weak disturbances being mistakenly filtered out during the preprocessing stage.

[0072] 3. By adopting an adaptive parameter analysis strategy based on physical atom types, corresponding parameter analysis methods are matched for different physical events, which can meet the extraction needs of various types of dynamic features such as frequency drift and damping evolution.

[0073] 4. By generating structured broadband harmonic event fingerprints, a standardized input data format is provided for subsequent analyses such as disturbance source tracing and risk assessment. Attached Figure Description

[0074] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention.

[0075] Figure 2 A schematic diagram illustrating the structured composition of a physics-inspired harmonic event atom library.

[0076] Figure 3 A flowchart illustrating the process of enhancing data quality through joint decomposition of two sparse spaces.

[0077] Figure 4 A schematic diagram of adaptive parameter parsing and event fingerprint generation driven by physical atom types. Detailed Implementation

[0078] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0079] Example 1

[0080] like Figures 1 to 4 As shown in the figure, this embodiment discloses a novel broadband harmonic event detection method for power systems based on structured sparse representation, the specific details of which are as follows:

[0081] S1. Obtain low-quality broadband harmonic monitoring data containing broadband harmonic disturbances.

[0082] At new energy power plants or grid connection points, broadband voltage and current signals are collected at a set sampling frequency through power quality monitoring terminals as observation signals to be processed. The observation signals contain one or more of the following broadband harmonic disturbances: a) random noise introduced by sensor thermal noise or electromagnetic interference; b) data loss caused by communication packet loss or storage failure; c) time-domain sparse abnormal spikes introduced by transient interference or equipment switching.

[0083] S2. Based on the broadband harmonic disturbance mechanism of new energy power electronic equipment, establish a structured overcomplete dictionary with clear physical mapping relationships, which includes a variety of physical atom sub-dictionaries.

[0084] The structured overcomplete dictionary consists of the following four sets of atoms generated based on a perturbation evolution model, each set of atoms corresponding to a different perturbation mechanism:

[0085] ;

[0086] In the formula, This is a structured, overcomplete dictionary, referred to as the physical heuristic harmonic event atom library; The set of steady-state harmonic atoms that describes a fixed-frequency harmonic is called the steady-state harmonic atom sub-dictionary; The set of dynamic interharmonic atoms that describes the perturbation of frequency over time is called the dynamic interharmonic atom sub-dictionary; The set of oscillating harmonic atoms that describes oscillations with damped or negatively damped characteristics is called the oscillating harmonic atom sub-dictionary; The set of transient harmonic atoms used to describe the mutation process is called the transient harmonic atom sub-dictionary.

[0087] The physical descriptions and parameterization models of each sub-dictionary are as follows:

[0088] Steady-state harmonic atomic dictionary : Used to describe the periodic perturbation components of power electronic devices, whose atoms are constructed using a cosine parameterized model:

[0089] ;

[0090] In the formula, For steady-state harmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; The initial amplitude, These are harmonic frequencies, corresponding to integer multiples of the fundamental frequency. This is the initial phase;

[0091] Dynamic Interharmonic Atomic Dictionary This is used to map the time-varying frequency disturbance components caused by phase-locked loop dynamics, weak grid interactions, or maximum power point tracking disturbances. Its atoms are constructed using a time-varying spectral parameterization model.

[0092] ;

[0093] In the formula, For the dynamic interharmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; For the time-varying trajectory of amplitude, For the center frequency, The rate of change of frequency characterizes the speed and direction of frequency drift.

[0094] Oscillating Harmonic Atomic Dictionary : Used to map the damped evolution component of control instability in grid-type converters, or subsynchronous or supersynchronous oscillations in new energy power plants. Its atoms are constructed using a decaying sinusoidal parameterized model.

[0095] ;

[0096] In the formula, For oscillatory harmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; The attenuation coefficient is... Corresponding to damped oscillation, Corresponding to divergent oscillation, Corresponding to constant amplitude oscillation, The oscillation frequency;

[0097] Transient Harmonic Atomic Dictionary Used for mapping short-term burst disturbance components caused by switching operations, fault clearing, and control mode switching; its atoms are constructed using a decay-modulated wavelet parameterization model.

[0098] ;

[0099] In the formula, Transient Harmonic Atomic Sub-Dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; For decay rate, These are wavelet basis functions with compact support properties;

[0100] Each sub-dictionary generates a set of redundant atoms by discretizing its own parameter space, forming an overcomplete representation basis covering typical broadband harmonic disturbance types; the discretization range and step size of each sub-dictionary parameter are determined according to the frequency domain resolution and time domain positioning accuracy requirements of the corresponding physical disturbance type.

[0101] S3. Establish a dual sparse space joint decomposition model that includes physical disturbance sparse space and data anomaly sparse space to decompose the broadband harmonic monitoring data, and decompose the data anomaly component and the physical disturbance component belonging to the structured overcomplete dictionary.

[0102] The dual sparse space joint decomposition model represents the observed signal as a superposition of physical perturbation components, data anomaly components, and random noise:

[0103] ;

[0104] In the formula, For observation signals; For a structured overcomplete dictionary; It is a sparse coefficient vector; For the structured overcomplete dictionary The physical disturbance component; These are time-domain sparse data anomalous components, including spikes and missing data segments; It is a random noise component; This represents the number of sampling points; Represents the set of real numbers;

[0105] Compared with traditional sparse representation models The only difference between the perturbation and noise terms is that the dual sparse space joint decomposition model explicitly introduces outlier components. This incorporates data quality issues into a unified mathematical framework, making data augmentation an integral part of the perturbation decomposition process.

[0106] The joint decomposition model of the two sparse spaces solves for the following joint optimization objective function:

[0107] ;

[0108] In the formula, To constrain the sparsity of the physical perturbation components, the first norm term is used. To constrain the sparsity of outlier components in the data, the second norm term is used. For signal reconstruction fidelity; This is an abnormal sparsity regularization parameter used to control the sparsity of abnormal data components; This is a data fidelity regularization parameter used to balance the signal reconstruction accuracy with the sparsity of the solution; The settings are based on the on-site measured noise level and model accuracy requirements; among them, with norm Constrained sparse coefficient vector This reflects the prior knowledge that actual physical events have sparse representations in the physical heuristic harmonic event atom library, with norm Constraint data outliers This reflects the prior that the spikes and gaps in the observed signal have a sparse distribution in the time domain.

[0109] The dual sparse space joint decomposition model is solved iteratively using either the alternating direction multiplier method or the block coordinate descent method, with the sparse coefficient vector being updated alternately in each iteration. Data anomaly components Using Lagrange multipliers, the solution is obtained by solving the problem until the convergence condition is met, thus obtaining the purified physical perturbation components. and the separated data anomaly components This completes the separation of data anomalies and the extraction of broadband harmonic physical components; wherein, the anomaly components They are explicitly separated into independent sparse spaces and no longer mixed with physical perturbation components. This process allows for data purification while preserving complete broadband harmonic dynamic information.

[0110] S4. Based on the physical atom type matched by the physical perturbation component and its corresponding physical atom sub-dictionary, perform adaptive parameter parsing to obtain the corresponding dynamic feature parameters.

[0111] The adaptive parameter parsing is performed based on the sparse coefficient vector. The positions of non-zero elements are determined, the activated atoms and their corresponding sub-dictionaries are located, and an adaptive mapping rule for the "physical atom type - parameter resolution strategy" is established:

[0112] When the activated atom belongs to the dynamic interharmonic atom sub-dictionary At that time, a time-varying frequency parameter estimation method is adopted, which uses Taylor series to model the time-varying amplitude and phase of the physical disturbance component, and extracts the instantaneous frequency trajectory through weighted least squares estimation. Rate of change of frequency and amplitude time-varying trajectory The time-varying frequency parameter estimation method can be implemented using the Taylor-Fourier expansion method, the synchronous compression transform method, or the Hilbert transform method.

[0113] When the activated atom belongs to the oscillating harmonic atom sub-dictionary At that time, the exponential decay mode parameter estimation method is used to construct the Hankel matrix for the physical disturbance components and perform singular value decomposition to extract the damping factor. and modal frequencies The method for estimating the parameters of the exponentially decaying mode can be implemented using the matrix bundle method, the Prony method, or a rotation-invariant signal parameter estimation method.

[0114] When the activated atom belongs to the steady-state harmonic atom sub-dictionary At that time, a high-resolution spectrum estimation method is used to extract the steady-state amplitude and phase parameters; wherein, the high-resolution spectrum estimation method can be implemented by a rotation-invariant signal parameter estimation method or a multiple signal classification method;

[0115] When the activated atom belongs to the transient harmonic atom sub-dictionary At that time, the transient duration was extracted using time-frequency analysis. and energy distribution characteristics.

[0116] S5. Integrate the dynamic feature parameters to generate a structured broadband harmonic event fingerprint containing physical semantics. The structured broadband harmonic event fingerprint is a unified event description vector constructed based on the dynamic feature parameters obtained from the analysis, using the following data model:

[0117] ;

[0118] In the formula, For structured broadband harmonic event fingerprinting; The event physical category includes at least steady-state harmonic events, dynamic interharmonic events, oscillating harmonic events, and transient harmonic events; The dominant frequency; The frequency drift trajectory over time; The amplitude is a time-varying trajectory; It is a set of damping characteristic parameters, which at least includes the damping factor and the damping ratio; Duration of the event; To test the confidence level.

[0119] The detection confidence level Based on reconstruction residual calculation:

[0120] ;

[0121] In the formula, To reconstruct the residuals Norm, representing the residual energy between the observed signal and the physical disturbance components and data anomaly components; For the observation signal Norm, representing the total energy of the observed signal; where, The range of values ​​is The closer the value is to 1, the more reliable the detection result; when the detection confidence level... When the result is below a preset threshold, the detection result is marked as low confidence, indicating that higher quality data or manual review is needed.

[0122] S6. Based on the physical category and dynamic characteristic parameters of the event in the structured broadband harmonic event fingerprint, determine and output the broadband harmonic event detection result, i.e., the broadband harmonic disturbance event type.

[0123] To further verify the effectiveness and reliability of the novel broadband harmonic event detection method for power systems described in this embodiment, a new energy power station is used as an example. This station is equipped with a grid-connected energy storage converter and employs a virtual synchronous machine control strategy. The station's grid connection point is in normal operation. Maintenance personnel need to perform online monitoring and event diagnosis of the broadband harmonic disturbance characteristics of the converter after grid connection to assess the dynamic interaction status of its control loop and potential oscillation risks.

[0124] Execution step S1: Obtain low-quality broadband harmonic monitoring data containing broadband harmonic disturbances.

[0125] At the power station's grid connection point, continuous instantaneous sampling data of the converter output voltage and current are collected via a power quality monitoring terminal at a set sampling frequency. The sampling frequency must cover broadband harmonic components near the switching frequency of the power electronic equipment, while also meeting the sampling requirements for the dynamic evolution of interharmonics and subsynchronous oscillations. The collected monitoring data includes random noise components introduced by sensor background noise and electromagnetic interference, as well as sporadic data gaps caused by transient packet loss in the communication link. The collected signals are used as observation signals. .

[0126] Step S2: Construct a physical heuristic harmonic event atom library.

[0127] Based on the typical disturbance mechanism of virtual synchronous machine control in grid-type converters, a structured overcomplete dictionary (i.e., a physical heuristic harmonic event atom library) is pre-constructed. .

[0128] Steady-state harmonic atomic dictionary Covering the frequency range from the fundamental frequency to typical harmonic orders, the frequency discretization step size corresponds to the fundamental frequency. Dynamic interharmonic atom dictionary. Covering the typical frequency band from subsynchronous to low harmonics, center frequency Discretize according to the set step size, frequency change rate The coverage area corresponds to the typical frequency drift rate of PLL bandwidth and weak network interaction. Oscillating harmonic atom dictionary. It covers the typical frequency band from subsynchronous to mid-frequency harmonics, with an oscillation frequency. Discretize according to the set step size, attenuation coefficient Covers a typical range from strong attenuation to weak divergence. Transient harmonic atom dictionary. A family of wavelets is used to construct transient processes at different time scales.

[0129] Execution step S3: Perform joint decomposition of the two sparse spaces.

[0130] The observed signal is represented as Regularization parameters and Based on the on-site noise level and model accuracy requirements, the following joint optimization objective function is solved:

[0131] ;

[0132] An iterative solution is performed using the alternating direction multiplier method, with a set convergence tolerance and a maximum number of iterations.

[0133] Upon completion of the solution, the following is obtained simultaneously: sparse coefficient vector Indicates the activated physical atoms and their corresponding sub-dictionaries; reconstructed physical perturbation components. ; Separated data anomaly components .

[0134] Execution step S4: Physical atom-driven adaptive parameter parsing.

[0135] Based on the sparse coefficient vector The positions of non-zero elements are determined, the activated atoms and their corresponding sub-dictionaries are located, and the corresponding parameter parsing method is selected according to the decision logic.

[0136] If the activated atom belongs to the dynamic interharmonic atom sub-dictionary The Taylor-Fourier expansion method is then used as the implementation method for estimating the time-varying frequency parameters. The Taylor series is used to model the time-varying amplitude and phase of the component, and the instantaneous frequency trajectory is obtained through sliding window weighted least squares estimation. Rate of change of frequency and amplitude time-varying trajectory .

[0137] If the activated atom belongs to the oscillating harmonic atom sub-dictionary The matrix bundle method is then used as the implementation method for estimating the parameters of the exponentially decaying mode. A Hankel matrix is ​​constructed and singular value decomposition is performed to extract the damping factor. and modal frequencies .

[0138] If the activated atom belongs to the steady-state harmonic atom sub-dictionary Therefore, the rotation-invariant signal parameter estimation method is adopted as the implementation method for high-resolution spectrum estimation to extract steady-state amplitude and phase parameters.

[0139] If the activated atom belongs to the transient harmonic atom sub-dictionary Therefore, the synchronous compression transform method is used as the implementation method for time-frequency analysis to extract the transient duration. and energy distribution characteristics.

[0140] Step S5: Generate event fingerprint and complete detection.

[0141] The dynamic feature parameters extracted in step S4 are integrated into a structured broadband harmonic event fingerprint according to the above data model. The detection confidence level is calculated based on the reconstructed residuals using the following formula. :

[0142]

[0143] If the detection confidence level is greater than or equal to a preset threshold, the detection result is considered reliable, and an event fingerprint report is output. The report includes the event's physical category. Dominant frequency Frequency drift trajectory over time Damping characteristic parameter set Duration and detection confidence level Information such as... If the detection confidence level is lower than the preset threshold, the detection result is marked as low confidence, indicating that higher quality data or manual review is needed.

[0144] Based on the physical category and dynamic characteristic parameters of the event in the event fingerprint report, the operation and maintenance personnel determine whether there are broadband harmonic disturbance events such as interharmonic frequency drift caused by PLL dynamic interaction and weak damping oscillation caused by control instability in the current system, and make corresponding adjustments to control parameters or optimization of operation mode.

[0145] Example 2

[0146] This embodiment discloses a novel power system broadband harmonic event detection system based on structured sparse representation, used to implement the novel power system broadband harmonic event detection method described in Embodiment 1, which includes the following functional modules:

[0147] The data acquisition module is used to acquire broadband harmonic monitoring data that includes broadband harmonic disturbances;

[0148] The dictionary construction module is used to build a structured, overcomplete dictionary with clear physical mapping relationships, containing multiple physical atom sub-dictionaries, based on the broadband harmonic disturbance mechanism of new energy power electronic equipment.

[0149] The joint decomposition module is used to establish a dual sparse space joint decomposition model that includes a physical disturbance sparse space and a data anomaly sparse space, and decompose the broadband harmonic monitoring data to decompose the data anomaly component and the physical disturbance component belonging to the structured overcomplete dictionary.

[0150] The parameter parsing module is used to perform adaptive parameter parsing to obtain the corresponding dynamic feature parameters based on the physical atom type matched by the physical perturbation component and its corresponding physical atom sub-dictionary.

[0151] The fingerprint generation module is used to integrate the dynamic feature parameters to generate a structured broadband harmonic event fingerprint containing physical semantics;

[0152] The event recognition module is used to determine and output the broadband harmonic event detection result based on the structured broadband harmonic event fingerprint.

[0153] Example 3

[0154] This embodiment discloses a storage medium storing a program, which, when executed by a processor, implements the novel power system broadband harmonic event detection method described in Embodiment 1.

[0155] The storage medium in this embodiment can be a disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), USB flash drive, portable hard drive, etc.

[0156] Example 4

[0157] This embodiment discloses a computing device, including a processor and a memory for storing processor-executable programs. When the processor executes the program stored in the memory, it implements the novel power system broadband harmonic event detection method described in Embodiment 1.

[0158] The computing device described in this embodiment may be a desktop computer, laptop computer, smartphone, PDA handheld terminal, tablet computer, programmable logic controller (PLC), or other terminal device with processor function.

[0159] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A novel method for detecting broadband harmonic events in power systems, characterized in that, include: S1. Obtain broadband harmonic monitoring data containing broadband harmonic disturbances; S2. Based on the broadband harmonic disturbance mechanism of new energy power electronic equipment, establish a structured overcomplete dictionary with clear physical mapping relationships that includes multiple physical atom sub-dictionaries; S3. Establish a dual sparse space joint decomposition model that includes physical disturbance sparse space and data anomaly sparse space to decompose the broadband harmonic monitoring data and decompose the data anomaly component and the physical disturbance component belonging to the structured overcomplete dictionary. S4. Based on the physical atom type matched by the physical perturbation component and its corresponding physical atom sub-dictionary, perform adaptive parameter parsing to obtain the corresponding dynamic feature parameters; S5. Integrate the dynamic feature parameters to generate a structured broadband harmonic event fingerprint containing physical semantics. The structured broadband harmonic event fingerprint is a unified event description vector constructed based on the dynamic feature parameters obtained from the analysis. S6. Based on the structured broadband harmonic event fingerprint, determine and output the broadband harmonic event detection result.

2. The novel broadband harmonic event detection method for power systems according to claim 1, characterized in that, The specific steps for obtaining broadband harmonic monitoring data containing broadband harmonic disturbances are as follows: At new energy power plants or grid connection points, broadband voltage and current signals are collected at a set sampling frequency as observation signals to be processed. The observation signals contain one or more of the following broadband harmonic disturbances: a) random noise introduced by sensor thermal noise or electromagnetic interference; b) data loss caused by communication packet loss or storage failure; c) time-domain sparse abnormal spikes introduced by transient interference or equipment switching.

3. A novel broadband harmonic event detection method for power systems according to claim 2, characterized in that, The structured overcomplete dictionary consists of the following four sets of atoms generated based on a perturbation evolution model, each set of atoms corresponding to a different perturbation mechanism: ； In the formula, This is a structured, overcomplete dictionary, referred to as the physical heuristic harmonic event atom library; The set of steady-state harmonic atoms that describes a fixed-frequency harmonic is called the steady-state harmonic atom sub-dictionary; The set of dynamic interharmonic atoms that describes the perturbation of frequency over time is called the dynamic interharmonic atom sub-dictionary; The set of oscillating harmonic atoms that describes oscillations with damped or negatively damped characteristics is called the oscillating harmonic atom sub-dictionary; The set of transient harmonic atoms used to describe the mutation process is called the transient harmonic atom sub-dictionary.

4. A novel broadband harmonic event detection method for power systems according to claim 3, characterized in that, The physical descriptions and parameterization models of each sub-dictionary are as follows: Steady-state harmonic atomic dictionary : Used to describe the periodic perturbation components of power electronic devices, whose atoms are constructed using a cosine parameterized model: ； In the formula, For steady-state harmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; The initial amplitude, These are harmonic frequencies, corresponding to integer multiples of the fundamental frequency. This is the initial phase; Dynamic Interharmonic Atomic Dictionary This is used to map the time-varying frequency disturbance components caused by phase-locked loop dynamics, weak grid interactions, or maximum power point tracking disturbances. Its atoms are constructed using a time-varying spectral parameterization model. ； In the formula, For the dynamic interharmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; For the time-varying trajectory of amplitude, For the center frequency, The rate of change of frequency characterizes the speed and direction of frequency drift. Oscillating Harmonic Atomic Dictionary : Used to map the damped evolution component of control instability in grid-type converters, or subsynchronous or supersynchronous oscillations in new energy power plants. Its atoms are constructed using a decaying sinusoidal parameterized model. ； In the formula, For oscillatory harmonic atomic sub-dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; The attenuation coefficient is... Corresponding to damped oscillation, Corresponding to divergent oscillation, Corresponding to constant amplitude oscillation, The oscillation frequency; Transient Harmonic Atomic Dictionary Used for mapping short-term burst disturbance components caused by switching operations, fault clearing, and control mode switching; its atoms are constructed using a decay-modulated wavelet parameterization model. ； In the formula, Transient Harmonic Atomic Sub-Dictionary At time 1000, the middle atom The instantaneous oscillation amplitude; For decay rate, These are wavelet basis functions with compact support properties; Each sub-dictionary generates a set of redundant atoms by discretizing its own parameter space, forming an overcomplete representation basis covering typical broadband harmonic disturbance types; the discretization range and step size of each sub-dictionary parameter are determined according to the frequency domain resolution and time domain positioning accuracy requirements of the corresponding physical disturbance type.

5. A novel broadband harmonic event detection method for power systems according to claim 4, characterized in that, The dual sparse space joint decomposition model represents the observed signal as a superposition of physical disturbance components, data anomaly components, and random noise: ； In the formula, For observation signals; For a structured overcomplete dictionary; It is a sparse coefficient vector; For the structured overcomplete dictionary The physical disturbance component; These are time-domain sparse data anomalous components, including spikes and missing data segments; It is a random noise component; This represents the number of sampling points; Represents the set of real numbers; introduces outlier components by explicitly specifying them. This incorporates data quality issues into a unified mathematical framework, making data augmentation an integral part of the perturbation decomposition process.

6. A novel broadband harmonic event detection method for power systems according to claim 5, characterized in that, The joint decomposition model of the two sparse spaces solves for the following joint optimization objective function: ； In the formula, To constrain the sparsity of the physical perturbation components, the first norm term is used. To constrain the sparsity of outlier components in the data, the second norm term is used. For signal reconstruction fidelity; This is an abnormal sparsity regularization parameter used to control the sparsity of abnormal data components; This is a data fidelity regularization parameter used to balance the signal reconstruction accuracy with the sparsity of the solution; where, with norm Constrained sparse coefficient vector This reflects the prior knowledge that actual physical events have sparse representations in the physical heuristic harmonic event atom library, with norm Constraint data outliers This reflects the prior that the spikes and gaps in the observed signal have a sparse distribution in the time domain.

7. A novel broadband harmonic event detection method for power systems according to claim 6, characterized in that, The dual sparse space joint decomposition model is solved iteratively using either the alternating direction multiplier method or the block coordinate descent method, with the sparse coefficient vector being updated alternately in each iteration. Data anomaly components Using Lagrange multipliers, the solution is obtained by solving the problem until the convergence condition is met, thus obtaining the purified physical perturbation components. and the separated data anomaly components The abnormal component They are explicitly separated into independent sparse spaces and no longer mixed with physical perturbation components. This process allows for data purification while preserving complete broadband harmonic dynamic information.

8. A novel broadband harmonic event detection method for power systems according to claim 7, characterized in that, The adaptive parameter parsing is performed based on the sparse coefficient vector. The positions of non-zero elements are determined, the activated atoms and their corresponding sub-dictionaries are located, and an adaptive mapping rule for the "physical atom type - parameter resolution strategy" is established: When the activated atom belongs to the dynamic interharmonic atom sub-dictionary At that time, a time-varying frequency parameter estimation method is adopted, which uses Taylor series to model the time-varying amplitude and phase of the physical disturbance component, and extracts the instantaneous frequency trajectory through weighted least squares estimation. Rate of change of frequency and amplitude time-varying trajectory The time-varying frequency parameter estimation method is implemented using the Taylor-Fourier expansion method, the synchronous compression transform method, or the Hilbert transform method. When the activated atom belongs to the oscillating harmonic atom sub-dictionary At that time, the exponential decay mode parameter estimation method is used to construct the Hankel matrix for the physical disturbance components and perform singular value decomposition to extract the damping factor. and modal frequencies The exponential decay mode parameter estimation method is implemented using a matrix bundle method, the Prony method, or a rotation-invariant technique signal parameter estimation method. When the activated atom belongs to the steady-state harmonic atom sub-dictionary In this process, a high-resolution spectrum estimation method is used to extract the steady-state amplitude and phase parameters; wherein, the high-resolution spectrum estimation method is implemented using a rotation-invariant signal parameter estimation method or a multiple signal classification method; When the activated atom belongs to the transient harmonic atom sub-dictionary At that time, the transient duration was extracted using time-frequency analysis. and energy distribution characteristics.

9. A novel broadband harmonic event detection method for power systems according to claim 8, characterized in that, The structured broadband harmonic event fingerprint uses the following data model: ； In the formula, For structured broadband harmonic event fingerprinting; The event physical category includes at least steady-state harmonic events, dynamic interharmonic events, oscillatory harmonic events, and transient harmonic events; The dominant frequency; The frequency drift trajectory over time; The amplitude is a time-varying trajectory; It is a set of damping characteristic parameters, which at least includes the damping factor and the damping ratio; Duration of the event; To test the confidence level.

10. A novel broadband harmonic event detection method for power systems according to claim 9, characterized in that, The detection confidence level Based on reconstruction residual calculation: ； In the formula, To reconstruct the residuals Norm, representing the residual energy between the observed signal and the physical disturbance components and data anomaly components; For the observed signal Norm, representing the total energy of the observed signal; where, The range of values ​​is The closer the value is to 1, the more reliable the detection result; when the detection confidence level... When the result is below a preset threshold, the detection result is marked as low confidence, indicating that higher quality data or manual review is needed. Finally, based on the physical category and dynamic characteristic parameters of the event in the structured broadband harmonic event fingerprint, the broadband harmonic event detection result, i.e., the broadband harmonic disturbance event type, is determined and output.

Images