Broadband harmonic traceability diagnosis method and device, storage medium and computer equipment

Abstract

According to the broadband harmonic traceability diagnosis method and device, the storage medium and the computer equipment provided by the invention, the time-varying harmonic characteristics of the broadband data are extracted, the harmonic frequency band energy of the broadband data is analyzed, and the energy entropy of each frequency band in the broadband data is determined. Feature extraction is carried out through two channels of harmonic feature extraction and frequency band capability analysis, and precise separation of sub / inter-harmonics is realized. And inputting the time-varying harmonic features, the environmental parameters and the energy entropy of each frequency band into a fault classifier, and extracting the fault probability of the inverter from an obtained fault probability vector. In combination with environmental parameters and by using a lightweight fault classifier, on one hand, temperature or irradiance interference can be reduced, on the other hand, calculation complexity can be reduced, and detection efficiency and detection precision can be improved. And then, under the condition that the inverter fault probability is greater than a first preset threshold value, calculating a harmonic wave responsibility index of each inverter so as to identify a fault source in each inverter. And accurate positioning of the inverter fault is realized.

Description

Technical Field

[0001] This application relates to the field of harmonic mitigation technology, and in particular to a broadband harmonic source tracing and diagnosis method, device, storage medium and computer equipment. Background Technology

[0002] With the large-scale grid connection of new energy sources, broadband harmonic issues have become a core bottleneck affecting power grid security. In harmonic analysis, the traditional fast Fourier transform method is limited by a fixed time window, resulting in large measurement errors under dynamic conditions such as sudden changes in illumination; while the fixed threshold judgment rule is difficult to adapt to the nonlinear harmonic characteristics caused by temperature-irradiance coupling, resulting in a high false alarm rate.

[0003] For broadband harmonic source tracing and diagnosis, existing improved methods still have significant shortcomings. For example, time-frequency analysis methods or adaptive threshold strategies can capture some time-varying features, but their frequency domain resolution is insufficient, making it difficult to separate dense interharmonics; while the latter can reduce the false alarm rate, it increases response delay. In other words, existing harmonic analysis methods have poor detection efficiency while maintaining full-band detection accuracy, making it difficult to balance detection accuracy and efficiency in broadband harmonic source tracing and diagnosis. Summary of the Invention

[0004] The purpose of this application is to at least solve one of the above-mentioned technical defects, in particular the technical defect that the existing harmonic analysis method has poor detection efficiency while maintaining full-band detection accuracy, which makes it difficult to balance detection accuracy and detection efficiency in broadband harmonic source tracing and diagnosis.

[0005] In a first aspect, this application provides a broadband harmonic source tracing and diagnosis method, the method comprising:

[0006] Acquire broadband data from new energy power plants and extract time-varying harmonic characteristics from the broadband data;

[0007] The harmonic frequency band energy of the broadband data is analyzed to determine the energy entropy of each frequency band in the broadband data;

[0008] The environmental parameters are obtained, and the time-varying harmonic features, the environmental parameters, and the energy entropy of each frequency band are input into a pre-trained fault classifier to obtain a fault probability vector. The inverter fault probability is then extracted from the fault probability vector.

[0009] When the inverter failure probability is greater than the first preset threshold, the harmonic responsibility index of each inverter is calculated based on the harmonic admittance matrix of each inverter in the new energy power plant.

[0010] Inverters with harmonic responsibility indicators greater than a second preset threshold are identified among all inverters, and the identified inverters are determined as the source of the fault.

[0011] In one embodiment, acquiring broadband data from the new energy power plant and extracting the time-varying harmonic features of the broadband data includes:

[0012] Collect broadband data from new energy power plants;

[0013] A preset feature extraction network is determined, and the broadband data is input into the feature extraction network. The dynamic characteristics of harmonics are captured by the dual-layer LSTM layer in the feature extraction network to obtain the time-varying harmonic characteristics of the broadband data.

[0014] In one embodiment, the method further includes, before inputting the broadband data into the feature extraction network:

[0015] Determine the Hanning window function;

[0016] The broadband data is windowed using the Hanning window function, and the resulting windowed broadband data is used as broadband data for subsequent applications.

[0017] In one embodiment, the step of analyzing the harmonic band energy of the broadband data to determine the energy entropy of each band in the broadband data includes:

[0018] Determine the preset wavelet basis functions and target frequency band;

[0019] The broadband data is decomposed using the wavelet basis function, and the frequency bands within the target frequency band from the decomposition results are divided into N sub-frequency bands;

[0020] Calculate the energy entropy of each frequency band in the N sub-bands to obtain the energy entropy of each frequency band in the broadband data.

[0021] In one embodiment, the step of inputting the time-varying harmonic features, the environmental parameters, and the energy entropy of each frequency band into a pre-trained fault classifier to obtain a fault probability vector includes:

[0022] A pre-trained fault classifier is determined, the fault classifier comprising a first convolutional layer, a second convolutional layer, and a fully connected layer;

[0023] The time-varying harmonic features, the environmental parameters, and the energy entropy of each frequency band are input into the first convolutional layer to extract local harmonic correlation features.

[0024] The local harmonic correlation features are input into the second convolutional layer for depth feature extraction to obtain the target depth features, and the target depth features are input into the fully connected layer to obtain the fault probability vector.

[0025] In one embodiment, the harmonic liability index of each inverter in the new energy power plant is calculated based on the harmonic admittance matrix of each inverter according to the following expression:

[0026]

[0027] In the formula, Indicates the first Harmonic admittance matrix at the grid connection point of the inverter. This represents the harmonic current deviation vector. This indicates the total number of inverters in a new energy power plant.

[0028] In one embodiment, the method further includes:

[0029] At preset time intervals, the inverters in the new energy power plant are tested using the harmonic injection method, and the parameters of the fault classifier are updated based on the test results.

[0030] Secondly, this application provides a broadband harmonic source tracing and diagnostic device, the device comprising:

[0031] The feature extraction module is used to acquire broadband data from new energy power plants and extract the time-varying harmonic features of the broadband data.

[0032] The frequency band analysis module is used to analyze the harmonic frequency band energy of the broadband data and determine the energy entropy of each frequency band in the broadband data.

[0033] The fault classification module is used to acquire environmental parameters, input the time-varying harmonic features, the environmental parameters and the energy entropy of each frequency band into a pre-trained fault classifier to obtain a fault probability vector, and extract the inverter fault probability from the fault probability vector.

[0034] The indicator calculation module is used to calculate the harmonic responsibility index of each inverter based on the harmonic admittance matrix of each inverter in the new energy power plant when the inverter failure probability is greater than the first preset threshold.

[0035] The fault source determination module is used to identify inverters whose harmonic responsibility index is greater than a second preset threshold among all inverters, and to determine the identified inverters as fault sources.

[0036] Thirdly, this application provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the broadband harmonic source tracing and diagnosis method as described in any of the above embodiments.

[0037] Fourthly, this application provides a computer device, including: one or more processors, and a memory;

[0038] The memory stores computer-readable instructions, and when the one or more processors execute the computer-readable instructions, they perform the steps of the broadband harmonic source tracing and diagnosis method as described in any of the above embodiments.

[0039] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0040] This application provides a broadband harmonic source tracing and diagnostic method, apparatus, storage medium, and computer equipment. It acquires broadband data from a new energy power plant and extracts time-varying harmonic features from the broadband data. Then, it analyzes the harmonic frequency band energy of the broadband data to determine the energy entropy of each frequency band. Feature extraction is performed through a dual-channel approach of harmonic feature extraction and frequency band capability analysis, fusing time-domain dynamic features and frequency band energy entropy distribution to achieve accurate separation of sub- and inter-harmonics, improving the accuracy of harmonic detection. Next, the time-varying harmonic features, environmental parameters, and the energy entropy of each frequency band are input into a pre-trained fault classifier to obtain a fault probability vector, and the inverter fault probability is extracted from the fault probability vector. Combining environmental parameters and using a lightweight fault classifier can mitigate temperature or irradiance interference and reduce end-to-end computational complexity, improving detection efficiency and accuracy. Subsequently, when the inverter fault probability exceeds a first preset threshold, the harmonic responsibility index of each inverter is calculated based on the harmonic admittance matrix of each inverter in the new energy power plant to identify the fault source in each inverter. By using the harmonic admittance matrix, each inverter is equivalent to an independent harmonic current source, thereby quantifying the harmonic responsibility ratio of each inverter and achieving accurate location of inverter faults. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A flowchart illustrating a broadband harmonic source tracing and diagnostic method provided in this application embodiment;

[0043] Figure 2 A schematic diagram of a broadband harmonic source tracing and diagnostic device provided in this application embodiment;

[0044] Figure 3This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0045] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0046] In one embodiment, this application provides a broadband harmonic source tracing and diagnostic method, which is illustrated in the following embodiments using a server as an example. It is understood that the broadband harmonic source tracing and diagnostic method can be performed on a single server or a server cluster consisting of multiple servers; this application does not impose any specific limitations on this.

[0047] like Figure 1 As shown, this application provides a broadband harmonic source tracing and diagnosis method, the method comprising:

[0048] S101: Acquire broadband data from new energy power plants and extract time-varying harmonic characteristics from the broadband data.

[0049] Wideband data refers to signal data collected over a wide frequency range, including variations in electrical parameters such as voltage and current at different frequencies. Time-varying harmonic characteristics refer to the properties of harmonic amplitude, phase, and other parameters that change over time.

[0050] In this step, the server can collect voltage and current signals through measurement devices pre-deployed at key nodes in the renewable energy power plant to form broadband data corresponding to the renewable energy power plant. This broadband data includes, but is not limited to, inverter DC-side voltage / current, inverter AC output voltage / current, and grid connection point voltage / current. Once the broadband data of the renewable energy power plant at this time is determined, harmonic features can be extracted from the broadband data through a pre-trained feature extraction network to obtain the time-varying harmonic features of the broadband data.

[0051] Specifically, time-varying harmonic features include the gradient of total harmonic distortion (THD), the standard deviation of phase jumps, and the fundamental amplitude volatility. The THD gradient measures the harmonic content in the data signal and is expressed as the ratio of the effective values ​​of all harmonic components to the effective value of the fundamental frequency. The standard deviation of phase jumps measures the dispersion of phase jumps, which can be used to assess the phase stability of renewable energy power plants under dynamic conditions. The fundamental amplitude volatility measures the magnitude of change in the fundamental amplitude per unit time and can be used to assess the output power stability of renewable energy power plants. Therefore, by extracting time-varying harmonic features from broadband data, the operating status of renewable energy power plants can be assessed from multiple perspectives, providing accurate data support for subsequent broadband harmonic source tracing and diagnosis.

[0052] S102: Analyze the harmonic band energy of broadband data to determine the energy entropy of each band in the broadband data.

[0053] Energy entropy is used to measure the uniformity of energy distribution of a signal within a specific frequency band.

[0054] In this step, a wavelet transform layer is used to divide the broadband data signal into several frequency bands, and the energy entropy of each frequency band is calculated. Specifically, the energy entropy can be calculated by determining the relative distribution of each frequency component within each frequency band. For example, the Shannon entropy formula can be used to quantify the uniformity of the energy distribution.

[0055] Furthermore, in the process of frequency band division and energy entropy calculation, the acquired broadband data can first be processed by wavelet transform. Wavelet transform decomposes the signal into wavelet coefficients of different scales, thus enabling analysis of the signal across different frequency bands. Specifically, wavelet transform can decompose broadband data into multiple levels of detail coefficients and approximation coefficients, each level corresponding to a specific frequency band range. For example, the first level of wavelet transform can decompose the signal into high-frequency detail coefficients (high-frequency band) and low-frequency approximation coefficients (low-frequency band), while each subsequent level of transform further subdivides the low-frequency band signal, resulting in a finer-grained frequency band division.

[0056] It is understandable that wavelet transform can provide multi-resolution analysis of signals in both time and frequency, thereby capturing the local features and dynamic changes of signals more accurately. Analyzing the harmonic band energy of broadband data using wavelet transform allows for a more detailed analysis of the signal's characteristics in different frequency bands. Furthermore, by quantifying the uniformity of energy distribution within each band, it helps identify major harmonic sources and potential power quality problems.

[0057] In addition, time-varying harmonic feature extraction and frequency band energy analysis are performed on broadband data separately. Broadband data is analyzed and processed simultaneously from two channels, thereby improving the detection rate of harmonics in broadband data and providing an accurate data foundation for subsequent source tracing and diagnosis.

[0058] S103: Obtain environmental parameters, input the time-varying harmonic features, environmental parameters and energy entropy of each frequency band into the pre-trained fault classifier to obtain the fault probability vector, and extract the inverter fault probability from the fault probability vector.

[0059] Environmental parameters refer to parameters related to the environment of new energy power plants, including but not limited to irradiance and temperature. The fault classifier is a lightweight convolutional classification model that can be used to assess the probability of various faults occurring.

[0060] In this step, environmental parameters of the new energy power plant are collected by monitoring devices such as sensors installed at the plant. These environmental parameters, time-varying harmonic characteristics, and energy entropy of various frequency bands are then integrated to meet the input specifications of the fault classifier. The integrated data is input into a pre-trained fault classifier to obtain a fault probability vector output by the classifier. This vector includes the probability of grid disturbance faults, inverter faults, component aging faults, and shadowing effect faults. Since the objective of this application is to detect inverter faults, the probability of inverter faults can be extracted from the fault probability vector as the inverter fault probability. It is understood that when detecting other fault conditions, the probability of the corresponding fault type can also be extracted for subsequent judgment and analysis. This application does not impose specific limitations in this regard.

[0061] Specifically, the fault classifier comprises a first convolutional layer, a second convolutional layer, and a fully connected layer. This fault classifier uses a combination of time-varying harmonic features, environmental parameters, and energy entropy of each frequency band as training samples, and the fault type corresponding to this combination as the sample label. It is then trained using a pre-set pre-trained model with a loss function. During training, the server can input the training samples and their corresponding labels into the pre-trained model for forward propagation, and use the loss function to adjust the parameters of the pre-trained model during backpropagation until the pre-trained model meets certain iteration conditions. The trained pre-trained model can then be used as the fault classifier.

[0062] Furthermore, when determining the fault probability vector, the fault type corresponding to the maximum occurrence probability can be selected as the fault type for this broadband harmonic diagnosis. Alternatively, an appropriate threshold can be set for each fault type, and fault types exceeding the corresponding threshold can be used as the fault types for this broadband harmonic diagnosis. This application does not impose specific limitations in this regard.

[0063] S104: When the inverter failure probability is greater than the first preset threshold, calculate the harmonic responsibility index of each inverter based on the harmonic admittance matrix of each inverter in the new energy power plant.

[0064] The first preset threshold is an empirical value that can be set and adjusted according to actual needs. In one example, it can be set to 0.8. The harmonic admittance matrix is ​​a matrix representation of the frequency domain admittance characteristics exhibited by the inverter under harmonic excitation at different frequencies. The harmonic responsibility index is used to quantify the degree of harmonic contribution of each inverter to the new energy power plant.

[0065] In this step, when the extracted inverter failure probability exceeds a first preset threshold, impedance spectrum sweep testing can be performed on each inverter, and the harmonic admittance matrix of each inverter can be determined based on the connection relationship of each inverter. Next, based on the harmonic admittance matrix of each inverter, the harmonic contribution of each inverter to the harmonics in the new energy power plant is calculated, i.e., the harmonic responsibility index.

[0066] Furthermore, when calculating the harmonic liability index, the harmonic current deviation vector is obtained, and then the harmonic admittance matrix and harmonic current deviation vector of each inverter are used to calculate the harmonic liability index of each inverter.

[0067] Specifically, when a high probability of inverter failure is detected, calculating the harmonic responsibility index can quickly identify which inverters are the main harmonic sources. For example, if an inverter has a high harmonic responsibility index, it indicates that the inverter contributes significantly to the harmonic level of the system. In this way, maintenance personnel can conduct targeted inspections and maintenance on the inverter, reducing the impact of harmonics on the system and the power grid.

[0068] S105: Identify inverters among all inverters whose harmonic responsibility index is greater than the second preset threshold, and determine the identified inverters as fault sources.

[0069] The first preset threshold is an empirical value that can be set and adjusted according to actual needs. In one example, it can be set to 0.25. A fault source refers to an inverter that generates harmonic pollution or other power quality problems due to its own malfunction or abnormal operating state.

[0070] In this step, when the harmonic responsibility index of each inverter is calculated, a threshold judgment is used to determine the inverter's operating status. When the inverter's harmonic responsibility index is greater than a second preset threshold, it indicates that the inverter may be in an abnormal operating state and is a major source of harmonic pollution in the new energy power plant. This achieves the goal of effectively identifying and locating inverters that have a significant adverse impact on the system's power quality. Furthermore, existing impedance analysis methods, when locating faults, suffer from mixed harmonic components from both the grid and inverter sides because grid background harmonics are reverse-coupled to the internal network of the new energy power plant through a common coupling point. Traditional impedance models can only reflect the total system impedance characteristics and cannot separate the independent contributions of grid background harmonics and specific inverter faults, thus making it difficult to distinguish fault sources. This application, however, establishes nodal harmonic admittance equations, treating each inverter as an independent harmonic current source and directly quantifying the harmonic responsibility proportion of each inverter. This decouples background harmonics and accurately locates internal inverter faults, improving the accuracy of fault location.

[0071] In the above embodiments, broadband data from the new energy power plant is acquired, and time-varying harmonic features of the broadband data are extracted. Then, the harmonic band energy of the broadband data is analyzed to determine the energy entropy of each band. Feature extraction is performed through a dual-channel approach of harmonic feature extraction and band capability analysis, fusing time-domain dynamic features and band energy entropy distribution to achieve accurate separation of sub- and inter-harmonics, improving the accuracy of harmonic detection. Next, the time-varying harmonic features, environmental parameters, and energy entropy of each band are input into a pre-trained fault classifier to obtain a fault probability vector, and the inverter fault probability is extracted from the fault probability vector. Combining environmental parameters and using a lightweight fault classifier can mitigate temperature or irradiance interference and reduce end-to-end computational complexity, improving detection efficiency and accuracy. Then, when the inverter fault probability exceeds a first preset threshold, the harmonic responsibility index of each inverter is calculated based on the harmonic admittance matrix of each inverter in the new energy power plant to identify the fault source in each inverter. By using the harmonic admittance matrix, each inverter is equivalent to an independent harmonic current source, thereby quantifying the harmonic responsibility ratio of each inverter and achieving accurate location of inverter faults.

[0072] In one embodiment, broadband data from a new energy power plant is acquired, and time-varying harmonic characteristics of the broadband data are extracted, including:

[0073] S1: Collect broadband data from new energy power plants.

[0074] S2: Determine the preset feature extraction network and input broadband data into the feature extraction network. Capture the dynamic characteristics of harmonics through the dual-layer LSTM layer in the feature extraction network to obtain the time-varying harmonic characteristics of the broadband data.

[0075] Among them, the feature extraction network is used to extract time-varying harmonic features. The feature extraction network includes a two-layer LSTM layer. The LSTM (Long Short-Term Memory) layer is a specially designed recurrent neural network that can effectively capture long-term dependencies in time series data.

[0076] In this embodiment, broadband data from the new energy power plant can be collected at a set sampling rate. Then, a feature extraction network consisting of two LSTM layers is determined, and this feature extraction network is used to capture the harmonic dynamic features in the broadband data, thereby obtaining the time-varying harmonic features corresponding to the broadband data.

[0077] Furthermore, the dual-layer LSTM structure, through the stacking of two LSTM layers, can further enhance the feature extraction network's ability to model complex dynamic characteristics. The first LSTM layer can initially extract time-series features from broadband data, capturing short-term dynamic changes in harmonics. The second LSTM layer then further processes this data, extracting deeper long-term dynamic characteristics, thus obtaining the time-varying harmonic features of the broadband data. These time-varying harmonic features can reflect the changes in harmonics over time in new energy power plants, providing important evidence for subsequent fault diagnosis and power quality assessment.

[0078] Specifically, the feature extraction network in this application can use time-series collected broadband data samples as training samples, time-varying harmonic features corresponding to the training samples as sample labels, and a pre-trained model trained using a loss function. During training, the server can input the training samples and their corresponding sample labels into the pre-trained model for forward propagation, and use the loss function to adjust the parameters of the pre-trained model during backpropagation until the pre-trained model meets certain iteration conditions. The trained pre-trained model can then be used as the feature extraction network.

[0079] In one embodiment, the method further includes, before inputting broadband data into the feature extraction network:

[0080] S1: Determine the Hanning window function.

[0081] S2: Use the Hanning window function to window the broadband data, and use the windowed broadband data as broadband data for subsequent use.

[0082] In this embodiment, the Hanning window function is used to window the broadband data. Specifically, each sample point of the broadband data is multiplied by the corresponding Hanning window function value. This process can be achieved through point-by-point multiplication. The windowed broadband data can effectively suppress spectral leakage and improve the accuracy of spectral analysis during frequency domain analysis.

[0083] In one example, the Hanning window function can be represented as follows:

[0084]

[0085] In the formula, This represents the Hanning window function. This indicates the number of sampling points within one second.

[0086] In one embodiment, the harmonic band energy of broadband data is analyzed to determine the energy entropy of each band in the broadband data, including:

[0087] S1: Determine the preset wavelet basis function and target frequency band.

[0088] S2: Use wavelet basis functions to decompose broadband data, and divide the frequency bands within the target frequency band from the decomposition results into N sub-bands.

[0089] S3: Calculate the energy entropy of each frequency band in the N sub-bands to obtain the energy entropy of each frequency band in the broadband data.

[0090] Where N is a positive integer greater than zero.

[0091] In this embodiment, a preset wavelet basis function and target frequency band are determined. The wavelet basis function is the core of wavelet transform, used to decompose a signal into components at different frequencies and time scales. Common wavelet basis functions include Daubechies, Symlets, and Coiflets. Then, based on the determined wavelet basis function, the broadband data is decomposed to extract the frequency band within the target frequency band, and this frequency band is further divided into N sub-bands. Finally, the energy entropy of each of the N sub-bands is calculated. Specifically, the sum of the energy of all wavelet coefficients in each band can be calculated, and then the energy entropy of each of the N sub-bands is determined by combining the definition of energy entropy.

[0092] Specifically, wavelet transform can provide multi-resolution analysis of signals in both time and frequency, thereby capturing the local features and dynamic changes of signals more accurately. By selecting appropriate wavelet basis functions and target frequency bands, broadband data can be decomposed into multiple sub-bands, each corresponding to a different frequency range. This allows for more detailed analysis of the characteristics of signals in different frequency bands, identifying major harmonic sources and potential power quality problems.

[0093] In one example, based on the application scenario of this application, the wavelet basis function can be the db8 wavelet basis function, and the target frequency band can be set to 0-5kHz.

[0094] In one example, assuming N is 16 and the target frequency band is 0-5kHz, then the bandwidth is 312.5Hz. The process of calculating the energy entropy of each of the N sub-bands can be expressed as follows:

[0095]

[0096]

[0097] In the formula, Let M represent the k-th wavelet coefficient in the j-th frequency band, where M is the total number of coefficients in that frequency band. This represents the wavelet packet energy of the j-th frequency band. This represents the energy entropy of each frequency band.

[0098] In one embodiment, time-varying harmonic features, environmental parameters, and energy entropy of each frequency band are input into a pre-trained fault classifier to obtain a fault probability vector, including:

[0099] S1: Determine a pre-trained fault classifier, which includes a first convolutional layer, a second convolutional layer, and a fully connected layer.

[0100] S2: Input the time-varying harmonic features, environmental parameters, and energy entropy of each frequency band into the first convolutional layer to extract local harmonic correlation features.

[0101] S3: Input the local harmonic correlation features into the second convolutional layer for depth feature extraction to obtain the target depth features, and input the target depth features into the fully connected layer to obtain the fault probability vector.

[0102] Among them, the local harmonic correlation features contain the local mode and structural information of the harmonic signal in time or frequency, which can reflect the mutual influence and correlation between harmonic components.

[0103] In this embodiment, time-varying harmonic features, environmental parameters, and energy entropy of each frequency band are first input into a fault classifier, which includes a first convolutional layer, a second convolutional layer, and a fully connected layer. Therefore, this data is first input into the first convolutional layer, which performs convolution operations on the input data using learned convolutional kernels to extract local harmonic correlation features. These features can capture local pattern and structural information in the input data, such as the frequency distribution and amplitude variations of harmonic components. Next, the local harmonic correlation features are input into the second convolutional layer for deep feature extraction. The second convolutional layer further abstracts and combines the local features to extract higher-level target deep features. These deep features contain more complex pattern and structural information in the input data, and can more comprehensively reflect the system's operating state and potential faults. Finally, the target deep features are input into the fully connected layer to obtain a fault probability vector. The fully connected layer can integrate and classify the target deep features, outputting the probability value for each fault category, thereby achieving fault identification and classification.

[0104] Understandably, prior harmonic analysis ensures the accuracy and comprehensiveness of time-varying harmonic characteristics and energy entropy across various frequency bands. Then, a lightweight fault classifier can quickly determine the fault probability vector. This allows for a balance between efficiency and accuracy in broadband harmonic diagnosis.

[0105] In one example, the first convolutional layer consists of 32 convolutional kernels, each with a receptive field size of 5×1, and the second convolutional layer consists of 64 convolutional kernels, each with a receptive field size of 3×1.

[0106] In another example, the failure probability vector can be represented as:

[0107]

[0108] In the formula, Represents the failure probability vector. This indicates the probability of a power grid disturbance fault occurring. This indicates the probability of an inverter failure occurring. This indicates the probability of component aging occurring. This indicates the probability of the shadow effect failure occurring.

[0109] In one embodiment, the harmonic liability index of each inverter in the new energy power plant is calculated based on the harmonic admittance matrix of each inverter according to the following expression:

[0110]

[0111] In the formula, Indicates the first Harmonic admittance matrix at the grid connection point of the inverter. This represents the harmonic current deviation vector. This represents the total number of inverters in a new energy power plant. The harmonic current deviation vector refers to the difference between the actual measured harmonic current and the expected (or standard) harmonic current.

[0112] In this embodiment, the harmonic admittance matrix reflects the inverter's response characteristics to harmonic currents. By multiplying it by the harmonic current deviation vector, the harmonic current generated by the corresponding inverter can be obtained, and then the harmonic responsibility index can be determined according to the proportion of harmonic current.

[0113] In one embodiment, the broadband harmonic source tracing and diagnosis method further includes:

[0114] At preset time intervals, the inverters in the new energy power plant are tested using the harmonic injection method, and the parameters of the fault classifier are updated based on the test results.

[0115] In this embodiment, a harmonic injection test is initiated at preset time intervals. The harmonic injection method involves injecting harmonic signals of known frequency and amplitude into the inverter's output terminal and recording the inverter's response to the harmonics. This response data includes the amplitude and phase changes of the harmonic current, and the inverter's protection actions. Based on the test results composed of this response data, the inverter's performance under different harmonic conditions is analyzed. If abnormal responses are found in the inverter under certain harmonic conditions, such as excessive harmonic current or protection malfunctions, these data can be recorded as training samples. These training samples can then be used to further optimize the fault classifier, improving its fault identification capability and accuracy, allowing the fault classifier to better adapt to the current actual operating state of the inverter.

[0116] In one embodiment, the broadband harmonic source tracing and diagnosis method further includes: comparing the diagnosis result consisting of the fault probability vector and the fault source with a preset historical fault database; if there is no record matching the diagnosis result in the historical fault database, then adding the diagnosis result to the historical fault database.

[0117] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0118] The broadband harmonic source tracing and diagnostic device provided in the embodiments of this application is described below. The broadband harmonic source tracing and diagnostic device described below can be referred to in correspondence with the broadband harmonic source tracing and diagnostic method described above.

[0119] like Figure 2 As shown, this application provides a broadband harmonic source tracing and diagnostic device 200, the device comprising:

[0120] The feature extraction module 201 is used to acquire broadband data of new energy power plants and extract time-varying harmonic features of broadband data;

[0121] The frequency band analysis module 202 is used to analyze the harmonic frequency band energy of broadband data and determine the energy entropy of each frequency band in the broadband data.

[0122] The fault classification module 203 is used to obtain environmental parameters, input the time-varying harmonic features, environmental parameters and energy entropy of each frequency band into the pre-trained fault classifier to obtain the fault probability vector, and extract the inverter fault probability from the fault probability vector.

[0123] The indicator calculation module 204 is used to calculate the harmonic responsibility index of each inverter based on the harmonic admittance matrix of each inverter in the new energy power plant when the inverter failure probability is greater than the first preset threshold.

[0124] The fault source determination module 205 is used to identify inverters in each inverter whose harmonic responsibility index is greater than a second preset threshold, and to determine the identified inverters as fault sources.

[0125] In the above embodiments, broadband data from the new energy power plant is acquired, and time-varying harmonic features of the broadband data are extracted. Then, the harmonic band energy of the broadband data is analyzed to determine the energy entropy of each band. Feature extraction is performed through a dual-channel approach of harmonic feature extraction and band capability analysis, fusing time-domain dynamic features and band energy entropy distribution to achieve accurate separation of sub- and inter-harmonics, improving the accuracy of harmonic detection. Next, the time-varying harmonic features, environmental parameters, and energy entropy of each band are input into a pre-trained fault classifier to obtain a fault probability vector, and the inverter fault probability is extracted from the fault probability vector. Combining environmental parameters and using a lightweight fault classifier can mitigate temperature or irradiance interference and reduce end-to-end computational complexity, improving detection efficiency and accuracy. Then, when the inverter fault probability exceeds a first preset threshold, the harmonic responsibility index of each inverter is calculated based on the harmonic admittance matrix of each inverter in the new energy power plant to identify the fault source in each inverter. By using the harmonic admittance matrix, each inverter is equivalent to an independent harmonic current source, thereby quantifying the harmonic responsibility ratio of each inverter and achieving accurate location of inverter faults.

[0126] In one embodiment, the feature extraction module includes:

[0127] The data acquisition submodule is used to collect broadband data from new energy power plants;

[0128] The feature extraction submodule is used to determine the preset feature extraction network and input broadband data into the feature extraction network. The two-layer LSTM layer in the feature extraction network captures the dynamic characteristics of harmonics to obtain the time-varying harmonic features of the broadband data.

[0129] In one embodiment, the feature extraction module further includes the following before executing the feature extraction submodule:

[0130] The function determination submodule is used to determine the Hanning window function;

[0131] The data processing submodule is used to perform windowing processing on broadband data using the Hanning window function, and then uses the windowed broadband data as broadband data for subsequent use.

[0132] In one embodiment, the frequency band analysis module includes:

[0133] The data determination submodule is used to determine the preset wavelet basis function and target frequency band;

[0134] The data decomposition submodule is used to decompose broadband data using wavelet basis functions and divide the frequency bands within the target frequency band from the decomposition results into N sub-frequency bands;

[0135] The calculation submodule is used to calculate the energy entropy of each frequency band in N sub-bands, so as to obtain the energy entropy of each frequency band in the broadband data.

[0136] In one embodiment, the fault classification module includes:

[0137] The classifier determination submodule is used to determine a pre-trained fault classifier, which includes a first convolutional layer, a second convolutional layer, and a fully connected layer.

[0138] The feature extraction submodule is used to input time-varying harmonic features, environmental parameters, and energy entropy of each frequency band into the first convolutional layer to extract local harmonic correlation features.

[0139] The deep extraction submodule is used to input the local harmonic correlation features into the second convolutional layer for deep feature extraction to obtain the target deep features, and then input the target deep features into the fully connected layer to obtain the fault probability vector.

[0140] In one embodiment, the indicator calculation module includes:

[0141]

[0142] In the formula, Indicates the first Harmonic admittance matrix at the grid connection point of the inverter. This represents the harmonic current deviation vector. This indicates the total number of inverters in a new energy power plant.

[0143] In one embodiment, the broadband harmonic source tracing and diagnostic device further includes:

[0144] The feedback optimization module is used to test the inverters in the new energy power plant using the harmonic injection method at preset time intervals, and update the parameters of the fault classifier based on the test results.

[0145] The division of modules in the above-described broadband harmonic source tracing and diagnostic device is merely illustrative. In other embodiments, the broadband harmonic source tracing and diagnostic device can be divided into different modules as needed to complete all or part of its functions. Each module in the above-described broadband harmonic source tracing and diagnostic device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0146] In one embodiment, this application also provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the broadband harmonic source tracing and diagnostic method as described in any of the above embodiments.

[0147] In one embodiment, this application also provides a computer device storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the broadband harmonic source tracing and diagnostic method as described in any of the above embodiments.

[0148] Indicatively, such as Figure 3 As shown, Figure 3 This is a schematic diagram of the internal structure of a computer device 300 provided in an embodiment of this application. The computer device 300 can be provided as a server. (Refer to...) Figure 3 The computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by memory 301 for storing instructions, such as application programs, that can be executed by the processing component 302. The application programs stored in memory 301 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 302 is configured to execute instructions to perform the broadband harmonic source tracing diagnostic method of any of the above embodiments.

[0149] The computer device 300 may also include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate on an operating system stored in memory 301, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.

[0150] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0151] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. 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 limitations, an element defined by the phrase "comprising a…" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. 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 the terms “comprising / including” or “having” specify the presence of the stated features, wholes, steps, operations, components, parts or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, 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.

[0152] 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.

[0153] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and 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 broadband harmonic traceability diagnostic method, characterized in that, The method comprises: acquiring wideband data of a new energy power plant and extracting time-varying harmonic features of the wideband data; analyzing harmonic band energy of the wideband data to determine energy entropy of each frequency band in the wideband data; acquiring environmental parameters, inputting the time-varying harmonic features, the environmental parameters and the energy entropy of each frequency band into a pre-trained fault classifier to obtain a fault probability vector, and extracting an inverter fault probability from the fault probability vector; when the inverter fault probability is greater than a first preset threshold, calculating a harmonic responsibility index of each inverter based on a harmonic admittance matrix of each inverter in the new energy power plant; identifying an inverter with a harmonic responsibility index greater than a second preset threshold in each inverter, and determining the identified inverter as a fault source.

2. The broadband harmonic traceability diagnostic method of claim 1, wherein, The acquisition of the wideband data of the new energy power plant and the extraction of the time-varying harmonic features of the wideband data comprise: collecting wideband data of a new energy power plant; determining a preset feature extraction network, and inputting the wideband data into the feature extraction network to capture harmonic dynamic characteristics through a double-layer LSTM layer in the feature extraction network to obtain time-varying harmonic features of the wideband data.

3. The broadband harmonic traceability diagnostic method of claim 2, wherein, Before inputting the wideband data into the feature extraction network, the method further comprises: determining a Hanning window function; performing windowing processing on the wideband data by using the Hanning window function, and taking the obtained wideband data subjected to the windowing processing as the wideband data used subsequently.

4. The broadband harmonic traceability diagnostic method of claim 1, wherein, The analysis of the harmonic band energy of the wideband data to determine the energy entropy of each frequency band in the wideband data comprises: determining a preset wavelet basis function and a target frequency band; decomposing the wideband data by using the wavelet basis function, and dividing a frequency band in the decomposition result within the target frequency band into N sub-frequency bands; calculating the energy entropy of each frequency band in the N sub-frequency bands to obtain the energy entropy of each frequency band in the wideband data.

5. The broadband harmonic traceability diagnostic method of claim 1, wherein, The inputting of the time-varying harmonic features, the environmental parameters and the energy entropy of each frequency band into the pre-trained fault classifier to obtain a fault probability vector comprises: determining a pre-trained fault classifier, wherein the fault classifier comprises a first convolutional layer, a second convolutional layer and a fully connected layer; inputting the time-varying harmonic features, the environmental parameters and the energy entropy of each frequency band into the first convolutional layer to extract local harmonic correlation features; inputting the local harmonic correlation features into the second convolutional layer for deep feature extraction to obtain target deep features, and inputting the target deep features into the fully connected layer to obtain a fault probability vector.

6. The broadband harmonic traceability diagnostic method of claim 1, wherein, The harmonic responsibility index of each inverter is calculated based on a harmonic admittance matrix of each inverter in the new energy power plant according to the following expression: In the formula, denotes the number of inverters in the new energy power generation field. denotes the harmonic admittance matrix of the grid-connected point of the inverter, denotes the harmonic current deviation vector, denotes the total number of inverters in the new energy power generation field.

7. The broadband harmonic traceability diagnostic method according to any of claims 1 to 6, characterized in that, The method further comprises: when an interval is preset for a time period, testing the inverters in the new energy power plant by a harmonic injection method, and updating parameters of the fault classifier according to a test result.

8. A broadband harmonic traceability diagnostic device, characterized by, The device comprises: a feature extraction module configured to acquire wideband data of a new energy power plant and extract time-varying harmonic features of the wideband data; a frequency band analysis module configured to analyze harmonic frequency band energy of the wideband data to determine energy entropy of each frequency band in the wideband data; a fault classification module configured to obtain environmental parameters, input the time-varying harmonic feature, the environmental parameters, and the energy entropy of each frequency band into a pre-trained fault classifier to obtain a fault probability vector, and extract an inverter fault probability from the fault probability vector; an index calculation module configured to, when the inverter fault probability is greater than a first preset threshold, calculate a harmonic responsibility index of each inverter based on a harmonic admittance matrix of each inverter in the new energy power plant; a fault source determination module configured to identify, among the inverters, an inverter with a harmonic responsibility index greater than a second preset threshold, and determine the identified inverter as a fault source.

9. A storage medium characterized by: The storage medium has computer readable instructions stored therein, and the computer readable instructions, when executed by one or more processors, cause the one or more processors to perform the steps of the wideband harmonic tracing diagnosis method according to any one of claims 1 to 7.

10. A computer device, comprising: comprise: one or more processors, and a memory; the memory has computer readable instructions stored therein, and the computer readable instructions, when executed by the one or more processors, perform the steps of the wideband harmonic tracing diagnosis method according to any one of claims 1 to 7.

Images