High-frequency resonance fault judgment method and device

By using convolutional neural network in a flexible DC system to extract the high-frequency resonance characteristics of the receiving converter station of the flexible DC distribution network, the problems of high-frequency oscillation determination complexity and information loss in the existing technology are solved, and efficient failures are achieved Improve judgment and system stability.

CN119066468APending Publication Date: 2024-12-03ELECTRIC POWER RESEARCH INSTITUTE OF STATE GRID JIBEI ELECTRIC POWER CO LTD +2
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411138106.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

High-frequency oscillation events in flexible DC systems lead to reduced system stability, increasing accident risk, and impact on the AC power grid. The existing high-frequency resonance fault determination methods have high quality requirements for feature extraction and are prone to loss of important information, resulting in complex and flexible mode judgment. Poor nature and difficult to adapt to the problem of big data classification.

Method used

By extracting the sensitive control signal of the current inner loop in the receiving converter station of the flexible DC distribution network, the high-frequency resonance characteristics are extracted using the convolutional neural network to generate two-dimensional feature samples, and determine whether a high-frequency resonance fault occurs, combined with Gla The M-angle and field are encoded and reconstructed, and the training and testing data are generated, the convolutional neural network model is generated, and feature extraction and fault judgment are embedded in the current inner loop.

Benefits of technology

It improves the accuracy and performance of high-frequency resonant feature extraction, avoids excessive feature extraction quality requirements and information loss, simplifies the pattern judgment process, adapts to big data classification, and enhances the stability and security of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066468A_ABST
    Figure CN119066468A_ABST
Patent Text Reader

Abstract

The invention provides a high-frequency resonance fault judgment method and device. The corresponding method comprises the following steps: acquiring a sensitive control signal from a current inner loop of a target receiving end converter station; extracting high-frequency resonance characteristics of the target receiving end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training a two-dimensional feature sample converted from historical data of the sensitive control signal of the target receiving end converter station; and judging whether the target receiving end converter station has a high-frequency resonance fault or not according to the high-frequency resonance characteristics. The method provided by the invention avoids the problem that the traditional extraction method is easy to lose important information, can accurately extract the high-frequency resonance characteristics of the control signal of the converter station, and provides an identification basis for the judgment of the high-frequency resonance phenomenon of the flexible direct current system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of flexible direct current power transmission, and in particular to a method and device for determining a high-frequency resonance fault. Background Art

[0002] With the rapid development of new energy power generation technology, the power system is developing towards "high proportion of renewable energy" and "high proportion of power electronic equipment". Among them, modular multilevel converter (MMC) has become the key equipment for flexible DC transmission due to its many advantages such as high modularity, low switching loss, convenient expansion capacity, and low output harmonics. It has been applied in high-voltage long-distance power transmission, new energy power generation grid connection, power supply in remote areas, asynchronous power grid interconnection, and urban power grid interconnection. It is not difficult to imagine that with the large-scale development of new energy such as hydropower, wind power, and photovoltaics, flexible DC transmission technology based on MMC topology will be more widely used.

[0003] However, due to the large number of cascaded sub-modules in the MMC bridge arm and the complex internal characteristics, the resulting system stability issues cannot be underestimated. For example, high-frequency oscillation events have occurred in multiple blocks. When such high-frequency oscillations occur, the system voltage and current are severely distorted, containing high-order harmonics with large amplitudes, which will not only reduce system stability and increase the risk of accidents. It will also bring safety hazards to the core power electronic equipment in the DC power grid, specifically, the high-frequency oscillation event will cause the converter station to be locked, and the resulting power surplus / deficit will have a serious impact on the connected AC power grid. Summary of the invention

[0004] The high-frequency resonance fault judgment method and device provided by the present invention are intended to extract the high-frequency characteristics of the control signal at the receiving-end converter station of the flexible DC system, and judge whether high-frequency resonance occurs at the receiving-end converter station of the flexible DC system through the extracted high-frequency characteristics, so as to provide preliminary technical support for the subsequent startup of the corresponding resonance suppression strategy.

[0005] Another object of the present invention is to provide a high-frequency resonance fault determination device. Another object of the present invention is to provide an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, the processor implementing the steps of the above-mentioned high-frequency resonance fault determination method when executing the computer program. Another object of the present invention is to provide a readable medium, on which a computer program is stored, and when the computer program is executed by the processor, the steps of the above-mentioned high-frequency resonance fault determination method are implemented.

[0006] In order to achieve the above-mentioned purpose, in a first aspect, this party provides a high-frequency resonance fault determination method applicable to a receiving-end converter station of a flexible DC distribution network, comprising:

[0007] Obtain sensitive control signals from the current inner loop of the target receiving-end converter station;

[0008] The high-frequency resonance characteristics of the target receiving-end converter station are extracted according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station;

[0009] It is determined whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0010] In one embodiment of the present invention, the step of generating the convolutional neural network includes:

[0011] Performing polar coordinate encoding on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data;

[0012] Generate the two-dimensional feature sample training data and the two-dimensional feature sample test data according to the Gram angle sum field and the Gram angle difference field;

[0013] The convolutional neural network is generated according to the two-dimensional feature sample training data, the two-dimensional feature sample test data and the initial model of the convolutional neural network.

[0014] In one embodiment of the present invention, polar coordinate encoding is performed on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data, including:

[0015] Performing polar coordinate encoding on the historical data to generate encoded historical data;

[0016] The encoded historical data is reconstructed to generate the Gram angle sum field and the Gram angle difference field.

[0017] In one embodiment of the present invention, reconstructing the encoded historical data includes:

[0018] Calculating angles and cosine values ​​between different historical data to generate the Gram angle and field;

[0019] The sine value of the angle difference between different historical data is calculated to generate the Gram angle difference field.

[0020] In an embodiment of the present invention, the sensitive control signal includes: an alternating current dq component, an alternating voltage dq component, and an alternating current reference value dq component.

[0021] In one embodiment of the present invention, a high-frequency resonance fault determination method applicable to a receiving-end converter station of a flexible DC distribution network further includes:

[0022] The convolutional neural network is arranged on the input link of the alternating current dq component of the current inner loop, the input link of the alternating voltage dq component and the input link of the alternating current reference value dq component to obtain the sensitive control signal.

[0023] In a second aspect, the present invention provides a high-frequency resonance fault determination device, which is applicable to a receiving-end converter station of a flexible DC distribution network, and the device comprises:

[0024] A sensitive control signal acquisition module is used to acquire sensitive control signals from the current inner loop of the target receiving-end converter station;

[0025] A high-frequency resonance feature extraction module, used to extract the high-frequency resonance feature of the target receiving-end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station;

[0026] A fault judgment module is used to judge whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0027] In one embodiment of the present invention, a high-frequency resonance fault determination device applicable to a receiving-end converter station of a flexible DC distribution network further includes:

[0028] A network generation module, used to generate the convolutional neural network, the network generation module comprising:

[0029] A historical data polar coordinate encoding unit, used for performing polar coordinate encoding on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data;

[0030] A training data generating unit, configured to generate the two-dimensional feature sample training data and the two-dimensional feature sample test data according to the Gram angle sum field and the Gram angle difference field;

[0031] A network generation unit is used to generate the convolutional neural network based on the two-dimensional feature sample training data, the two-dimensional feature sample test data and the initial model of the convolutional neural network.

[0032] In one embodiment of the present invention, the historical data polar coordinate encoding unit includes:

[0033] A coded data generating unit, used for performing polar coordinate coding on the historical data to generate coded historical data;

[0034] The historical data reconstruction unit is used to reconstruct the encoded historical data to generate the Gram angle sum field and the Gram angle difference field.

[0035] In one embodiment of the present invention, the historical data reconstruction unit includes:

[0036] A Gram angle and field generating unit, used for calculating angles and cosine values ​​between different historical data to generate the Gram angle and field;

[0037] The Gram angle difference field generating unit is used to calculate the sine value of the angle difference between different historical data to generate the Gram angle difference field.

[0038] In an embodiment of the present invention, the sensitive control signal includes: an alternating current dq component, an alternating voltage dq component, and an alternating current reference value dq component.

[0039] In one embodiment of the present invention, a high-frequency resonance fault determination device applicable to a receiving-end converter station of a flexible DC distribution network further includes:

[0040] A network setting module is used to set the convolutional neural network on the input link of the AC current dq component of the current inner loop, the input link of the AC voltage dq component and the input link of the AC current reference value dq component to obtain the sensitive control signal.

[0041] In a third aspect, the present invention provides a computer program product, comprising a computer program / instruction, which implements the steps of a high-frequency resonance fault determination method when executed by a processor.

[0042] In a fourth aspect, the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a high-frequency resonance fault determination method when executing the program.

[0043] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of a high-frequency resonance fault determination method when executed by a processor.

[0044] From the above description, it can be seen that the high-frequency resonance fault determination method and device provided by the embodiment of the present invention are applicable to the receiving-end converter station of the flexible DC distribution network, and the corresponding method includes: first, obtaining a sensitive control signal from the current inner loop of the target receiving-end converter station; then, extracting the high-frequency resonance characteristics of the target receiving-end converter station according to the sensitive control signal and the pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station; finally, judging whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0045] The present invention overcomes the problems in the existing high-frequency resonance fault judgment method that the quality requirements for high-frequency resonance feature extraction are too high and important information is easily lost, resulting in a complex pattern judgment process, poor flexibility, and difficulty in adapting to big data classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 The process diagram of the high frequency resonance fault determination method provided in the embodiment of the present invention is as follows Figure 1 ;

[0048] Figure 2 The process diagram of the high frequency resonance fault determination method provided in the embodiment of the present invention is as follows Figure 2 ;

[0049] Figure 3 4 is a flow chart of step 400 of the high frequency resonance fault determination method in an embodiment of the present invention;

[0050] Figure 4 This is a flow chart of step 401 of the high-frequency resonance fault determination method in an embodiment of the present invention;

[0051] Figure 5 This is a flow chart of step 4012 of the high-frequency resonance fault determination method in an embodiment of the present invention;

[0052] Figure 6 The process diagram of the high frequency resonance fault determination method provided in the embodiment of the present invention is as follows Figure 3 ;

[0053] Figure 7 It is an electrical topology connection diagram of a flexible DC system in a specific application example of the present invention;

[0054] Figure 8 It is a flow chart of a method for determining a high-frequency resonance fault in a specific application example of the present invention;

[0055] Fig. 9 A mind map of a high-frequency resonance fault determination method in a specific application example of the present invention;

[0056] Fig.10 is a schematic diagram of the AC voltage d component in a specific application example of the present invention;

[0057] Fig.11 It is a schematic diagram of the d component of the alternating current in a specific application example of the present invention;

[0058] Fig.12 It is a schematic diagram of the reference value d component of the alternating current in a specific application example of the present invention;

[0059] Fig.13 It is a VF control structure diagram of a receiving-end converter station in a specific application example of the present invention;

[0060] Fig.14 It is a schematic diagram of the d-axis component of the AC voltage in a steady-state process in a specific application example of the present invention;

[0061] Fig.15 It is a schematic diagram of the d-axis component of the AC voltage during the transition process in a specific application example of the present invention;

[0062] Fig.16 It is a schematic diagram of the d-axis component of the AC voltage during the oscillation process in a specific application example of the present invention;

[0063] Fig.17 It is a schematic diagram of the d-axis component of the alternating current in a steady-state process in a specific application example of the present invention;

[0064] Fig.18 It is a schematic diagram of the d-axis component of the alternating current in the transition process in a specific application example of the present invention;

[0065] Fig.19 It is a schematic diagram of the d-axis component of the alternating current in the oscillation process in a specific application example of the present invention;

[0066] Fig. 20 It is a schematic diagram of the d-axis component of the AC current reference value in the steady-state process in a specific application example of the present invention;

[0067] Fig.21 It is a schematic diagram of the d-axis component of the AC current reference value in the transition process in a specific application example of the present invention;

[0068] Fig. 22 It is a schematic diagram of the d-axis component of the AC current reference value in the oscillation process in a specific application example of the present invention;

[0069] Fig.23 It is a schematic diagram of a training image corresponding to the d-axis component of the AC voltage in a steady-state process in a specific application example of the present invention;

[0070] Fig.24 It is a schematic diagram of a training image corresponding to the d-axis component of the AC voltage in the transition process in a specific application example of the present invention;

[0071] Fig.25 It is a schematic diagram of a training image corresponding to the d-axis component of the AC voltage in the oscillation process in a specific application example of the present invention;

[0072] Fig.26 It is a schematic diagram of a training image corresponding to the d-axis component of the alternating current in a steady-state process in a specific application example of the present invention;

[0073] Fig. 27 It is a schematic diagram of a training image corresponding to the d-axis component of the alternating current in the transition process in a specific application example of the present invention;

[0074] Fig.28 It is a schematic diagram of a training image corresponding to the d-axis component of the alternating current in the oscillation process in a specific application example of the present invention;

[0075] Fig.29 It is a schematic diagram of a training image corresponding to the d-axis component of the AC current reference value in a steady-state process in a specific application example of the present invention;

[0076] Fig.30 It is a schematic diagram of a training image corresponding to the d-axis component of the AC current reference value in the transition process in a specific application example of the present invention;

[0077] Fig.31 It is a schematic diagram of a training image corresponding to the d-axis component of the AC current reference value in the oscillation process in a specific application example of the present invention;

[0078] Fig.32 It is a schematic diagram of the training result corresponding to the d-axis component of the AC voltage in a specific application example of the present invention;

[0079] Fig.33 It is a schematic diagram of the training result corresponding to the d-axis component of the alternating current in a specific application example of the present invention;

[0080] Fig.34 It is a schematic diagram of the training result corresponding to the reference value of the d-axis component of the alternating current in a specific application example of the present invention;

[0081] Fig.35 The structure of the high frequency resonance fault determination device in the embodiment of the present invention is shown in FIG. Figure 1 ;

[0082] Fig.36 The structure of the high frequency resonance fault determination device in the embodiment of the present invention is shown in FIG. Figure 2 ;

[0083] Fig.37 A schematic diagram of the structure of the network generation module 40 in an embodiment of the present invention;

[0084] Fig.38 Schematic diagram of the structure of the historical data polar coordinate encoding unit 40a in an embodiment of the present invention;

[0085] Fig.39 Schematic diagram of the structure of the historical data reconstruction unit 40a2 in an embodiment of the present invention;

[0086] Fig.40 The structure of the high frequency resonance fault determination device in the embodiment of the present invention is shown in FIG. Figure 3 ;

[0087] Fig.41 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0088] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0090] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or apparatuses.

[0091] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0092] In view of the technical pain points of MMC described in the background technology section, it is necessary to extract the high-frequency characteristics of the control signal at the receiving-end converter station of the flexible DC system, and use feature extraction technology to determine whether high-frequency resonance occurs at the receiving-end converter station of the flexible DC system, so as to provide preliminary technical support for the subsequent launch of the corresponding resonance suppression strategy.

[0093] In the study of high-frequency resonance feature extraction of flexible DC systems, existing high-frequency resonance feature extraction methods for flexible DC systems mainly include time domain, frequency domain, time-frequency domain and other methods, such as empirical mode decomposition, wavelet transform, power spectrum entropy, etc.; traditional fault classification methods include BP neural network, support vector machine (SVM), random forest, etc. However, traditional methods have too high requirements on the quality of feature extraction and insufficient learning depth, which leads to complex pattern judgment process, poor flexibility, and difficulty in adapting to big data classification. For this reason, it is necessary to find a new method for extracting high-frequency resonance features at the receiving-end converter station of a flexible DC system. Based on this, an embodiment of the present invention provides a specific implementation method of a high-frequency resonance fault judgment method, which is applicable to the receiving-end converter station of a flexible DC distribution network, see. Figure 1 , the method specifically includes the following contents:

[0094] Step 100: Acquire a sensitive control signal from the current inner loop of the target receiving-end converter station;

[0095] Step 200: extracting the high-frequency resonance characteristics of the target receiving-end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station;

[0096] Step 300: Determine whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0097] From the above description, it can be seen that the high-frequency resonance fault determination method provided by the embodiment of the present invention includes: first, obtaining a sensitive control signal from the current inner loop of the target receiving-end converter station; then, extracting the high-frequency resonance characteristics of the target receiving-end converter station based on the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station; finally, judging whether a high-frequency resonance fault occurs in the target receiving-end converter station based on the high-frequency resonance characteristics.

[0098] The present invention overcomes the problems in the existing high-frequency resonance fault judgment method that the quality requirements for high-frequency resonance feature extraction are too high and important information is easily lost, resulting in a complex pattern judgment process, poor flexibility, and difficulty in adapting to big data classification.

[0099] In some embodiments of the present application, the current inner loop of step 100 is a control mechanism suitable for voltage and current control in non-power electronic equipment and industrial automation systems. The voltage outer loop and current inner loop control are mainly divided into two levels in principle, namely voltage outer loop control and current inner loop control. Accurate voltage and current control can be achieved through these two levels of control. In the voltage outer loop control, the goal of the controller is to regulate and maintain the stability of the voltage signal. The voltage outer loop control usually uses a PI controller (proportional integral controller) to correct the voltage error. The output of the PI controller adjusts the voltage output based on the accumulation of the current voltage error and the integral voltage error. When the voltage error is large, the PI controller will output a larger correction signal to increase the voltage output. When the voltage error is small, the controller will output a smaller correction signal to reduce the voltage output. Through this adjustment method, the voltage can be kept near the set value.

[0100] The current inner loop control is based on the voltage outer loop control, and its goal is to regulate and maintain the stability of the current signal. The current inner loop control usually uses a PI controller to correct the current error. The output of the PI controller adjusts the current output based on the accumulation of the current current error and the integral current error. When the current error is large, the PI controller will output a larger correction signal to increase the current output. When the current error is small, the I controller will output a smaller correction signal to reduce the current output. Through this regulation method, the current can be kept near the set value.

[0101] It should be pointed out that in the voltage outer loop and current inner loop control, the voltage outer loop control is responsible for maintaining the stability of the voltage, and the current inner loop control is responsible for maintaining the stability of the current. In practical applications, the relationship between voltage and current is mutually influential. When the voltage changes, it will cause the current to change; when the current changes, it will also affect the change of voltage.

[0102] In addition, the sensitive control signal of the current inner loop needs to be general and not change with the change of the control mode of the receiving converter station.

[0103] In some embodiments of the present application, the convolutional neural network (CNN) of step 200 is a deep learning model, which is mainly used for image recognition and processing tasks, and realizes automatic feature extraction and classification of images by simulating the processing mode of the human visual system.

[0104] CNN is mainly composed of convolutional layer, pooling layer, fully connected layer and activation function. The convolutional layer is the core part of CNN, which extracts features from the input image through convolution operation. The convolution operation uses a set of learnable filters (also called convolution kernels) to perform sliding window calculations on the input image to obtain a series of feature maps, which represent the features of the input image at different positions and scales. The pooling layer is used to reduce the size of the feature map, reduce the computational complexity, and make the features translation invariant. Commonly used pooling operations include maximum pooling and average pooling, which select the maximum value or average value in the local area as the output respectively. The fully connected layer connects the outputs of the convolutional layer and the pooling layer, and implements the final classification or regression task through a series of fully connected operations. The activation function introduces nonlinear transformations to increase the expressive power of the model. Commonly used activation functions include ReLU, Sigmoid, and Tanh.

[0105] CNN gradually extracts low-level to high-level features of the image through multi-layer convolution and pooling operations, and performs classification or regression tasks through fully connected layers. It has the characteristics of parameter sharing, local perception, and spatial translation invariance, which makes it perform well in image processing tasks. Based on this, this application selects CNN as an effective way to identify the high-frequency resonance characteristics of the current inner loop sensitive control signal. As a supervised learning deep learning algorithm, CNN has obvious advantages over traditional feature recognition methods in the field of feature extraction. Using CNN for feature extraction and recognition, the frequency domain features of the current inner loop sensitive control signal can be automatically extracted through training after the training is completed, and a classification model corresponding to the characteristic sample data of the steady-state process, transition process and oscillation process is established, so that the feature extraction accuracy and performance are improved.

[0106] In some embodiments of the present application, see Figure 2 , a high frequency resonance fault determination method further includes:

[0107] Step 400: Generate the convolutional neural network, see Figure 3 , step 400 comprises:

[0108] Step 401: performing polar coordinate encoding on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data;

[0109] Preferably, the historical data is a time series. It is understood that the Gramian angular field (GAF) is composed of the inner product of two vectors, which can preserve the time dependence of the time series. Before the Gramian matrix transformation, the time series needs to be spatially transformed, and the specific operation is to convert the Cartesian coordinate system into a polar coordinate system. This can effectively distinguish between value information and Gaussian noise.

[0110] Step 402: Generate the two-dimensional feature sample training data and the two-dimensional feature sample test data according to the Gram angle sum field and the Gram angle difference field;

[0111] GAF is selected as a way to convert time series signals into two-dimensional images. GAF is a coding method that combines coordinate transformation and Gram matrix to transform time series into images. The Gram matrix is ​​composed of the inner product of two vectors and can preserve the time dependency of the time series. In addition, before performing the Gram matrix transformation, the time series needs to be spatially transformed, and the specific operation is to convert the Cartesian coordinate system into the polar coordinate system. This can effectively distinguish between value information and Gaussian noise.

[0112] Step 403: Generate the convolutional neural network according to the two-dimensional feature sample training data, the two-dimensional feature sample test data and the initial model of the convolutional neural network.

[0113] Specifically, in step 402 and step 403, two-dimensional feature sample data is generated according to the Gram angle sum field and the Gram angle difference field, and the two-dimensional feature sample data is randomly divided into two-dimensional feature sample training data and two-dimensional feature sample test data, the two-dimensional feature sample training data is used to train the convolutional neural network, and the two-dimensional feature sample test data is used to test the training result, and when the training result reaches a preset expected value, the training and testing of the convolutional neural network are stopped. Preferably, the two-dimensional feature sample training data is a two-dimensional picture.

[0114] CNN is preferred as an effective way to identify the high-frequency resonance characteristics of the current inner loop sensitive control signal. As a supervised learning deep learning algorithm, CNN has obvious advantages over traditional feature recognition methods in the field of feature extraction. Using CNN for feature extraction and recognition, the frequency domain characteristics of the current inner loop sensitive control signal can be automatically extracted after training, and a classification model corresponding to the characteristic sample data of the steady-state process, transition process and oscillation process can be established, so that the feature extraction accuracy and performance are improved.

[0115] In some embodiments of the present application, see Figure 4 , step 401 comprises:

[0116] Step 4011: performing polar coordinate encoding on the historical data to generate encoded historical data;

[0117] Step 4012: reconstruct the encoded historical data to generate the Gram angle sum field and the Gram angle difference field.

[0118] In some embodiments of the present application, see Figure 5 , step 4012 includes:

[0119] Step 40121: Calculate the angle and cosine value between different historical data to generate the Gram angle and field;

[0120] Step 40122: Calculate the sine value of the angle difference between different historical data to generate the Gram angle difference field.

[0121] In some embodiments of the present application, the sensitive control signal includes: an AC current dq component, an AC voltage dq component, and an AC current reference value dq component.

[0122] It should be pointed out that the sensitive control signal of the current inner loop is general and does not change with the change of the control mode of the receiving converter station. In the controller current inner loop of each receiving converter station of the flexible DC system, the AC current dq component I d ,I q ; AC voltage dq component u d 、u q And the AC current reference value dq component I dref and I qref It is the current inner loop sensitive control signal that this application focuses on.

[0123] In some embodiments of the present application, see Figure 6 , a high-frequency resonance fault determination method applicable to a receiving-end converter station of a flexible DC distribution network, further comprising:

[0124] Step 500: Setting the convolutional neural network on the input link of the AC current dq component of the current inner loop, the input link of the AC voltage dq component and the input link of the AC current reference value dq component to obtain the sensitive control signal.

[0125] In order to make the extraction result of the current inner loop sensitive control signal more reliable, the convolutional neural network is embedded in the receiving end converter station u d 、u q Input link, I d ,I q Component input link and I dref ,I qref Input link.

[0126] Based on the Gram angular field and convolutional neural network theory, the present invention proposes a method for extracting high-frequency resonance features of a receiving-end converter station of a flexible DC system based on modular multi-level converter control signal analysis. First, a two-dimensional feature sample set generated by a Gramian angular summation field (GASF) and a Gramian angular difference field (GADF) is obtained by combining the GAF matrix with polar coordinates for CNN training. Then, the trained CNN is embedded in the inner current loop of the receiving-end converter station control system in the form of a discriminator to determine whether the sensitive control signal has a high-frequency resonance component, thereby completing the extraction of high-frequency resonance features of the receiving-end converter station of the flexible DC system. The method provided by the present invention avoids the problem that traditional extraction methods are prone to losing important information. It can accurately extract the high-frequency resonance features of the converter station control signal, providing an identification basis for determining the high-frequency resonance phenomenon of the flexible DC system.

[0127] To further illustrate the present solution, the present invention provides a specific application example of a high-frequency resonance fault determination method applicable to a receiving-end converter station of a flexible DC distribution network, which specifically includes the following contents.

[0128] See also Figure 7 The present invention first provides a flexible DC system, which includes a sending-end converter station, a DC transmission line, and a receiving-end converter station. At present, the method for extracting high-frequency resonance characteristics of the receiving-end converter station of the flexible DC system faces many problems, the main problems are:

[0129] 1. The algorithms used in the prior art for extracting high-frequency resonance characteristics of the receiving-end converter station of the flexible DC system are usually installed in the electrical topology after having the extraction capability, and cannot extract the high-frequency resonance characteristics of the receiving-end converter station from the source at the control loop level.

[0130] Second, the algorithms used to extract high-frequency resonance features in the receiving-end converter station of the flexible DC system in the past often have too high requirements on the quality of feature extraction and insufficient learning depth, which leads to defects such as complex pattern judgment process, easy loss of important data, and difficulty in adapting to big data classification, which hinders the promotion of various feature recognition algorithms in engineering.

[0131] Based on the above reasons, in the process of extracting high-frequency resonance characteristics of the receiving-end converter station of the flexible DC system, it is necessary to focus on the discriminator's current inner loop (based on Fig.13The invention uses the VF control mode in the current inner loop as an example to determine the sensitive control signal, and puts an algorithm that can simply and accurately reflect the frequency domain characteristics of the current inner loop sensitive control signal into the extraction application of the above signal. To this end, the present invention applies the method provided by the invention to the extraction of high-frequency resonance characteristics of the current inner loop sensitive control signal, thereby avoiding the limitations of traditional methods such as too high requirements on the quality of feature extraction and easy loss of important data.

[0132] Next, see Figure 8 as well as Fig. 9 The specific application example of the high-frequency resonance fault determination method for a receiving-end converter station of a flexible DC distribution network provided by the present invention comprises the following steps:

[0133] S1: Select sensitive control signals in the current inner loop of the receiving-end converter station controller;

[0134] It can be understood that the current inner loop sensitive control signal is general and does not change with the change of the control mode of the receiving end converter station. Preferably, the I in the current inner loop of the controller of each receiving end converter station of the flexible DC system is d ,I q 、u d 、u q ,I dref and I qref As the preferred current inner loop sensitive control signal.

[0135] During the steady-state process, transition process and oscillation process of the current inner loop in the converter station at the receiving end of the flexible DC system, the sensitive control signals of the current inner loop (such as Figures 10 to 12 The transition process here refers to the intermediate process between the steady-state process and the oscillation process due to the change of working conditions.

[0136] Specifically, combined Fig.13 ,u d 、u q , i d 、i q , and are the dq axis components of the AC voltage and AC current and their respective command values. fd and u qf is the dq axis component of the differential mode voltage. vac (s) and H iac (s) is the transfer function of the voltage outer loop and current inner loop PI controller. Under VF control mode, and byu d 、u q And the corresponding instruction value is obtained.

[0137] After the current command value enters the current inner loop, the output of the inner loop PI controller is d(q) and ωLi q(d) The dq axis components of the AC current are calculated. For VF control, the outer loop control target is the AC voltage amplitude and frequency, so u d 、u q Subtract the command value and enter the outer loop to obtain and As shown in formula (1). In view of this, the present application selects the current inner loop related control quantity: the dq component of the AC current (I d ,I q ), AC voltage dq component (u d 、u q ) and the dq component of the AC current reference value (I dref ,I qref ) is used as the current inner loop sensitive control signal as shown.

[0138]

[0139] S2: Establish a two-dimensional feature sample set of sensitive control signals of the receiving converter station;

[0140] Specifically, step S2 includes the following contents: firstly, the time domain data of the current inner loop sensitive control signal is normalized, specifically: I d ,I q 、u d 、u q ,I dref and I qref Normalize the one-dimensional structured data X consisting of 6 variables n =[x1,x2,...,x i ], n = 1...6. Here x i represents the time series observation value corresponding to the i-th time point, n represents the six current inner loop sensitive control signals, and X n Mapped to the interval [0,1], here represents the values ​​of the six current inner loop sensitive control signals after normalization, as shown in formula (2).

[0141]

[0142] Then the time series will be scaled to [-1,1] using the min-max scaler as shown in equation (3). Here Represents the values ​​of the six current inner loop sensitive control signals after minimum-maximum normalization, X n(0)Represents the normalized current inner loop sensitive control signal sequence calculated by formula (2). After data processing by formula (2) and formula (3), the current inner loop sensitive data is mapped to the first and second quadrants of the polar coordinates, which can better describe the relationship between the time points and their time series observation values.

[0143] Next, the time domain data of the current inner loop sensitive control signal is intercepted for multiple times with the same length; the intercepted time domain data of the current inner loop sensitive control signal is Figures 14 to 22 , three types of labels, steady-state process, transition process and oscillation process, are used to construct a two-dimensional feature sample training set of the control signal of the receiving-end converter station; specifically: 0.06s (three power frequency cycles) of time domain data of each state are selected for data classification and label division.

[0144] Finally, the intercepted time domain signal of the current inner loop sensitive control signal is used to construct a data set of polar coordinate features of the current inner loop sensitive control signal through the Gram angle field and the data set is divided into a training set and a test set;

[0145] Will The value of is encoded as a cosine angle and re-expressed in polar coordinates as φ and r, where φ and r are the polar angle and polar diameter, as shown in equation (5).

[0146]

[0147] After calculation by formula (5), the polar angle and radius of the polar coordinates are obtained, and the current inner loop sensitive control signal is converted into polar coordinates, which retains the key characteristic information to the greatest extent. After obtaining the polar angle and radius, the characteristic image is obtained by the definition of the Gram matrix shown in formula (5). The Gramian angular field (GAF) is divided into the Gramian angular sum field (GASF) and the Gramian angular difference field (GADF), each of which defines its own special inner product, that is, the "inner product" between two timing points is the cosine of the sum of the polar angles of the two timing points after the polar coordinates are transformed, and the calculation is shown in formula (6).

[0148]

[0149] By performing triangular transformation on the Gram matrix G in equation (6) as shown in equation (7), the simplified forms of GASF and GADF can be obtained, as shown in equations (8) and (9):

[0150]

[0151] As time goes by, the corresponding data can find corresponding points in the polar coordinate system. That is, for a fixed time series, the mapping in the polar coordinate system has one and only one unique result. In addition, the polar coordinate system retains the degree of correlation between the previous and next time points through the radius size. After calculation and coloring, a two-dimensional feature sample graph is obtained, such as Figures 23 to 31 shown.

[0152] S3: Randomly divide the two-dimensional feature sample set into a training set and a test set. After the CNN is trained, the accuracy of the discriminator is tested.

[0153] Construct a CNN framework, and input the training set into CNN for feature extraction training. Use 70% of the data in the two-dimensional feature sample set as the training set, and 30% of the data in the two-dimensional feature sample set as the test set. Train and test the model, and use the trained model to judge the operating status of the current inner loop sensitive control signal in the test set. If the verification accuracy does not meet the requirements, adjust the interception length of the current inner loop sensitive control signal and the framework of the convolution model; repeat steps S1 to S2 until the verification accuracy meets the requirements.

[0154] S4: If the discriminator meets the accuracy requirement, the discriminator is embedded in the corresponding control link of the current inner loop to extract the high-frequency resonance characteristic samples of the receiving-end converter station. Otherwise, the above steps are repeated.

[0155] The trained CNN discriminator is embedded into the corresponding control link of the current inner loop, such as Fig.13 As shown, the current inner loop is sensitive to the control signal I d ,I q 、u d 、u q ,I dref and I qref Extract and identify high-frequency resonance features.

[0156] First, after training the CNN discriminator according to the parameters listed in Table 1, the model accuracy is used as the evaluation criterion for the discriminator after the discriminator training is completed. Figure 32 to Figure 34 , after training, the accuracy of CNN is ≥95%, which means it can be considered to have the ability to extract high-frequency resonance features of sensitive control signals of the inner loop of the current in the receiving-end converter station of the flexible DC system.

[0157] Table 1

[0158] Operation layer (type) Number of dimensions Total number of learnable parameters The first convolutional layer 227*227*8 392 First Maximal Pooling 225*225*8 0 The second convolutional layer 223*223*16 2064 Second Max Pooling 223*223*16 0 The third convolutional layer 223*223*32 8224 Fully connected layer 1*1*8 12730632 Normalized Mapping 1*1*8 0

[0159] Next, the trained discriminator is added to the inner loop of the control system current. d ,I q 、u d 、u q ,I dref and Iqref The input discriminator judges in real time whether the current inner loop sensitive control signal has the characteristics of high-frequency resonance.

[0160] Based on the same inventive concept, the embodiments of the present application also provide a high-frequency resonance fault determination device, which can be used to implement the method described in the above embodiments, such as the following embodiments. Since the principle of solving the problem by the high-frequency resonance fault determination device is similar to that of the high-frequency resonance fault determination method, the implementation of the high-frequency resonance fault determination device can refer to the implementation of the high-frequency resonance fault determination method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0161] The embodiment of the present invention provides a specific implementation of a high-frequency resonance fault determination device capable of implementing a high-frequency resonance fault determination method, see Fig.35 A high-frequency resonance fault determination device for a receiving-end converter station of a flexible DC distribution network specifically includes the following contents:

[0162] A sensitive control signal acquisition module 10 is used to acquire a sensitive control signal from the current inner loop of the target receiving-end converter station;

[0163] A high-frequency resonance feature extraction module 20 is used to extract the high-frequency resonance feature of the target receiving-end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station;

[0164] The fault judgment module 30 is used to judge whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0165] In one embodiment of the present invention, see Fig.36 , a high-frequency resonance fault determination device applicable to a receiving-end converter station of a flexible DC distribution network, further comprising:

[0166] The network generation module 40 is used to generate the convolutional neural network, see Fig.37 , the network generation module 40 includes:

[0167] A historical data polar coordinate encoding unit 40a, configured to perform polar coordinate encoding on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data;

[0168] A training data generating unit 40b, configured to generate the two-dimensional feature sample training data and the two-dimensional feature sample test data according to the Gram angle sum field and the Gram angle difference field;

[0169] The network generation unit 40c is used to generate the convolutional neural network according to the two-dimensional feature sample training data, the two-dimensional feature sample test data and the initial model of the convolutional neural network.

[0170] In one embodiment of the present invention, see Fig.38 , the historical data polar coordinate encoding unit 40a includes:

[0171] The encoded data generating unit 40a1 is used to perform polar coordinate encoding on the historical data to generate encoded historical data;

[0172] The historical data reconstruction unit 40a2 is used to reconstruct the encoded historical data to generate the Gram angle sum field and the Gram angle difference field.

[0173] In one embodiment of the present invention, see Fig.39 , the historical data reconstruction unit 40a2 includes:

[0174] A Gram angle and field generating unit 40a21, used for calculating angles and cosine values ​​between different historical data to generate the Gram angle and field;

[0175] The Gram angle difference field generating unit 40a22 is used to calculate the sine value of the angle difference between different historical data to generate the Gram angle difference field.

[0176] In an embodiment of the present invention, the sensitive control signal includes: an alternating current dq component, an alternating voltage dq component, and an alternating current reference value dq component.

[0177] In one embodiment of the present invention, see Fig.40 , a high-frequency resonance fault determination device applicable to a receiving-end converter station of a flexible DC distribution network, further comprising:

[0178] The network setting module 50 is used to set the convolutional neural network on the input link of the AC current dq component of the current inner loop, the input link of the AC voltage dq component and the input link of the AC current reference value dq component to obtain the sensitive control signal.

[0179] From the above description, it can be seen that the high-frequency resonance fault judgment device provided in the embodiment of the present invention is suitable for the receiving-end converter station of the flexible DC distribution network, including: a sensitive control signal acquisition module, used to obtain a sensitive control signal from the current inner loop of the target receiving-end converter station; a high-frequency resonance feature extraction module, used to extract the high-frequency resonance characteristics of the target receiving-end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station; a fault judgment module, used to judge whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0180] The present invention overcomes the problems in the existing high-frequency resonance fault judgment method that the quality requirements for high-frequency resonance feature extraction are too high and important information is easily lost, resulting in a complex pattern judgment process, poor flexibility, and difficulty in adapting to big data classification.

[0181] The embodiment of the present application also provides a specific implementation of an electronic device capable of implementing all steps in the high-frequency resonance fault determination method in the above embodiment, see Fig.41 , electronic equipment specifically includes the following:

[0182] Processor (processor) 1201, memory (memory) 1202, communication interface (CommunicationsInterface) 1203 and bus 1204;

[0183] The processor 1201 , the memory 1202 , and the communication interface 1203 communicate with each other via the bus 1204 ; the communication interface 1203 is used to realize information transmission between relevant devices such as the server-side device, the power measurement device, and the user-side device.

[0184] The processor 1201 is used to call the computer program in the memory 1202. When the processor executes the computer program, all the steps in the high-frequency resonance fault determination method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0185] Step 100: Acquire a sensitive control signal from the current inner loop of the target receiving-end converter station;

[0186] Step 200: extracting the high-frequency resonance characteristics of the target receiving-end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station;

[0187] Step 300: Determine whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0188] The embodiment of the present application also provides a computer-readable storage medium capable of implementing all the steps in the high-frequency resonance fault determination method in the above embodiment. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, all the steps in the high-frequency resonance fault determination method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0189] Step 100: Acquire a sensitive control signal from the current inner loop of the target receiving-end converter station;

[0190] Step 200: extracting the high-frequency resonance characteristics of the target receiving-end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station;

[0191] Step 300: Determine whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

[0192] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the hardware + program embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0193] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0194] Although the present application provides method operation steps such as embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in the order of the method shown in the embodiments or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment).

[0195] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0196] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0197] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0198] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A high-frequency resonance fault determination method, applicable to a receiving-end converter station of a flexible DC distribution network, characterized in that: include: Obtain sensitive control signals from the current inner loop of the target receiving-end converter station; The high-frequency resonance characteristics of the target receiving-end converter station are extracted according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station; It is determined whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

2. The high-frequency resonance fault determination method according to claim 1, characterized in that: The steps of generating the convolutional neural network include: Performing polar coordinate encoding on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data; Generate the two-dimensional feature sample training data and the two-dimensional feature sample test data according to the Gram angle sum field and the Gram angle difference field; The convolutional neural network is generated according to the two-dimensional feature sample training data, the two-dimensional feature sample test data and the initial model of the convolutional neural network.

3. The high-frequency resonance fault determination method according to claim 2, characterized in that: Polar coordinate encoding is performed on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data, including: Performing polar coordinate encoding on the historical data to generate encoded historical data; The encoded historical data is reconstructed to generate the Gram angle sum field and the Gram angle difference field.

4. The high-frequency resonance fault determination method according to claim 3, characterized in that: Reconstructing the encoded historical data includes: Calculating angles and cosine values ​​between different historical data to generate the Gram angle and field; The sine value of the angle difference between different historical data is calculated to generate the Gram angle difference field.

5. The high-frequency resonance fault determination method according to claim 1, characterized in that: The sensitive control signal includes: an alternating current dq component, an alternating voltage dq component and an alternating current reference value dq component.

6. The high-frequency resonance fault determination method according to claim 5, characterized in that: Also includes: The convolutional neural network is arranged on the input link of the alternating current dq component of the current inner loop, the input link of the alternating voltage dq component and the input link of the alternating current reference value dq component to obtain the sensitive control signal.

7. A high-frequency resonance fault determination device, suitable for a receiving-end converter station of a flexible DC distribution network, characterized in that: include: A sensitive control signal acquisition module is used to acquire sensitive control signals from the current inner loop of the target receiving-end converter station; A high-frequency resonance feature extraction module, used to extract the high-frequency resonance feature of the target receiving-end converter station according to the sensitive control signal and a pre-generated convolutional neural network; wherein the convolutional neural network extraction is generated by training two-dimensional feature samples converted from historical data of the sensitive control signal of the target receiving-end converter station; A fault judgment module is used to judge whether a high-frequency resonance fault occurs in the target receiving-end converter station according to the high-frequency resonance characteristics.

8. The high-frequency resonance fault determination device according to claim 7, characterized in that: Also includes: A network generation module, used to generate the convolutional neural network, the network generation module comprising: A historical data polar coordinate encoding unit, used for performing polar coordinate encoding on the historical data to generate a Gram angle sum field and a Gram angle difference field corresponding to the historical data; A training data generating unit, configured to generate the two-dimensional feature sample training data and the two-dimensional feature sample test data according to the Gram angle sum field and the Gram angle difference field; A network generation unit is used to generate the convolutional neural network based on the two-dimensional feature sample training data, the two-dimensional feature sample test data and the initial model of the convolutional neural network.

9. The high-frequency resonance fault determination device according to claim 8, characterized in that: The historical data polar coordinate encoding unit comprises: A coded data generating unit, used for performing polar coordinate coding on the historical data to generate coded historical data; The historical data reconstruction unit is used to reconstruct the encoded historical data to generate the Gram angle sum field and the Gram angle difference field.

10. The high-frequency resonance fault determination device according to claim 9, characterized in that: The historical data reconstruction unit comprises: A Gram angle and field generating unit, used for calculating angles and cosine values ​​between different historical data to generate the Gram angle and field; The Gram angle difference field generating unit is used to calculate the sine value of the angle difference between different historical data to generate the Gram angle difference field.

11. The high-frequency resonance fault determination device according to claim 7, characterized in that: The sensitive control signal includes: an alternating current dq component, an alternating voltage dq component and an alternating current reference value dq component.

12. The high-frequency resonance fault determination device according to claim 11, characterized in that: Also includes: A network setting module is used to set the convolutional neural network on the input link of the AC current dq component of the current inner loop, the input link of the AC voltage dq component and the input link of the AC current reference value dq component to obtain the sensitive control signal.

13. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the high-frequency resonance fault determination method according to any one of claims 1 to 6 are implemented.

14. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the high-frequency resonance fault determination method according to any one of claims 1 to 6 are implemented.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the high-frequency resonance fault determination method according to any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • High-frequency resonance control method, device, equipment and storage medium

    CN114498717A

  • Bearing fault diagnosis method based on lightweight neural network and dimension expansion

    CN115761398A

  • Motor fault diagnosis method and equipment

    CN117872127A

  • Single-ended fault positioning method and system for high-voltage direct-current transmission line of hybrid network

    US20220196720A1