DC fault arc detection method, device, equipment, medium and chip

Abstract

The invention relates to the field of power transmission line partial discharge detection, in particular to a DC fault arc detection method, device and equipment, a medium and a chip, and the DC fault arc detection method comprises the steps: collecting an electric signal of a DC power system; the electric signal comprises a normal signal and a fault arc signal; extracting signal features of the electric signals, and constructing a sample set based on the signal features; the signal features comprise two or more of a time-frequency domain feature, a singular value feature, a complexity feature and an energy feature; training a DC fault arc identification network based on the sample set; and performing DC fault arc detection by using the trained DC fault arc identification network. According to the technical scheme, the multi-dimensional mixed feature set is extracted, the DC fault arc recognition network is trained, and then the trained DC fault arc recognition network is used for fault judgment, so that efficient detection of the DC fault arc is realized, and the detection accuracy and real-time performance are improved.

Description

[0001] Technology Neighborhood

[0002] The present disclosure relates to the field of partial discharge detection of power transmission lines, and in particular to a method, device, equipment, medium and chip for detecting direct current fault arcs. Background Art

[0003] With the continuous improvement of modern power demand and technology, DC systems have become increasingly critical in the energy supply, regulation and backup of important industries and facilities. As an important part of the power system, DC substations are used by many power secondary equipment and communication equipment. Once the DC system fails, it will affect the safe and stable operation of the substation secondary system.

[0004] Due to the lack of zero-crossing characteristics, the generation and continued existence of DC arcs can cause serious equipment damage and safety hazards, and may even cause fires. The current DC fault arc detection technology has problems such as low detection accuracy, poor real-time performance, and insufficient anti-interference ability in practical applications. A more effective detection method is urgently needed to improve the detection level of DC fault arcs. Summary of the invention

[0005] In order to solve the problems in the related art, the embodiments of the present disclosure provide a DC fault arc detection method, device, equipment, medium and chip.

[0006] In a first aspect, the present disclosure provides a DC fault arc detection method, comprising:

[0007] Collecting electrical signals of a DC power system; the electrical signals include normal signals and fault arc signals;

[0008] Extracting signal features of the electrical signal, and constructing a sample set based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features, and energy features;

[0009] Training a DC fault arc identification network based on the sample set;

[0010] The trained DC arc fault recognition network is used to detect DC arc faults.

[0011] In one implementation of the present disclosure, the signal features include time-frequency domain features, singular value features, complexity features, and energy features;

[0012] The constructing a sample set based on the signal feature comprises:

[0013] Constructing a multi-dimensional hybrid feature set based on the signal features;

[0014] The multi-dimensional mixed feature set is subjected to dimensionality reduction processing to obtain a low-dimensional feature data set as a sample set.

[0015] In an implementation of the present disclosure, performing dimensionality reduction processing on the multi-dimensional mixed feature set to obtain a low-dimensional feature data set as a sample set includes:

[0016] The multi-dimensional mixed feature set is processed using a locality preserving projection algorithm to obtain a low-dimensional feature data set as a sample set.

[0017] In an implementation of the present disclosure, in the locality preserving projection algorithm, a kernel density estimation method is used to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multi-dimensional mixed feature set.

[0018] In an implementation of the present disclosure, in the locality preserving projection algorithm, using a kernel density estimation method to calculate neighborhood parameters of data sample points to implement dimensionality reduction processing of the multidimensional mixed feature set includes:

[0019] Calculate the neighborhood probability density as:

[0020]

[0021] The neighborhood parameters are calculated as:

[0022]

[0023] Among them, x i is the target point in the sample set for which the probability density is to be estimated, x d is the data sample point in the sample set; N is the number of samples in the sample set; φ(x) is the window function, and satisfies φ(x)≥0,∫φ(x)dx=0; h is the window width; V is the window volume; is the average probability density of the sample set; ε0 is the initial neighborhood parameter.

[0024] In one implementation of the present disclosure, the DC fault arc identification network is a convolutional neural network.

[0025] In a second aspect, the present disclosure provides a DC fault arc detection device, comprising:

[0026] A collection module is configured to collect electrical signals of a DC power system; the electrical signals include normal signals and fault arc signals;

[0027] A construction module is configured to extract signal features of the electrical signal and construct a sample set based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features and energy features;

[0028] A training module, configured to train a DC fault arc identification network based on the sample set;

[0029] The detection module is configured to perform DC fault arc detection using a trained DC fault arc recognition network.

[0030] In one implementation of the present disclosure, the signal features include time-frequency domain features, singular value features, complexity features, and energy features;

[0031] The building blocks include:

[0032] A construction unit, configured to construct a multi-dimensional hybrid feature set based on the signal features;

[0033] The dimensionality reduction processing unit is configured to perform dimensionality reduction processing on the multi-dimensional mixed feature set to obtain a low-dimensional feature data set as a sample set.

[0034] In one implementation of the present disclosure, the dimension reduction processing unit includes:

[0035] The dimensionality reduction processing subunit is configured to process the multi-dimensional mixed feature set using a local preservation projection algorithm to obtain a low-dimensional feature data set as a sample set.

[0036] In an implementation of the present disclosure, in the locality preserving projection algorithm, a kernel density estimation method is used to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multi-dimensional mixed feature set.

[0037] In an implementation of the present disclosure, in the locality preserving projection algorithm, using a kernel density estimation method to calculate neighborhood parameters of data sample points to implement dimensionality reduction processing of the multidimensional mixed feature set includes:

[0038] Calculate the neighborhood probability density as:

[0039]

[0040] The neighborhood parameters are calculated as:

[0041]

[0042] Among them, x i is the target point in the sample set for which the probability density is to be estimated, x d is the data sample point in the sample set; N is the number of samples in the sample set; φ(x) is the window function, and satisfies φ(x)≥0,∫φ(x)dx=0; h is the window width; V is the window volume; is the average probability density of the sample set; ε0 is the initial neighborhood parameter.

[0043] In one implementation of the present disclosure, the DC fault arc identification network is a convolutional neural network.

[0044] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer instructions, and wherein the one or more computer instructions are executed by the processor to implement a method as described in any one of the first aspects.

[0045] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the method described in any one of the first aspects is implemented.

[0046] In a fifth aspect, an embodiment of the present disclosure provides a chip, the chip including a processor, the processor being used to call a computer program in a memory to execute the above-mentioned DC fault arc detection method.

[0047] The technical effects provided by the embodiments of the present disclosure may include the following beneficial effects:

[0048] According to the technical solution provided by the embodiment of the present disclosure, a DC fault arc detection method includes collecting electrical signals of a DC power system; the electrical signals include normal signals and fault arc signals; extracting signal features of the electrical signals, and constructing a sample set based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features, and energy features; training a DC fault arc recognition network based on the sample set; and performing DC fault arc detection using the trained DC fault arc recognition network. In the above technical solution, by extracting a multi-dimensional mixed feature set, that is, extracting features of at least two dimensions of time-frequency domain features, singular value features, complexity features, and energy features, training a DC fault arc recognition network, and then using the trained DC fault arc recognition network for fault judgment, efficient detection of DC fault arcs is achieved, detection accuracy and real-time performance are improved, anti-interference capability is enhanced, and the DC power distribution system can be adapted to different scenarios.

[0049] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings.

[0051] Figure 1 A flow chart of a DC fault arc detection method according to an embodiment of the present disclosure is shown.

[0052] Figure 2A flow chart of a DC fault arc detection method according to a specific embodiment of the present disclosure is shown.

[0053] Figure 3 A structural block diagram of a DC fault arc detection device according to an embodiment of the present disclosure is shown.

[0054] Figure 4 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0055] Figure 5 A schematic diagram showing the structure of a computer system suitable for implementing the method according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0056] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts irrelevant to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0057] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate the presence of features, numbers, steps, behaviors, components, parts, or a combination thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other features, numbers, steps, behaviors, components, parts, or a combination thereof exist or are added.

[0058] It should also be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0059] Although the existing fault arc detection is relatively mature in AC fault arc detection, it has certain defects in DC fault arc detection. DC fault arcs are random and uncertain, which leads to a complex detection process, long data processing time, and low detection accuracy. The decomposition of DC fault arc signals is a key step in determining the detection speed and detection accuracy. For example, the traditional method of fault diagnosis that relies solely on electrical parameters is easily interfered by environmental noise, and when the power grid fluctuates or surges occur, the electrical parameters in the system change, which can easily cause fault omissions and misjudgments.

[0060] Taking the above-mentioned defects into consideration, the DC fault arc detection method provided by the present invention extracts a multi-dimensional mixed feature set, namely, extracts features of at least two dimensions among time-frequency domain features, singular value features, complexity features and energy features, trains a DC fault arc recognition network, and then uses the trained DC fault arc recognition network to perform fault judgment, thereby realizing efficient detection of DC fault arcs, improving detection accuracy and real-time performance, enhancing anti-interference capability, and being able to adapt to different DC power distribution system scenarios.

[0061] Figure 1 A flow chart of a DC fault arc detection method according to an embodiment of the present disclosure is shown.

[0062] like Figure 1 As shown, the DC fault arc detection method includes the following steps S110-S140:

[0063] In step S110, an electrical signal of a DC power system is collected; the electrical signal includes a normal signal and a fault arc signal;

[0064] In step S120, the signal features of the electrical signal are extracted, and a sample set is constructed based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features, and energy features;

[0065] In step S130, a DC arc fault identification network is trained based on the sample set;

[0066] In step S140, the trained DC arc fault identification network is used to perform DC arc fault detection.

[0067] In the disclosed method, the electrical signal in step S110 may be a current signal or a voltage signal. Specifically, a high-precision current sensor and a voltage sensor are used to collect the current signal and the voltage signal in the DC power system in real time.

[0068] In the disclosed method, the time-frequency domain features in step S120 may be time-frequency domain features such as the amplitude of the current signal and the current change rate. The singular value features may be singular value features of the electrical signal extracted using the singular value decomposition technique. The complexity features may be complexity features such as the amplitude entropy of the electrical signal. The energy features may be features such as the energy mean and energy distribution of the electrical signal. Among them, amplitude entropy (AE) is an indicator for measuring the complexity of a signal, and is often used to analyze the randomness and structure of a signal. The core idea of ​​amplitude entropy is whether the change in signal amplitude has a certain regularity or structure. If the amplitude change of a signal is very regular, its amplitude entropy is low; if the amplitude change of the signal has greater randomness or complexity, its amplitude entropy is high. Specifically, amplitude entropy considers the amplitude change of the signal, and does not care about the phase or frequency components of the signal. It is a measure of the change in signal amplitude and reflects the complexity of the signal.

[0069] In the disclosed method, the sample set includes training samples and test samples. When the DC fault arc recognition network is trained based on the sample set in step S130, the DC fault arc recognition network is trained using the training samples and tested using the test samples until the accuracy of the model prediction reaches the preset requirement, a trained DC fault arc recognition network is obtained.

[0070] In one implementation of the present disclosure, the signal features include time-frequency domain features, singular value features, complexity features and energy features. Specifically, the signal features may include five types of feature quantities: the amplitude of the current signal, the current change rate, the singular value feature, the amplitude entropy and the energy mean.

[0071] The step S120 of constructing a sample set based on the signal features includes:

[0072] Constructing a multi-dimensional hybrid feature set based on the signal features;

[0073] The multi-dimensional mixed feature set is subjected to dimensionality reduction processing to obtain a low-dimensional feature data set as a sample set.

[0074] In the disclosed method, a multidimensional mixed feature set including five types of feature quantities, namely, the amplitude of the current signal, the current change rate, the singular value feature, the amplitude entropy and the energy mean, is constructed, and then the multidimensional mixed feature set is subjected to dimensionality reduction processing to obtain a low-dimensional feature data set, and the optimal discriminant feature is selected as the sample set, thereby improving the accuracy of fault arc detection.

[0075] In the disclosed method, principal component analysis (PCA) may be used to perform data dimensionality reduction. As a classic algorithm for data dimensionality reduction, PCA can effectively remove linear correlations between features while maintaining the main information of the original features.

[0076] In the disclosed method, considering that principal component analysis only considers the global distribution characteristics, that is, only considers the maximization of the coefficient variance after projection, and fails to fully consider the preservation of local information, the variables obtained after PCA dimensionality reduction can reflect most of the global information of the original variables, but will disrupt the local neighborhood structure of the original data, resulting in the loss of the ability to distinguish between different categories and thus reducing the fault recognition rate. Therefore, in the disclosed method, locality preserving projections (LPP) can also be used to replace PCA for data dimensionality reduction. LPP is a linear approximation of the nonlinear method Laplacian Eigenmap. As a new subspace analysis method, it not only solves the problem that the principal component analysis method is difficult to maintain the local structure of the nonlinear manifold of the original data, but also solves the problem that it cannot remove the nonlinear correlation between indicators.

[0077] In an implementation of the present disclosure, performing dimensionality reduction processing on the multi-dimensional mixed feature set to obtain a low-dimensional feature data set as a sample set includes:

[0078] The multi-dimensional mixed feature set is processed using a locality preserving projection algorithm to obtain a low-dimensional feature data set as a sample set.

[0079] In the disclosed method, the LPP algorithm constructs the distance relationship between each sample point in the space and finds the corresponding projection matrix so that the sum of the weighted distance squares between neighboring data points in the low-dimensional space is minimized, and the local neighborhood structure of the spatial sample is maintained during the projection process. The dimensionality reduction result obtained by the LPP algorithm can well maintain the manifold structure of the original data sample. At the same time, the algorithm has low requirements on the data structure and can effectively avoid the divergence of the data set.

[0080] In an implementation of the present disclosure, in the locality preserving projection algorithm, a kernel density estimation method is used to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multi-dimensional mixed feature set.

[0081] In the disclosed method, in the process of data dimensionality reduction using the LPP algorithm, it is necessary to determine the neighborhood parameter ε. The neighborhood parameter ε is a parameter for calculating the weight matrix. If the distance between two data sample points is less than ε, the two are considered to be neighbors. When a unified neighborhood parameter is manually selected, the dimensionality reduction effect is uncontrollable due to the uneven distribution of the data set in the high-dimensional space. Therefore, it is necessary to find a method that can adaptively adjust the neighborhood parameters according to the distribution of the data, so that the resulting low-dimensional fault feature data set can better reflect the manifold characteristics of the data. In the disclosed method, an adaptive LPP algorithm is proposed, that is, a kernel density estimation method is used to calculate the neighborhood parameters of data sample points. Specifically, the kernel density estimation is the Parzen window method. The Parzen window method can estimate the overall probability density of the data sample when the distribution of the data sample is uncertain. This feature can be used to adjust the neighborhood parameters of the data sample points in the LPP algorithm, so that the neighborhood parameters can be adaptively adjusted to achieve a better dimensionality reduction effect.

[0082] In an implementation of the present disclosure, in the locality preserving projection algorithm, using a kernel density estimation method to calculate neighborhood parameters of data sample points to implement dimensionality reduction processing of the multidimensional mixed feature set includes:

[0083] Calculate the neighborhood probability density as:

[0084]

[0085] The neighborhood parameters are calculated as:

[0086]

[0087] Among them, x i is the target point in the sample set for which the probability density is to be estimated, x d is the data sample point in the sample set; N is the number of samples in the sample set; φ(x) is the window function, and satisfies φ(x)≥0,∫φ(x)dx=0; h is the window width; V is the window volume; is the average probability density of the sample set; ε0 is the initial neighborhood parameter.

[0088] In the disclosed method, the initial neighborhood parameter ε0 can be set to the same value as the window width h. The window function φ(x) can select a normal window function with good smoothness, specifically:

[0089]

[0090] Through the above adaptive adjustment method, the neighborhood parameters can be dynamically adjusted according to the data distribution to ensure that the manifold structure of the data is better maintained during the data dimensionality reduction process. The low-dimensional feature data set obtained in this way can more accurately reflect the intrinsic characteristics of the original high-dimensional data.

[0091] In one implementation of the present disclosure, the DC fault arc identification network is a convolutional neural network.

[0092] In the disclosed method, the convolutional neural network may include a convolutional layer, a pooling layer, and a fully connected layer. The network model may be constructed by referring to the prior art, which is not the focus of the disclosed method. The low-dimensional feature data set after dimensionality reduction is input into the convolutional neural network for training and testing, and the trained DC fault arc recognition network is used to classify the new input data to determine whether a DC fault arc is generated.

[0093] The DC fault arc detection method provided by the present disclosure further improves the accuracy of fault arc detection by constructing a multi-dimensional hybrid feature set and combining LPP dimension reduction and convolutional neural network fault discrimination. LPP dimension reduction processing reduces the feature dimension, and convolutional neural network improves the calculation efficiency and enhances the real-time performance of detection. LPP dimension reduction processing can also effectively remove redundant information, and then use convolutional neural network to extract effective features in the signal, thereby improving the anti-interference ability of the system. The DC fault arc detection method provided by the present disclosure can be applied to DC power systems of different scales and complexities, and can effectively improve the detection capability and detection accuracy of DC fault arcs, shorten the detection time, and reduce the misjudgment and missed judgment caused by environmental interference.

[0094] Figure 2 A flow chart of a DC fault arc detection method according to a specific embodiment of the present disclosure is shown.

[0095] like Figure 2 As shown, the DC fault arc detection method comprises the following steps:

[0096] Signal acquisition: In this step, high-precision current and voltage sensors are used to collect current and voltage signals in the DC system in real time. The collected signal data is then divided into training samples and test samples to ensure the diversity and representativeness of the sample data.

[0097] Feature extraction: In this step, five types of feature quantities are extracted from the collected signal data, namely, the amplitude of the current signal, the current change rate, the singular value feature, the amplitude entropy and the energy mean.

[0098] Feature dimension reduction: The above five types of feature quantities are integrated to obtain a multidimensional mixed feature set, and the multidimensional mixed feature set is reduced in dimension using the adaptive LPP algorithm to select the optimal discriminant feature.

[0099] Fault identification: The low-dimensional feature data set after dimensionality reduction is input into the convolutional neural network for training and testing, and then the trained DC fault arc identification network is used to classify the new input data to determine whether a DC fault arc occurs.

[0100] Figure 3 The structure block diagram of the DC fault arc detection device according to the embodiment of the present disclosure is shown. The device can be implemented as part or all of the electronic device through software, hardware or a combination of both.

[0101] like Figure 3 As shown, the DC fault arc detection device 300 includes:

[0102] The acquisition module 310 is configured to acquire electrical signals of the DC power system; the electrical signals include normal signals and fault arc signals;

[0103] The construction module 320 is configured to extract the signal features of the electrical signal and construct a sample set based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features and energy features;

[0104] A training module 330 is configured to train a DC arc fault identification network based on the sample set;

[0105] The detection module 340 is configured to perform DC arc fault detection using the trained DC arc fault identification network.

[0106] The DC fault arc detection device provided by the present invention extracts a multi-dimensional mixed feature set, that is, extracts features of at least two dimensions among time-frequency domain features, singular value features, complexity features and energy features, trains a DC fault arc recognition network, and then uses the trained DC fault arc recognition network to perform fault judgment, thereby achieving efficient detection of DC fault arcs, improving detection accuracy and real-time performance, enhancing anti-interference capability, and being able to adapt to different DC distribution system scenarios.

[0107] In one implementation of the present disclosure, the signal features include time-frequency domain features, singular value features, complexity features, and energy features;

[0108] The building blocks include:

[0109] A construction unit, configured to construct a multi-dimensional hybrid feature set based on the signal features;

[0110] The dimensionality reduction processing unit is configured to perform dimensionality reduction processing on the multi-dimensional mixed feature set to obtain a low-dimensional feature data set as a sample set.

[0111] In one implementation of the present disclosure, the dimension reduction processing unit includes:

[0112] The dimensionality reduction processing subunit is configured to process the multi-dimensional mixed feature set using a local preservation projection algorithm to obtain a low-dimensional feature data set as a sample set.

[0113] In an implementation of the present disclosure, in the locality preserving projection algorithm, a kernel density estimation method is used to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multi-dimensional mixed feature set.

[0114] In an implementation of the present disclosure, in the locality preserving projection algorithm, using a kernel density estimation method to calculate neighborhood parameters of data sample points to implement dimensionality reduction processing of the multidimensional mixed feature set includes:

[0115] Calculate the neighborhood probability density as:

[0116]

[0117] The neighborhood parameters are calculated as:

[0118]

[0119] Among them, x i is the target point in the sample set for which the probability density is to be estimated, x d is the data sample point in the sample set; N is the number of samples in the sample set; φ(x) is the window function, and satisfies φ(x)≥0,∫φ(x)dx=0; h is the window width; V is the window volume; is the average probability density of the sample set; ε0 is the initial neighborhood parameter.

[0120] In one implementation of the present disclosure, the DC fault arc identification network is a convolutional neural network.

[0121] The present disclosure also discloses an electronic device, Figure 4 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0122] like Figure 4As shown, the electronic device includes a memory and a processor, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method according to an embodiment of the present disclosure.

[0123] The DC fault arc detection method comprises:

[0124] Collecting electrical signals of a DC power system; the electrical signals include normal signals and fault arc signals;

[0125] Extracting signal features of the electrical signal, and constructing a sample set based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features, and energy features;

[0126] Training a DC fault arc identification network based on the sample set;

[0127] The trained DC arc fault recognition network is used to detect DC arc faults.

[0128] In one implementation of the present disclosure, the signal features include time-frequency domain features, singular value features, complexity features, and energy features;

[0129] The constructing a sample set based on the signal feature comprises:

[0130] Constructing a multi-dimensional hybrid feature set based on the signal features;

[0131] The multi-dimensional mixed feature set is subjected to dimensionality reduction processing to obtain a low-dimensional feature data set as a sample set.

[0132] In an implementation of the present disclosure, performing dimensionality reduction processing on the multi-dimensional mixed feature set to obtain a low-dimensional feature data set as a sample set includes:

[0133] The multi-dimensional mixed feature set is processed using a locality preserving projection algorithm to obtain a low-dimensional feature data set as a sample set.

[0134] In an implementation of the present disclosure, in the locality preserving projection algorithm, a kernel density estimation method is used to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multi-dimensional mixed feature set.

[0135] In an implementation of the present disclosure, in the locality preserving projection algorithm, using a kernel density estimation method to calculate neighborhood parameters of data sample points to implement dimensionality reduction processing of the multidimensional mixed feature set includes:

[0136] Calculate the neighborhood probability density as:

[0137]

[0138] The neighborhood parameters are calculated as:

[0139]

[0140] Among them, x i is the target point in the sample set for which the probability density is to be estimated, x d is the data sample point in the sample set; N is the number of samples in the sample set; φ(x) is the window function, and satisfies φ(x)≥0,∫φ(x)dx=0; h is the window width; V is the window volume; is the average probability density of the sample set; ε0 is the initial neighborhood parameter.

[0141] In one implementation of the present disclosure, the DC fault arc identification network is a convolutional neural network.

[0142] Figure 5 A schematic diagram showing the structure of a computer system suitable for implementing the method according to an embodiment of the present disclosure is shown.

[0143] like Figure 5 As shown, the computer system includes a processing unit, which can perform the various methods in the above-mentioned embodiments according to the program stored in the read-only memory (ROM) or the program loaded from the storage part into the random access memory (RAM). In the RAM, various programs and data required for the operation of the computer system are also stored. The processing unit, ROM and RAM are connected to each other through a bus. The input / output (I / O) interface is also connected to the bus.

[0144] The following components are connected to the I / O interface: an input part including a keyboard, a mouse, etc.; an output part including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage part including a hard disk, etc.; and a communication part including a network interface card such as a LAN card, a modem, etc. The communication part performs a communication process via a network such as the Internet. The drive is also connected to the I / O interface as needed. Removable media, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on the drive as needed so that the computer program read therefrom is installed into the storage part as needed. Among them, the processing unit can be implemented as a processing unit such as a CPU, a GPU, a TPU, an FPGA, an NPU, etc.

[0145] In particular, according to an embodiment of the present disclosure, the method described above can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes a program code for executing the above method. In such an embodiment, the computer program can be downloaded and installed from a network through a communication part, and / or installed from a removable medium.

[0146] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0147] The units or modules involved in the embodiments described in the present disclosure may be implemented by software or programmable hardware. The units or modules described may also be set in a processor, and the names of these units or modules do not constitute limitations on the units or modules themselves in some cases.

[0148] As another aspect, the present disclosure further provides a chip, which includes at least one processor and can be used to implement the method involved in the above system embodiment.

[0149] In one possible design, the chip also includes a memory, which is used to store program instructions and data, and the memory is located inside or outside the processor.

[0150] As another aspect, the present disclosure further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the electronic device or computer system in the above embodiment; or a computer-readable storage medium that exists independently and is not assembled into a device. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the method described in the present disclosure.

[0151] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to).

Claims

1. A DC fault arc detection method, characterized in that: include: Collect electrical signals from DC power systems; The electrical signal includes a normal signal and a fault arc signal; Extracting signal features of the electrical signal, and constructing a sample set based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features, and energy features; Training a DC fault arc identification network based on the sample set; The trained DC arc fault recognition network is used to detect DC arc faults.

2. The DC fault arc detection method according to claim 1, characterized in that: The signal characteristics include time-frequency domain characteristics, Singular value features, complexity features, and energy features; The constructing a sample set based on the signal feature comprises: Constructing a multi-dimensional hybrid feature set based on the signal features; The multi-dimensional mixed feature set is subjected to dimensionality reduction processing to obtain a low-dimensional feature data set as a sample set.

3. The DC fault arc detection method according to claim 2, characterized in that: The dimensionality reduction process is performed on the multi-dimensional mixed feature set to obtain a low-dimensional feature data set as a sample set, including: The multi-dimensional mixed feature set is processed using a locality preserving projection algorithm to obtain a low-dimensional feature data set as a sample set.

4. The DC fault arc detection method according to claim 3, characterized in that: In the locality preserving projection algorithm, a kernel density estimation method is used to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multi-dimensional mixed feature set.

5. The DC fault arc detection method according to claim 4, characterized in that: In the locality preserving projection algorithm, the use of a kernel density estimation method to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multidimensional mixed feature set includes: Calculate the neighborhood probability density as: The neighborhood parameters are calculated as: Among them, x i is the target point in the sample set for which the probability density is to be estimated, x d is the data sample point in the sample set; N is the number of samples in the sample set; φ(x) is the window function, and satisfies φ(x)≥0,∫φ(x)dx=0; h is the window width; V is the window volume; is the average probability density of the sample set; ε0 is the initial neighborhood parameter.

6. The DC fault arc detection method according to claim 1, characterized in that: The DC fault arc identification network is a convolutional neural network.

7. A DC fault arc detection device, characterized in that: include: A collection module configured to collect electrical signals of a DC power system; The electrical signal includes a normal signal and a fault arc signal; A construction module is configured to extract signal features of the electrical signal and construct a sample set based on the signal features; the signal features include two or more of time-frequency domain features, singular value features, complexity features and energy features; A training module, configured to train a DC fault arc identification network based on the sample set; The detection module is configured to perform DC fault arc detection using a trained DC fault arc recognition network.

8. The DC fault arc detection device according to claim 7, characterized in that: The signal characteristics include time-frequency domain characteristics, Singular value features, complexity features, and energy features; The building blocks include: A construction unit, configured to construct a multi-dimensional hybrid feature set based on the signal features; The dimensionality reduction processing unit is configured to perform dimensionality reduction processing on the multi-dimensional mixed feature set to obtain a low-dimensional feature data set as a sample set.

9. The DC fault arc detection device according to claim 8, characterized in that: The dimension reduction processing unit comprises: The dimensionality reduction processing subunit is configured to process the multi-dimensional mixed feature set using a local preservation projection algorithm to obtain a low-dimensional feature data set as a sample set.

10. The DC fault arc detection device according to claim 9, characterized in that: In the locality preserving projection algorithm, a kernel density estimation method is used to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multi-dimensional mixed feature set.

11. The DC fault arc detection device according to claim 10, characterized in that: In the locality preserving projection algorithm, the use of a kernel density estimation method to calculate neighborhood parameters of data sample points to achieve dimensionality reduction processing of the multidimensional mixed feature set includes: Calculate the neighborhood probability density as: The neighborhood parameters are calculated as: Among them, x i is the target point in the sample set for which the probability density is to be estimated, x d is the data sample point in the sample set; N is the number of samples in the sample set; φ(x) is the window function, and satisfies φ(x)≥0,∫φ(x)dx=0; h is the window width; V is the window volume; is the average probability density of the sample set; ε0 is the initial neighborhood parameter.

12. The DC fault arc detection device according to claim 7, characterized in that: The DC fault arc identification network is a convolutional neural network.

13. An electronic device, characterized in that: The method comprises a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method according to any one of claims 1 to 6.

14. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

15. A chip, characterized in that: include: At least one processor, configured to implement the method according to any one of claims 1 to 6.

Images