Relay protection fault diagnosis method and system based on BNDT

Abstract

The invention provides a BNDT-based relay protection fault diagnosis method and system, and relates to the technical field of power system fault diagnosis, and the method comprises the steps: obtaining multi-source power parameters when a power system fails; carrying out normalization processing on the data; performing feature depth enhancement by using a deep neural network; the original data and the enhanced features are jointly input into a BNDT model to train and construct a fault prediction model; and performing fault diagnosis by using the model. Through fusion of the deep neural network and the BNDT, the problem of insufficient diagnosis precision caused by fuzzy fault data in a new energy power grid is effectively solved, and the accuracy and robustness of relay protection fault diagnosis are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system fault diagnosis, in particular to a relay protection fault diagnosis method and system based on BNDT. BACKGROUND

[0002] Relay protection is the first line of defense to ensure the safe and stable operation of the power system, and its core function is to quickly and accurately identify and isolate fault components when the system fails. With the large-scale integration of renewable energy such as wind power and photovoltaic power into the power grid, the topology and operation mode of the power grid are becoming increasingly complex, resulting in an explosive increase in operation and fault data and inevitably accompanied by problems such as data ambiguity and uncertainty. Traditional relay protection fault diagnosis methods rely mainly on expert systems or shallow machine learning models, and when dealing with such massive, high-dimensional, and ambiguous data, there are key technical problems such as insufficient diagnostic accuracy and poor fault tolerance, making it difficult to meet the higher requirements of new power systems for fault diagnosis accuracy and reliability. SUMMARY

[0003] To solve the technical problem of insufficient diagnostic accuracy caused by data ambiguity in the traditional relay protection fault diagnosis method in the background, the present application introduces a deep neural network (DNN) for feature enhancement and combines a Bayesian network decision tree (BNDT) to handle data ambiguity, and proposes a relay protection fault diagnosis scheme based on BNDT, which significantly improves the accuracy and robustness of fault diagnosis.

[0004] To achieve the above purpose, the first aspect of the present application provides a relay protection fault diagnosis method based on BNDT, comprising: obtaining power parameter data during power system failure; normalizing the power parameter data to obtain normalized data; inputting the normalized data into a deep neural network for feature deep enhancement processing to obtain enhanced feature data; inputting the normalized data and the enhanced feature data into a BNDT model for training to obtain a fault prediction model; using the fault prediction model to predict power system failure.

[0005] Further, the power parameters include oscillation period frequency deviation OFD, goose alarm information AG, SV alarm information AS, device service life AD, device service time ST, and whether a communication terminal CO appears.

[0006] Furthermore, the normalization process for the power parameter data is performed using the following formula: ; in, For the current eigenvalue, , These represent the minimum and maximum values ​​of the feature, respectively. The value is the normalized data value, and its range is (0, 1).

[0007] Furthermore, the normalized data is input into a deep neural network for feature depth enhancement processing, including: Use a deep neural network structure with an input layer, 20 hidden layers, and an output layer; The normalized data is linearly transformed using an activation function to obtain the output result. The calculation formula is: ; in, Represents the weight of each hidden layer, where, ; Indicates the first There are activation functions, where... ， ; The model number is represented by the first... The layer bias, where, .

[0008] Furthermore, the training of the deep neural network employs a learning rate decay mechanism, and the learning rate is calculated using the following formula: ; in, and These represent the minimum learning rate and the maximum learning rate, respectively. Indicates the number of training iterations. This represents the learning decay rate.

[0009] Furthermore, the step of inputting the normalized data and the enhanced feature data into the BNDT model for training includes: Multi-state Bayesian networks are used to process fuzzy data by adding Bayesian nodes between adjacent attribute nodes in traditional decision trees. Calculate the posterior probability using Bayes' theorem: ; in, ; The loss function is optimized using stochastic gradient descent to obtain the fault prediction model.

[0010] A second aspect of the present invention provides a BNDT-based relay protection fault diagnosis system, comprising: The data acquisition module is used to acquire power parameter data during power system faults. The normalization processing module is used to normalize the power parameter data to obtain normalized data; The feature enhancement module is used to input the normalized data into a deep neural network for feature depth enhancement processing to obtain enhanced feature data; The model training module is used to input the normalized data and the enhanced feature data into the BNDT model for training to obtain the fault prediction model. The fault prediction module is used to predict power system faults using the fault prediction model.

[0011] A third aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps in the BNDT-based relay protection fault diagnosis method as described in the first aspect of the present invention.

[0012] A fourth aspect of the present invention provides a computer-readable storage medium having a program stored thereon that, when executed by a processor, implements the steps of the BNDT-based relay protection fault diagnosis method as described in the first aspect of the present invention.

[0013] A fifth aspect of the present invention provides a computer program product comprising software code, wherein the program in the software code performs the steps of the BNDT-based relay protection fault diagnosis method as described in the first aspect of the present invention.

[0014] Compared with existing technologies, the relay protection fault diagnosis method and system based on BNDT provided by this invention have the following advantages: (1) To address the technical problem of decreased diagnostic accuracy caused by the ambiguity and high uncertainty of fault data due to the access of new energy sources, this invention utilizes a deep neural network to enhance the feature depth of normalized data and inputs the data into a BNDT model for training. By deeply mining the inherent correlation and high-order features of the data through the DNN, the discriminative power of the features is effectively improved. Furthermore, the Bayesian network nodes built into the BNDT model are used to perform probabilistic reasoning and uncertainty quantification on the ambiguity data, thereby realizing the effective utilization and identification of ambiguity information. Based on the synergy of the two, a significant improvement in fault diagnosis accuracy is achieved in complex data environments.

[0015] (2) In view of the technical problems of low training efficiency and easy overfitting of traditional models, this invention adopts a learning rate decay mechanism to train DNN. By dynamically adjusting the learning rate, the model can converge quickly in the early stage of training and stabilize near the optimal solution in the later stage of training. This effectively avoids the oscillation of the training process and suppresses the overfitting phenomenon, thereby improving the training efficiency and generalization ability of the model. Attached Figure Description

[0016] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0017] Figure 1 This is a flowchart of the relay protection fault diagnosis method based on BNDT provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the BNDT model structure provided in Embodiment 1 of the present invention; Figure 3 This is a diagram illustrating the architecture of a BNDT-based relay protection fault diagnosis system provided in Embodiment 2 of the present invention. Detailed Implementation

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form as well. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0021] All data acquisition in this embodiment is carried out in accordance with laws and regulations and with user consent, and the data is used legally.

[0022] Before describing the embodiments of the present invention in detail, the technical terms involved in the present invention are explained and defined as follows: 1. BNDT (Bayesian Network Decision Tree): This refers to a hybrid machine learning model that combines Bayesian networks and decision trees. Its core idea is to introduce Bayesian network nodes into the node splitting process of traditional decision trees to handle dependencies and uncertainties between attributes. By calculating posterior probabilities through Bayesian inference, BNDT can effectively handle fuzzy and incomplete data during training and prediction, improving the model's classification accuracy and robustness in complex scenarios.

[0023] 2. Feature Depth Enhancement: In this invention, it specifically refers to the process of automatically extracting high-level, abstract features from raw power parameter data using deep neural networks (DNNs). This process is achieved through multi-layer nonlinear transformations of DNNs, which can uncover deep, complex patterns and correlations in the data that are difficult to design manually, thereby generating an enhanced feature set that is more discriminative than the original features, providing a higher-quality data foundation for subsequent fault classification.

[0024] 3. Data Ambiguity: In this invention, this specifically refers to the uncertainty and inaccuracy present in relay protection fault diagnosis data. Its sources include, but are not limited to: signal fluctuations caused by the access of new energy sources, sensor measurement errors, packet loss or delay in communication transmission, and correlations and conflicts between various alarm information. This ambiguity makes it difficult for simple judgments based on Boolean logic or traditional classification models to make accurate diagnoses.

[0025] 4. Learning Rate Decay Mechanism: This refers to an optimization strategy that dynamically adjusts the learning rate during the training of deep neural networks. Its basic principle is to gradually decrease the learning rate as training iterations progress. A larger learning rate is used in the early stages of training to accelerate convergence; a smaller learning rate is used in the later stages to allow the model parameters to more stably approach the optimal solution, avoiding oscillations around the optimal value, thereby effectively improving training efficiency and the generalization performance of the final model.

[0026] Example 1 like Figure 1 This embodiment provides a relay protection fault diagnosis method based on BNDT, including: Acquire power parameter data during power system faults; The power parameter data is normalized to obtain normalized data; The normalized data is input into a deep neural network for feature depth enhancement processing to obtain enhanced feature data. The normalized data and the enhanced feature data are input into the BNDT model for training to obtain the fault prediction model. The fault prediction model is used to predict power system faults.

[0027] Specifically, the power parameters include oscillation frequency deviation OFD, goose alarm information AG, SV alarm information AS, equipment service life AD, equipment usage time ST, and whether a communication terminal CO has occurred.

[0028] By extensively collecting various and multi-dimensional parameters reflecting system fault states (such as oscillation frequency deviation (OFD), goose alarm information (AG), SV alarm information (AS), equipment service life (AD), equipment usage time (ST), and whether a communication terminal has occurred (CO), comprehensive and original information input is provided for subsequent diagnostic models. This overcomes the technical problem that single-type data often cannot comprehensively and accurately depict complex fault states, leading to insufficient diagnostic basis. By constructing a multi-source, heterogeneous dataset, the completeness and timeliness of the diagnostic foundation data are ensured, laying a solid data foundation for high-precision and robust fault diagnosis.

[0029] Specifically, the normalization process for the power parameter data is performed using the following formula: ; in, For the current eigenvalue, , These represent the minimum and maximum values ​​of the feature, respectively. The value is the normalized data value, and its range is (0, 1).

[0030] Normalizing each feature value to the (0,1) interval eliminates the influence of differences in the units and numerical ranges of different power parameters. This addresses the problem that if the original data features have inconsistent scales and are directly input into the model, features with large numerical ranges will dominate the training process, preventing the model from effectively learning other important features and affecting convergence speed and final performance. Unifying all features to the same scale accelerates the convergence process of the subsequent DNN model, improves the stability and efficiency of model training, and ensures a balanced contribution of all features to the model's decision-making.

[0031] Specifically, the normalized data is input into a deep neural network for feature depth enhancement processing, including: Use a deep neural network structure with an input layer, 20 hidden layers, and an output layer; The normalized data is linearly transformed using an activation function to obtain the output result. The calculation formula is: ; in, Represents the weight of each hidden layer, where, ; Indicates the first There are activation functions, where... ， ; The model number is represented by the first... The layer bias, where, .

[0032] A deep network structure with 20 hidden layers is constructed, and layer-by-layer nonlinear transformations are performed using activation functions (such as sigmoid). This process automatically learns and mines deep, complex, and nonlinear patterns and correlations in the original data. Automatic feature optimization and enhancement are achieved, generating higher-order abstract features that are more expressive and discriminative than the original features. This overcomes the limitations and subjectivity of manually designed features, providing an information foundation for high-precision diagnosis. Traditional feature extraction methods (such as manually designed or shallow models) struggle to effectively extract highly discriminative features from massive, high-dimensional fault data, limiting the upper limit of diagnostic model performance.

[0033] Specifically, the training of the deep neural network employs a learning rate decay mechanism, and the learning rate is calculated using the following formula: ; in, and These represent the minimum learning rate and the maximum learning rate, respectively. Indicates the number of training iterations. This represents the learning decay rate.

[0034] Use a larger learning rate in the early stages of training. To quickly approximate the optimal solution region; as the number of iterations increases... Increase the learning rate, then exponentially decay, using a smaller learning rate later in the training process. This allows for a fine-grained search near the optimal solution. It addresses the technical problems that a fixed learning rate can easily lead to oscillations in the convergence process, excessively long training times, or getting stuck in local optima or overfitting due to improper learning rate settings when training deep neural networks. This mechanism effectively balances training speed and convergence stability, avoiding drastic oscillations during training and enabling the model parameters to converge more smoothly and stably to the global optimum or a better solution. Simultaneously, it suppresses overfitting to some extent and improves the model's generalization ability.

[0035] Specifically, the step of inputting the normalized data and the enhanced feature data into the BNDT model for training includes: Multi-state Bayesian networks are used to process fuzzy data by adding Bayesian nodes between adjacent attribute nodes in traditional decision trees. Calculate the posterior probability using Bayes' theorem: ; in, ; The loss function is optimized using stochastic gradient descent to obtain the fault prediction model.

[0036] The BNDT model adds Bayesian nodes between adjacent attribute nodes in a traditional decision tree (DT), forming a hybrid structure. It utilizes decision trees for efficient preliminary classification and rule generation, while simultaneously employing Bayesian networks to handle fuzzy dependencies between attributes. Based on Bayes' theorem, it performs probabilistic inference, combining prior knowledge about faults (such as historical statistics) with current observational evidence (such as various alarm information) to calculate the most probable posterior probability of the fault state. This overcomes the significant fuzziness and uncertainty inherent in fault data in new energy power grid environments, where traditional "hard decision" models (such as ordinary decision trees and support vector machines) have poor fault tolerance and are prone to misdiagnosis and missed diagnosis. The BNDT model, combining the efficient classification capabilities of decision trees with the probabilistic inference advantages of Bayesian networks in handling uncertainty, enables the final fault prediction model to perform more intelligent and robust inference when faced with fuzzy, incomplete, or even contradictory fault data, thus significantly improving the accuracy and robustness of diagnosis.

[0037] In one specific embodiment, the method provided by the present invention specifically includes the following steps: Step 1: Data Acquisition.

[0038] Step 1.1: When a power system fault occurs, the power parameters will change significantly. The following power parameters are used as input data: oscillation frequency deviation (OFD), goose alarm information (AG), SV alarm information (AS), equipment service life (AD), equipment usage time (ST), and whether a communication terminal (CO) has appeared. Data from the past 3 years is collected as historical data, with a sampling time of less than 30 seconds.

[0039] Step 2 uses a DNN to perform feature depth enhancement on the input data; Step 2.1: First, the input data is dimensionless. This invention uses the following formula for normalization.

[0040]

[0041] in, For the current eigenvalue, , These represent the minimum and maximum values ​​of the feature, respectively. The data values ​​are normalized, ranging from (0, 1). The normalized data is represented as follows:

[0042] Step 2.2: Determine the structure of the DNN: input layer, 20 hidden layers, and output layer.

[0043] Step 2.3: Use a DNN to enhance the features of the data from Step 1.1. Pass the data through the input layer to hidden layer 1, where an activation function performs a linear transformation. Then pass the data to hidden layer 2, repeating this process until the output result is obtained. .

[0044]

[0045] In the formula, This represents the output of the DNN; Represents the weight of each hidden layer, where, ; Indicates the first There are activation functions, where... ， ; The model number is represented by the first... The layer bias, where, .

[0046] Step 2.4: To avoid overfitting in the DNN model and improve training efficiency, a learning rate decay mechanism is used to train the DNN. The calculation formula is as follows:

[0047] In the formula, and These represent the minimum learning rate and the maximum learning rate, respectively. This indicates the number of iterations for training the DNN model; This represents the learning decay rate.

[0048] Step 3: Use the original data and the output of the DNN as input to the BNDT model to establish a relay fault prediction model.

[0049] Step 3.1: BNDT uses a multi-state Bayesian network to process fuzzy data. Bayesian nodes are added between adjacent attribute nodes in traditional DT, and the fuzziness of the data is determined based on the consistency of prior and posterior information in the Bayesian network.

[0050] Step 3.1.1: The formula for BN in the BNDT model is as follows:

[0051]

[0052] The dataset is sequentially input into BNDT for training, and the loss function is optimized using stochastic gradient descent to obtain the fault prediction network.

[0053] Step 4: Use the BNDT model to predict the fault.

[0054] The beneficial effects of this invention are that, compared with the prior art, this invention uses DNN to enhance the features of the data, fully explores the relationship between the data, and improves the accuracy of the model; it uses BNDT to train the fault prediction model, which, compared with other networks, makes better use of fuzzy data, greatly improves the prediction accuracy of the model, greatly improves the data fuzziness problem in new power systems, and ensures the stable and reliable operation of the power system.

[0055] Example 2 like Figure 3 As shown, this embodiment provides a relay protection fault diagnosis system based on BNDT, including: The data acquisition module is used to acquire power parameter data during power system faults. The normalization processing module is used to normalize the power parameter data to obtain normalized data; The feature enhancement module is used to input the normalized data into a deep neural network for feature depth enhancement processing to obtain enhanced feature data; The model training module is used to input the normalized data and the enhanced feature data into the BNDT model for training to obtain the fault prediction model. The fault prediction module is used to predict power system faults using the fault prediction model.

[0056] Example 3 Embodiment 3 of the present invention provides an electronic device.

[0057] An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements the steps in the BNDT-based relay protection fault diagnosis method as described in Embodiment 1 of the present invention.

[0058] The detailed steps are the same as those of the BNDT-based relay protection fault diagnosis method provided in Example 1, and will not be repeated here.

[0059] Example 4 Embodiment 4 of the present invention provides a computer-readable storage medium.

[0060] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the BNDT-based relay protection fault diagnosis method as described in Embodiment 1 of the present invention.

[0061] The detailed steps are the same as those of the BNDT-based relay protection fault diagnosis method provided in Example 1, and will not be repeated here.

[0062] Example 5 Embodiment 5 of the present invention provides a computer program product.

[0063] A computer program product includes software code, wherein the program in the software code performs the steps of the BNDT-based relay protection fault diagnosis method as described in Embodiment 1 of the present invention.

[0064] The detailed steps are the same as those of the BNDT-based relay protection fault diagnosis method provided in Example 1, and will not be repeated here.

[0065] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can 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. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0066] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0067] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0069] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0070] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0071] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A relay protection fault diagnosis method based on BNDT, characterized in that, include: Acquire power parameter data during power system faults; The power parameter data is normalized to obtain normalized data; The normalized data is input into a deep neural network for feature depth enhancement processing to obtain enhanced feature data. The normalized data and the enhanced feature data are input into the BNDT model for training to obtain the fault prediction model. The fault prediction model is used to predict power system faults.

2. The method as described in claim 1, characterized in that, The power parameters include oscillation frequency deviation OFD, goose alarm information AG, SV alarm information AS, equipment service life AD, equipment usage time ST, and whether a communication terminal CO has occurred.

3. The method as described in claim 1, characterized in that, The power parameter data is normalized using the following formula: ； in, For the current eigenvalue, , These represent the minimum and maximum values ​​of the feature, respectively. The value is the normalized data value, and its range is (0, 1).

4. The method as described in claim 1, characterized in that, The normalized data is input into a deep neural network for feature depth enhancement processing, including: Use a deep neural network structure with an input layer, 20 hidden layers, and an output layer; The normalized data is linearly transformed using an activation function to obtain the output result. The calculation formula is: ； in, Represents the weight of each hidden layer, where, ; Indicates the first There are activation functions, where... , ; The model number is represented by the first... The layer bias, where, .

5. The method as described in claim 4, characterized in that, The deep neural network is trained using a learning rate decay mechanism, and the learning rate is calculated using the following formula: ； in, and These represent the minimum learning rate and the maximum learning rate, respectively. Indicates the number of training iterations. This represents the learning decay rate.

6. The method as described in claim 4, characterized in that, The step of inputting the normalized data and the enhanced feature data into the BNDT model for training includes: Multi-state Bayesian networks are used to process fuzzy data by adding Bayesian nodes between adjacent attribute nodes in traditional decision trees. Calculate the posterior probability using Bayes' theorem: ; in, ; The loss function is optimized using stochastic gradient descent to obtain the fault prediction model.

7. A relay protection fault diagnosis system based on BNDT, characterized in that, include: The data acquisition module is used to acquire power parameter data during power system faults. The normalization processing module is used to normalize the power parameter data to obtain normalized data; The feature enhancement module is used to input the normalized data into a deep neural network for feature depth enhancement processing to obtain enhanced feature data; The model training module is used to input the normalized data and the enhanced feature data into the BNDT model for training to obtain the fault prediction model. The fault prediction module is used to predict power system faults using the fault prediction model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps of the BNDT-based relay protection fault diagnosis method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the BNDT-based relay protection fault diagnosis method as described in any one of claims 1 to 6.

10. A computer program product, comprising software code, characterized in that, The program in the software code executes the steps of the BNDT-based relay protection fault diagnosis method as described in any one of claims 1 to 6.

Images