HVDC system fault diagnosis method and system based on ISTOA-CNN

Abstract

The application provides an ISTOA-CNN-based HVDC system fault diagnosis method and system, an improved ternarized Gull optimization algorithm is introduced to intelligently optimize parameters of the CNN model, and an optimized ISTOA-CNN model is obtained; the ISTOA-CNN model is trained through a training set, the ISTOA-CNN model is tested through a test set until a preset fault recognition accuracy is reached, and a final ISTOA-CNN-based HVDC system fault diagnosis model is obtained, which is used for fault diagnosis of a high-voltage direct-current transmission system, thereby solving the problem that parameter selection error of the CNN model has a great influence on classification effect. The improved ternarized Gull optimization algorithm is introduced to intelligently optimize parameters of the CNN, so as to improve fault recognition accuracy of the CNN on the HVDC system; the ISTOA is introduced to intelligently optimize parameters of the CNN, so as to reduce error existing in parameter selection of the CNN and affect classification accuracy of the CNN, and effectively improve accuracy of various fault diagnoses of the HVDC system.

Description

Technical Field

[0001] This application relates to the field of power system fault diagnosis technology, and in particular to a fault diagnosis method and system for HVDC systems based on ISTOA-CNN. Background Technology

[0002] As HVDC projects continue to increase their proportion in my country's power grid, ensuring the safe and stable operation of HVDC systems is becoming increasingly important. High-voltage direct current (HVDC) transmission protection systems are responsible for rapidly detecting faults and ensuring the safe and stable operation of HVDC transmission projects. HVDC technology boasts high transmission efficiency; therefore, improving the reliability, stability, and security of HVDC systems has significant socio-economic implications.

[0003] Researchers worldwide have developed numerous effective strategies for fault diagnosis in HVDC systems. Support vector machines (SVMs) can be used for nonlinear classification and regression, and have already been applied in some data modeling fields. In recent years, knowledge graph-based power system fault diagnosis methods have been continuously proposed, generating positive impacts in power system fault diagnosis. By combining database technology and data mining methods, connections and patterns between data can be discovered, thereby improving the performance of fault diagnosis. Key knowledge elements extracted from fault data collected through various channels can be effectively integrated through standardized representations and various relationships, continuously expanding the scope of domain knowledge and providing foundational elements for intelligent knowledge applications and visualization. Currently, the problems in the development of these technologies are low accuracy and inadequate utilization of hidden data information, which to some extent cannot guarantee the safe and stable operation of the power system. Summary of the Invention

[0004] This invention provides a fault diagnosis method and system for HVDC systems based on ISTOA-CNN. It introduces an improved Sooty Tern Optimization Algorithm (ISTOA) to intelligently optimize the parameters of the CNN, thereby addressing the problem that parameter selection errors in the CNN model significantly impact classification performance. This method achieves high accuracy in classifying various faults in HVDC systems, and compared to other methods, the false diagnosis rate is reduced to a certain extent.

[0005] In a first aspect, the fault diagnosis method for HVDC systems based on ISTOA-CNN described in this invention is used for diagnosing faults in high-voltage direct current transmission systems. The method includes:

[0006] S1: Extract signal features from the fault sample set using wavelet transform;

[0007] S2: Building a CNN model;

[0008] S3: An improved ternary optimization algorithm is introduced to intelligently optimize the parameters of the CNN model, resulting in an optimized ISTOA-CNN model;

[0009] S4: Label various fault types in the high-voltage direct current transmission system to obtain a fault sample set;

[0010] S5: Divide the fault sample set into a training set and a test set, train the ISTOA-CNN model using the training set, and test the ISTOA-CNN model using the test set until the preset fault recognition accuracy is achieved, thereby obtaining the final HVDC system fault diagnosis model based on ISTOA-CNN.

[0011] S6: The fault diagnosis model of HVDC system based on ISTOA-CNN is used to diagnose faults in high voltage direct current transmission systems.

[0012] This invention utilizes wavelet transforms to extract signal features from a fault sample set; it constructs a CNN model to classify HVDC fault datasets; simultaneously, to reduce errors in CNN parameter selection that could affect classification accuracy, it introduces ISTOA for intelligent parameter optimization; finally, the HVDC system fault sample set is divided into training and testing sets, and the ISTOA-CNN is trained to obtain the fault identification accuracy. The fault diagnosis method described in this invention achieves high accuracy in classifying various faults in HVDC systems, and compared to other methods, the false diagnosis rate is reduced to a certain extent, providing a new tool for HVDC system fault diagnosis.

[0013] In some implementations, step S1 utilizes wavelet transform to extract signal features from the fault sample set, specifically including:

[0014] Perform wavelet transform on the fault signal, let ;

[0015] If a and b change continuously, then a corresponding cluster Ψ(a,b)(t) is obtained;

[0016] The wavelet transform of the fault waveform signal x(t)∈L2(R) in a high-voltage direct current transmission system is defined as:

[0017] ;

[0018] Discretizing a and b, the discrete wavelet transform of the HVDC system fault waveform signal x(t) is calculated as follows:

[0019] ;

[0020] In the formula: a and b are discretization constants, and a > 0; t is the discretization time variable; Ψ(t) is the wavelet function; L2(R) refers to the function space to which the fault waveform signal x(t) belongs.

[0021] In some implementations, step S2, which involves building the CNN model, specifically includes:

[0022] The CNN model consists of convolutional layers, sampling layers, and fully connected layers; among which...

[0023] The output of the convolutional layer is: ;

[0024] The output of the sampling layer is: ;

[0025] The outputs of the convolutional and sampling layers are rearranged into X = {x1, x2, ..., xn} ∈ Rn and used as the input to the fully connected layer. The output of the fully connected layer is: ;

[0026] Y ∈ Rm×1, and is used as the input to the output layer of the CNN model, uses the Softmax regression function as the activation function to obtain the output result of the output layer as follows: ;

[0027] Wherein, the Softmax regression function is ;

[0028] In the above formula, Mj represents the selected output combination of the previous layer; klij represents the convolution kernel used between the i-th input and the j-th output; bj represents the bias corresponding to the j-th output; f is the activation function; down() represents the downsampling function; β represents the weight; W1∈Rm×n and b1∈Rm×1 are the weights and biases from the input layer to the hidden layer, respectively; W2∈Rk×n and b2∈Rk×1 are the weights and biases from the input layer to the hidden layer, respectively; Sj represents the probability that the classification category is the j-th type; and aj represents the j-th result of the fully connected layer output.

[0029] Step S3 introduces an improved tern optimization algorithm to intelligently optimize the parameters of the CNN model, specifically including:

[0030] S31: Initialize the variable parameters SA of the CNN model;

[0031] S32: Obtain the initial population based on the improved optimization algorithm for the Black-crowned Tern;

[0032] S33: Improve the variable parameter SA to obtain the improved objective function value;

[0033] S34: Update the location information of individual Black-tern terns through migration and attacks;

[0034] S35: Recalculate the fitness value of the objective function based on the location information of the individual, and retain the best individual; obtain the optimized ISTOA-CNN model based on the best individual.

[0035] In some implementations, step S5 yields the final HVDC system fault diagnosis model based on ISTOA-CNN. The specific implementation process includes:

[0036] S51: Train the ISTOA-CNN model using the training set, and output the globally optimal CNN model parameters of the improved tern optimization algorithm when the maximum number of iterations is reached;

[0037] S52: Establish an identification model using the globally optimal CNN model parameters, and identify the operating status of the high voltage DC transmission system through a test set until the preset fault identification accuracy is achieved, and output the optimal parameter results;

[0038] S53: Use the optimal parameter results as the parameters for the final HVDC system fault diagnosis model based on ISTOA-CNN;

[0039] The optimal parameter results include at least the Black-tailed Tern population, the number of iterations, and the settings for the CNN learning rate, dropout rate, hidden layer weights, and biases.

[0040] To address the issue that parameter selection errors in CNNs significantly impact classification performance, this invention introduces ISTOA for intelligent parameter optimization of CNNs. This method achieves high accuracy in classifying various faults in HVDC systems, and compared to other methods, the false diagnosis rate is reduced to a certain extent.

[0041] Secondly, the present invention also provides a diagnostic system for a fault diagnosis method for HVDC systems based on ISTOA-CNN, the diagnostic system comprising at least:

[0042] The data processing unit uses wavelet transform to extract signal features from the fault sample set, labels various fault types of the high voltage direct current transmission system, and divides the fault sample set into a training set and a test set.

[0043] The first optimization unit introduces an improved tern optimization algorithm to intelligently optimize the parameters of the CNN model, resulting in an optimized ISTOA-CNN model.

[0044] The second optimization unit trains the ISTOA-CNN model using the training set and tests the ISTOA-CNN model using the test set until a preset fault identification accuracy is achieved, thereby obtaining the final HVDC system fault diagnosis model based on ISTOA-CNN.

[0045] The fault diagnosis unit uses the aforementioned ISTOA-CNN-based HVDC system fault diagnosis model to perform fault diagnosis on the high-voltage direct current transmission system.

[0046] Thirdly, the present invention also provides a fault diagnosis model for fault diagnosis of high voltage direct current transmission systems; the fault diagnosis model is the ISTOA-CNN-based HVDC system fault diagnosis model in the ISTOA-CNN-based HVDC system fault diagnosis method described above.

[0047] Fourthly, the present invention also provides a fault diagnosis method for a high-voltage direct current transmission system, comprising the following steps:

[0048] Step 1: Collect data from multiple power equipment terminals and extract fault samples from the data;

[0049] Step 2: Input the fault samples into the HVDC system fault diagnosis model based on ISTOA-CNN, and calculate the fault type of each power equipment terminal through the HVDC system fault diagnosis model based on ISTOA-CNN;

[0050] Step 3: Save the fault samples to the data training set for optimization training of the HVDC system fault diagnosis model based on ISTOA-CNN;

[0051] The fault diagnosis model for HVDC systems based on ISTOA-CNN adopts the fault diagnosis model for HVDC systems based on ISTOA-CNN described above.

[0052] Fifthly, the present invention also provides a fault diagnosis device, the fault diagnosis device comprising at least:

[0053] A data acquisition module that communicates with multiple power equipment terminals in a high-voltage direct current transmission system;

[0054] A first processing unit connected to the data acquisition module is used to extract fault samples from the data information sent by the acquisition module, divide the fault samples into training set and test set, and store them in the storage unit.

[0055] The system also includes a data processing unit that calls upon data from the storage unit to perform real-time optimization of the data processing model and responds to diagnostic commands from the host computer to perform fault diagnosis on the power equipment terminal.

[0056] The data processing unit further includes:

[0057] When a diagnostic command is received from the host computer, real-time data information of the data acquisition module within a preset time period is obtained, and the real-time data information is input into the fault diagnosis model of the HVDC system based on ISTOA-CNN to calculate the fault type of each power equipment terminal; the fault diagnosis model of the HVDC system based on ISTOA-CNN adopts the fault diagnosis model of the HVDC system based on ISTOA-CNN in the fault diagnosis method of the HVDC system based on ISTOA-CNN as described above.

[0058] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0059] 1. This invention introduces the Improved Sooty Tern Optimization Algorithm (ISTOA) to intelligently optimize the parameters of CNN, thereby improving the accuracy of CNN in identifying faults in HVDC systems.

[0060] 2. This invention introduces ISTOA to intelligently optimize the parameters of CNN, thereby reducing the problem of errors in CNN parameter selection that affect the classification accuracy of CNN.

[0061] 3. In this invention, the fault sample set of the HVDC system is divided into a training set and a test set, and the ISTOA-CNN is trained to obtain the fault identification accuracy. This method has a high accuracy in classifying various faults in the HVDC system, and the false diagnosis rate is reduced to a certain extent compared with other methods. Attached Figure Description

[0062] Figure 1 This is a flowchart of the HVDC system fault diagnosis method based on ISTOA-CNN described in this invention.

[0063] Figure 2 for Figure 1 The flowchart of fault diagnosis for the HVDC system based on ISTOA-CNN.

[0064] Figure 3 This is a schematic diagram of the CNN model described in this invention.

[0065] Figure 4 This is a schematic diagram illustrating the intelligent parameter optimization of the CNN model using the improved tern optimization algorithm described in this invention.

[0066] Figure 5 This is a schematic diagram of the HVDC system fault diagnosis system based on ISTOA-CNN described in this invention.

[0067] Figure 6 This is a flowchart of the fault diagnosis method for high voltage direct current transmission systems described in this invention.

[0068] Figure 7 This is a schematic diagram of the fault diagnosis device described in this invention. Detailed Implementation

[0069] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0070] Example 1:

[0071] Please refer to Figure 1-2 The diagram below is a schematic of a fault diagnosis method for an HVDC system based on ISTOA-CNN, provided in the first embodiment of this application.

[0072] Specifically, the HVDC system fault diagnosis method based on ISTOA-CNN described in this invention is used to diagnose faults in high-voltage direct current transmission systems. The method includes:

[0073] S1: Extract signal features from fault sample set using wavelet transform.

[0074] In some implementations, step S1 utilizes wavelet transform to extract signal features from the fault sample set, specifically including:

[0075] Perform wavelet transform on the fault signal, let .

[0076] If a and b change continuously, then a corresponding cluster Ψ(a,b)(t) is obtained.

[0077] The wavelet transform of the fault waveform signal x(t)∈L2(R) in a high-voltage direct current transmission system is defined as:

[0078] .

[0079] Discretize a and b: Select a set of discrete parameters a and b, which control the expansion and translation of the wavelet function.

[0080] The discrete wavelet transform of the fault waveform signal x(t) of the HVDC system is calculated as follows:

[0081] .

[0082] In the formula: a and b are discretization constants, and a > 0; t is the discretization time variable; Ψ(t) is the wavelet function; L2(R) refers to the function space to which the fault waveform signal x(t) belongs.

[0083] By extracting signal features from a fault sample set using wavelet transform, the characteristics of the fault signals can be analyzed and described from both frequency and time domain perspectives. These features can be used for fault diagnosis, classification, and analysis, assisting in fault handling and maintenance of HVDC systems. It should be noted that the specific feature extraction and analysis methods can be selected and implemented based on actual needs and problems.

[0084] S2: Building a CNN model;

[0085] In some implementations, step S2, which involves building the CNN model, specifically includes:

[0086] Please refer to Figure 3 The CNN model consists of convolutional layers, sampling layers, and fully connected layers; wherein,

[0087] The output of the convolutional layer is: ;

[0088] The output of the sampling layer is: ;

[0089] The outputs of the convolutional and sampling layers are rearranged into X = {x1, x2, ..., xn} ∈ Rn and used as the input to the fully connected layer. The output of the fully connected layer is: ;

[0090] Y ∈ Rm×1, and is used as the input to the output layer of the CNN model, uses the Softmax regression function as the activation function to obtain the output result of the output layer as follows: ;

[0091] Wherein, the Softmax regression function is ;

[0092] In the above formula, Mj represents the selected output combination of the previous layer; klij represents the convolution kernel used between the i-th input and the j-th output; bj represents the bias corresponding to the j-th output; f is the activation function; down() represents the downsampling function; β represents the weight; W1∈Rm×n and b1∈Rm×1 are the weights and biases from the input layer to the hidden layer, respectively; W2∈Rk×n and b2∈Rk×1 are the weights and biases from the input layer to the hidden layer, respectively; Sj represents the probability that the classification category is the j-th type; and aj represents the j-th result of the fully connected layer output.

[0093] S3: An improved ternary optimization algorithm is introduced to intelligently optimize the parameters of the CNN model, resulting in an optimized ISTOA-CNN model;

[0094] In some implementations, step S3 introduces an improved tern optimization algorithm to intelligently optimize the parameters of the CNN model, such as... Figure 4 As shown, it specifically includes:

[0095] S31: Initialize the variable parameters SA of the CNN model;

[0096] The variable parameter SA can also include the convolution kernel size, pooling operation, number of layers in the neural network, and number of nodes.

[0097] S32: Obtain the initial population based on the improved optimization algorithm for the Black-crowned Tern;

[0098] The improved Black-tern optimization algorithm is an optimization algorithm based on the strategies of Black-terns in nature. It simulates the behavior of flocks searching for food and finds the optimal solution through search and update operations. In this step, an initial population is generated according to the algorithm's design.

[0099] S33: Improve the variable parameter SA to obtain the improved objective function value;

[0100] In this step, the parameters of the CNN model are updated and changed according to the fitness function or objective function defined in the algorithm, and the corresponding objective function value is calculated.

[0101] S34: Update the location information of individual Black-tern terns through migration and attacks;

[0102] Among them, the improved tern optimization algorithm updates the location information of individuals by simulating the behavior of terns, including migration and attack, in order to achieve the optimization goal.

[0103] S35: Recalculate the fitness value of the objective function based on the location information of the individual, and retain the best individual; obtain the optimized ISTOA-CNN model based on the best individual.

[0104] In this step, the fitness value of the objective function is recalculated based on the iteratively updated individual location information, and the optimal individual is retained. Finally, the optimized ISTOA-CNN model is obtained based on the parameter values ​​of the optimal individual.

[0105] To address the issue that parameter selection errors in CNNs significantly impact classification performance, this invention introduces ISTOA for intelligent parameter optimization of CNNs. This method achieves high accuracy in classifying various faults in HVDC systems, and compared to other methods, the false diagnosis rate is reduced to a certain extent.

[0106] S4: Label various fault types in the high-voltage direct current transmission system to obtain a fault sample set;

[0107] To address the impact of CNN parameter selection errors on classification performance, ISTOA is introduced to intelligently optimize CNN parameters, thereby improving classification accuracy and reducing false positive rates. By improving the Black Terrier Optimization Algorithm, the parameter space of the CNN model can be efficiently searched, and the optimal individual parameters can be found to optimize the ISTOA-CNN model.

[0108] In some implementations, HVDC system fault types are labeled and mapped, as shown in Table 1: Label Mapping Table.

[0109]

[0110] S5: Divide the fault sample set into a training set and a test set, train the ISTOA-CNN model using the training set, and test the ISTOA-CNN model using the test set until the preset fault recognition accuracy is achieved, thereby obtaining the final HVDC system fault diagnosis model based on ISTOA-CNN.

[0111] In some implementations, step S5 yields the final HVDC system fault diagnosis model based on ISTOA-CNN. The specific implementation process includes:

[0112] S51: Train the ISTOA-CNN model using the training set, and output the globally optimal CNN model parameters of the improved tern optimization algorithm when the maximum number of iterations is reached;

[0113] In this step, the training set is used as input to train the ISTOA-CNN model, and the value of the loss function is recorded in each iteration. After reaching the predetermined maximum number of iterations, the parameters of the globally optimal CNN model obtained by the improved ternary optimization algorithm are output.

[0114] S52: Establish an identification model using the globally optimal CNN model parameters, and identify the operating status of the high voltage DC transmission system through a test set until the preset fault identification accuracy is achieved, and output the optimal parameter results;

[0115] In this step, a recognition model is built based on the parameters of the globally optimal CNN model, namely, an HVDC system fault diagnosis model based on ISTOA-CNN. Then, the model is validated and evaluated using a test set. According to the preset fault recognition accuracy requirements, iterative training and parameter adjustment are performed until the preset accuracy requirements are met, and the optimal parameter results are output.

[0116] S53: Use the optimal parameter results as the parameters for the final HVDC system fault diagnosis model based on ISTOA-CNN;

[0117] In this step, the obtained optimal parameter results are used as parameters for the ISTOA-CNN model to form the final HVDC system fault diagnosis model.

[0118] The optimal parameter results include at least the Black-tailed Tern population, the number of iterations, and the settings for the CNN learning rate, dropout rate, hidden layer weights, and biases. These parameters have a significant impact on the model's training and performance. By improving the optimal parameter results obtained through the Black-tailed Tern optimization algorithm, the accuracy and reliability of the model can be improved, thereby enhancing the effectiveness of fault diagnosis in the HVDC system.

[0119] S6: The fault diagnosis model of HVDC system based on ISTOA-CNN is used to diagnose faults in high voltage direct current transmission systems.

[0120] In some implementations, this invention utilizes wavelet transforms to extract signal features from the fault sample set; it constructs a CNN model to classify the HVDC fault dataset; simultaneously, to reduce errors in CNN parameter selection that could affect classification accuracy, it introduces ISTOA for intelligent parameter optimization of the CNN; finally, the HVDC system fault sample set is divided into training and testing sets, and the ISTOA-CNN is trained to obtain the fault identification accuracy. The fault diagnosis method described in this invention achieves high accuracy in classifying various faults in the HVDC system, and compared to other methods, the false diagnosis rate is reduced to a certain extent, providing a new tool for HVDC system fault diagnosis.

[0121] Furthermore, to better illustrate step S3: introducing an improved tern optimization algorithm to intelligently optimize the parameters of the CNN model and obtain the optimized ISTOA-CNN model, the preferred implementation process is as follows:

[0122] Step 1: Migration behavior (global search).

[0123] The algorithm achieves global search by simulating the migration process of a flock of black terns, and three conditions need to be met during the migration phase.

[0124] (1) Avoid conflict.

[0125] To prevent collisions between flocks of terns, the algorithm uses an additional variable, SA, to calculate the new position of the terns. The formula is as follows:

[0126] ;

[0127] In the formula: Cst is the new migratory location of the tern that does not conflict with other terns; Pst(Z) is the current location of the tern; SA is the movement mode of the tern in the given space.

[0128] ;

[0129] In the formula: Cf can adjust the control variable SA, whose value decreases linearly from 2 to 0; Z is the current iteration number; Miteration is the maximum iteration number.

[0130] (2) Gathering.

[0131] After avoiding collisions with other terns during movement, individual terns will move towards the direction of the optimal position, as shown by the following formula:

[0132] ;

[0133] In the formula: mst is the direction of the optimal position; pbst(Z) is the optimal position of the tern; CB is a random number to make the exploration more comprehensive.

[0134] ;

[0135] In the formula: Rrand is a random number in the range [0,1].

[0136] (3) Update.

[0137] The trajectory of the Black-crowned Tern moving towards its optimal position is given by the formula:

[0138] ;

[0139] Step 2: Attack behavior (local search).

[0140] During migration, the Black-winged Tern relies on its wings and weight to maintain altitude and can continuously adjust its attack angle and speed. Its circling behavior in the air when attacking prey can be defined by the following mathematical model.

[0141] ;

[0142] ;

[0143] ;

[0144] ;

[0145] In the formula: R is the radius of each helix; i is a random angle in the range [0, 2π]; u and v are constants related to the shape of the helix.

[0146] Based on the attack positions of the terns mentioned above, we can conclude that:

[0147] ;

[0148] In the formula: Ps(Z) is the attack position of the black tern.

[0149] Step 3: Improve the Black-and-Tern algorithm.

[0150] In the optimization process, the variable SA acts as an adaptive inertia weight. A larger weight provides good global search capability and can accelerate convergence, but it is difficult to find the optimal solution. To enhance local search capability, the weight value of variable SA is appropriately reduced as the number of iterations increases, as shown in the following equation:

[0151] ;

[0152] ;

[0153] Example 2:

[0154] This invention also provides a diagnostic system for HVDC system fault diagnosis based on ISTOA-CNN, such as... Figure 5 As shown, the diagnostic system includes at least:

[0155] The data processing unit uses wavelet transform to extract signal features from the fault sample set, labels various fault types of the high voltage direct current transmission system, and divides the fault sample set into a training set and a test set.

[0156] The first optimization unit introduces an improved tern optimization algorithm to intelligently optimize the parameters of the CNN model, resulting in an optimized ISTOA-CNN model.

[0157] The second optimization unit trains the ISTOA-CNN model using the training set and tests the ISTOA-CNN model using the test set until a preset fault identification accuracy is achieved, thereby obtaining the final HVDC system fault diagnosis model based on ISTOA-CNN.

[0158] The fault diagnosis unit uses the aforementioned ISTOA-CNN-based HVDC system fault diagnosis model to perform fault diagnosis on the high-voltage direct current transmission system.

[0159] The diagnostic system based on ISTOA-CNN for HVDC system fault diagnosis, as described in this invention, is used to classify various faults in HVDC systems. Addressing the issue that parameter selection errors in Convolutional Neural Networks (CNNs) significantly impact classification performance, an Improved Sooty Tern Optimization Algorithm (ISTOA) is introduced to intelligently optimize CNN parameters, thereby improving the accuracy of CNN in identifying faults in HVDC systems.

[0160] This method significantly improves the accuracy of fault diagnosis in HVDC systems. First, wavelet transforms are used to extract signal features from the fault sample set. Then, a CNN is built to classify the HVDC fault dataset. Simultaneously, to reduce errors in CNN parameter selection that could affect classification accuracy, ISTOA is introduced for intelligent parameter optimization. Finally, the HVDC system fault sample set is divided into training and test sets, and the ISTOA-CNN is trained to obtain the fault identification accuracy. This method achieves high accuracy in classifying various faults in HVDC systems, and compared to other methods, the false diagnosis rate is reduced to a certain extent.

[0161] Example 3:

[0162] The present invention also provides a fault diagnosis model for fault diagnosis of high voltage direct current transmission systems; the fault diagnosis model is the ISTOA-CNN-based HVDC system fault diagnosis model in the ISTOA-CNN-based HVDC system fault diagnosis method described above.

[0163] A fault diagnosis model based on ISTOA-CNN, after training and optimization, can accurately diagnose faults in high-voltage direct current (HVDC) transmission systems. It utilizes the ISTOA-CNN model and employs the Black-Tern optimization algorithm to intelligently optimize its parameters for better performance and accuracy. The model uses wavelet transform to extract signal features from the fault sample set, which are then trained and tested together with training and test sets until a preset fault identification accuracy is achieved. By using this fault diagnosis model, faults in HVDC transmission systems can be effectively and accurately diagnosed, improving the system's reliability and stability.

[0164] Example 4:

[0165] This invention also provides a fault diagnosis method for high-voltage direct current transmission systems, such as... Figure 6 As shown, it includes the following steps:

[0166] Step 1: Collect data from multiple power equipment terminals and extract fault samples from the data;

[0167] During data acquisition, data from power equipment terminals can be collected using sensors, monitoring devices, or data instruments. This data can include various parameters such as voltage, current, power, frequency, temperature, and pressure. The collected data is then recorded and stored. Data recording systems, databases, or cloud platforms can be used to save the data, ensuring its integrity and reliability.

[0168] Based on known fault conditions or expert knowledge, the collected data is labeled to determine which data samples belong to fault samples. Specialized labeling tools or software can be used for labeling.

[0169] Preferably, during feature extraction, signal processing techniques, statistical analysis methods, or machine learning algorithms can be used to extract fault features from data samples. Commonly used methods include wavelet transform, spectrum analysis, time-frequency analysis, and time-domain feature extraction.

[0170] Preferably, the fault sample set is divided into a training set and a test set. Typically, most samples are used to train the model, while a smaller portion is used to test the model's accuracy and generalization ability.

[0171] By collecting data from multiple power equipment terminals and extracting fault samples, a rich sample set of fault data is established for training and testing fault diagnosis models. This effectively helps identify and predict various fault conditions, improving the safety and reliability of power systems.

[0172] Step 2: Input the fault samples into the HVDC system fault diagnosis model based on ISTOA-CNN, and calculate the fault type of each power equipment terminal through the HVDC system fault diagnosis model based on ISTOA-CNN;

[0173] In some implementations, for each fault sample of a power equipment terminal, preprocessing and format conversion are performed according to model requirements to conform to the model's input requirements. This may include data normalization, standardization, or other specific processing operations. The preprocessed fault samples are then input into an ISTOA-CNN-based HVDC system fault diagnosis model. The model uses pre-trained weights and parameters to calculate and process the input data.

[0174] The model calculates based on the input fault samples and outputs the corresponding fault type results. These fault types may include various types such as overload, short circuit, and ground fault.

[0175] The fault type of each power equipment terminal can be interpreted based on the model's output. The model's output can be interpreted according to predefined labels or codes.

[0176] By using an HVDC system fault diagnosis model based on ISTOA-CNN, the fault type of each power equipment terminal can be calculated and determined. This provides timely fault diagnosis, helping operation and maintenance personnel to quickly locate and resolve problems, ensuring the reliable operation of the power system.

[0177] Step 3: Save the fault samples to the data training set for optimization training of the HVDC system fault diagnosis model based on ISTOA-CNN;

[0178] In some implementations, extracted fault samples are saved to a training set. To ensure the integrity and accuracy of the sample data, appropriate data formats and storage methods can be used, such as CSV files, databases, or specialized dataset management tools. Necessary data preprocessing is performed on the saved fault samples. This includes data cleaning, feature selection, and sample balancing to ensure the quality and usability of the training set. The saved fault samples are then used to train an HVDC system fault diagnosis model based on ISTOA-CNN. The model is trained by inputting the training set into the model and performing backpropagation, parameter optimization, and other steps. During training, the model is adjusted and optimized based on its performance and metrics. Different combinations of hyperparameters, loss functions, and optimization algorithms can be tried to achieve better performance and accuracy. The optimized and trained model is then validated and tested using different evaluation metrics. Independent test sets or cross-validation methods can be used to evaluate the model's performance and ensure its accuracy and generalization ability on unseen data.

[0179] By storing fault samples in a training dataset and using these samples for model optimization training, the performance and accuracy of the ISTOA-CNN-based HVDC system fault diagnosis model can be improved. This training process allows the model to better identify and classify different types of faults, thereby enhancing the accuracy and reliability of fault diagnosis.

[0180] The fault diagnosis model for HVDC systems based on ISTOA-CNN adopts the fault diagnosis model for HVDC systems based on ISTOA-CNN described above.

[0181] Example 5:

[0182] The present invention also provides a fault diagnosis device, such as... Figure 7 As shown, the fault diagnosis device includes at least:

[0183] A data acquisition module that communicates with multiple power equipment terminals in a high-voltage direct current transmission system;

[0184] A first processing unit connected to the data acquisition module is used to extract fault samples from the data information sent by the acquisition module, divide the fault samples into training set and test set, and store them in the storage unit.

[0185] The system also includes a data processing unit that calls upon data from the storage unit to perform real-time optimization of the data processing model and responds to diagnostic commands from the host computer to perform fault diagnosis on the power equipment terminal.

[0186] The data processing unit further includes:

[0187] When a diagnostic command is received from the host computer, real-time data information of the data acquisition module within a preset time period is obtained, and the real-time data information is input into the fault diagnosis model of the HVDC system based on ISTOA-CNN to calculate the fault type of each power equipment terminal; the fault diagnosis model of the HVDC system based on ISTOA-CNN adopts the fault diagnosis model of the HVDC system based on ISTOA-CNN in the fault diagnosis method of the HVDC system based on ISTOA-CNN as described above.

[0188] In some implementations, the fault diagnosis equipment operates as follows:

[0189] Step 1: Acquire Real-Time Data: Following instructions, acquire the required real-time data from the power equipment terminal via the data acquisition module. This data may include parameters such as voltage, current, power, frequency, temperature, and pressure, and the time range is within a preset time period.

[0190] Step 2: Data Preprocessing: Perform necessary preprocessing and format conversion on the acquired real-time data to meet the input requirements of the ISTOA-CNN-based HVDC system fault diagnosis model. This may include data normalization, standardization, or other specific processing operations.

[0191] Step 3: Input Data into the Model: Input the preprocessed real-time data into the ISTOA-CNN-based HVDC system fault diagnosis model. The model will use the pre-trained weights and parameters to calculate and process the input data.

[0192] Step 4: Calculate the fault type: The HVDC system fault diagnosis model based on ISTOA-CNN will calculate the fault type of each power equipment terminal based on the input real-time data. These fault types may include overload, short circuit, ground fault, and other types.

[0193] Step 5: Calculate the fault type: The HVDC system fault diagnosis model based on ISTOA-CNN will calculate the fault type of each power equipment terminal based on the input real-time data. These fault types may include overload, short circuit, ground fault, and other types.

[0194] The ISTOA-CNN-based fault diagnosis model for HVDC systems employs the ISTOA-CNN-based model from the aforementioned ISTOA-CNN-based HVDC system fault diagnosis method. By calculating real-time data, it can accurately determine the fault type of each power equipment terminal. This helps operators to promptly detect and handle faults, ensuring the safe and reliable operation of the power system.

[0195] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of the invention. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the invention as claimed in the appended claims.

[0196] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0197] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0198] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules according to embodiments of the present invention. The present invention can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing some or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0199] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0200] Although the invention has been described in conjunction with the specific embodiments described above, it will be apparent to those skilled in the art that many substitutions, modifications, and variations can be made based on the foregoing. Therefore, all such substitutions, modifications, and variations are included within the spirit and scope of the appended claims.

Claims

1. An ISTOA-CNN-based HVDC system fault diagnosis method for diagnosing faults of a high-voltage direct current (HVDC) power transmission system, characterized in that, The method comprises: S1: extracting signal features of a fault sample set by wavelet variation; S2: building a CNN model; S3: introducing an improved ternary optimization algorithm of artificial swallow optimization to intelligently optimize parameters of the CNN model to obtain an optimized ISTOA-CNN model; S4: labeling various fault types of a high-voltage direct-current (HVDC) power transmission system to obtain a fault sample set; S5: dividing the fault sample set into a training set and a test set, training the ISTOA-CNN model by using the training set, testing the ISTOA-CNN model by using the test set until a preset fault recognition accuracy is reached, and obtaining a final HVDC system fault diagnosis model based on ISTOA-CNN; S6: using the HVDC system fault diagnosis model based on ISTOA-CNN for fault diagnosis of the HVDC power transmission system. The step S3 further comprises: S31: initialize variable parameters of the CNN model S A ; S32: obtaining an initial population according to the improved ternary optimization algorithm of artificial swallow optimization; S33: updating the variable parameter according to the target function value S A improvement, obtaining an improved target function value; S34: updating position information of the artificial swallow individuals by migration and attack; S35: recalculating fitness values of the objective function according to the position information of the individuals, retaining an optimal individual, and obtaining the optimized ISTOA-CNN model according to the optimal individual.

2. The ISTOA-CNN based HVDC system fault diagnosis method of claim 1, wherein, The step S1 further comprises: Wavelet transform is performed on the fault signal, and let ; If a, b vary, a cluster of solutions is obtained corresponding thereto Ψ (a,b) ( t ) Defining fault waveform signals of a high voltage direct current transmission system x ( t )∈ L 2 ( R ) wavelet transform is: ； Discretizing a, b, calculating the HVDC system fault waveform signal x ( t ) of the discrete wavelet variation is: ； where a and b are discretized constants, and a > 0; t is a discretized time variable; Ψ ( t ) is a wavelet function; L 2 ( R ) refers to the function space to which the fault waveform signal x(t) belongs.

3. The ISTOA-CNN based HVDC system fault diagnosis method of claim 2, wherein, The step S2 further comprises that the CNN model is composed of a convolution layer, a sampling layer, and a fully connected layer; wherein, The output of the convolution layer is: ; The output of the sampling layer is: ; The output of the convolutional and pooling layers is arranged into X ={ x 1, x 2,..., x n}∈R n As input to the fully connected layer, the output of the fully connected layer is: ; The Y ∈R m×1 , and as the input of the output layer of the CNN model, a Softmax regression function is used as the activation function, and the output result of the output layer is: ; In the formula, the Softmax regression function is ; In the above formula, M j represents the selected last layer output combination: k l ij represents the convolution kernel used between the i th input and the j th output; b j represents the bias corresponding to the j th output; f is an activation function; down () represents a down-sampling function, β represents a weight; W 1∈R m×n , b 1∈R m×1 are the weights and biases from the input layer to the hidden layer, respectively; W 2∈R k×n , b 2∈R k×1 are the weights and biases from the input layer to the hidden layer, respectively; S j represents the probability that the classification category is the j th; a j represents the j th result of the fully connected layer output.

4. The ISTOA-CNN based HVDC system fault diagnosis method of claim 3, wherein, The step S5 further comprises: S51: training the ISTOA-CNN model by using the training set, outputting global optimal CNN model parameters of the improved ternary optimization algorithm of artificial swallow optimization when a maximum number of iterations is reached; S52: establishing an identification model by using the global optimal CNN model parameters, and identifying an operating state of the HVDC power transmission system by using the test set until a preset fault recognition accuracy is reached, and outputting optimal parameter results; S53: using the optimal parameter results as parameters of the final HVDC system fault diagnosis model based on ISTOA-CNN; The optimal parameter results at least include a swallow population, a number of iterations, a CNN learning rate, a dropout rate, hidden layer weights, and biases.

5. The diagnostic system of the ISTOA-CNN based HVDC system fault diagnosis method according to any one of claims 1-4, characterized in that, The diagnosis system at least comprises: a data processing unit that extracts signal features of a fault sample set by wavelet variation, labels various fault types of a high-voltage direct-current (HVDC) power transmission system, and divides the fault sample set into a training set and a test set; a first optimization unit that introduces an improved ternary optimization algorithm of artificial swallow optimization to intelligently optimize parameters of a CNN model to obtain an optimized ISTOA-CNN model; a second optimization unit that trains the ISTOA-CNN model by using the training set, tests the ISTOA-CNN model by using the test set until a preset fault recognition accuracy is reached, and obtains a final HVDC system fault diagnosis model based on ISTOA-CNN; a fault diagnosis unit that uses the HVDC system fault diagnosis model based on ISTOA-CNN for fault diagnosis of the HVDC power transmission system.

6. A failure diagnostic model characterized by, The fault diagnosis model is used for fault diagnosis of the high-voltage direct current power transmission system; and the fault diagnosis model is the ISTOA-CNN-based HVDC system fault diagnosis model in the ISTOA-CNN-based HVDC system fault diagnosis method according to any one of claims 1-5.

7. A method of diagnosing a fault in a high voltage direct current power transmission system, characterized by, The method comprises the following steps: Step 1: collecting data information of a plurality of power equipment terminals, and extracting fault samples in the data information; Step 2: inputting the fault samples into the ISTOA-CNN-based HVDC system fault diagnosis model, and calculating fault types of each of the power equipment terminals by the ISTOA-CNN-based HVDC system fault diagnosis model; Step 3: saving the fault samples to a data training set for optimized training of the ISTOA-CNN-based HVDC system fault diagnosis model; The ISTOA-CNN-based HVDC system fault diagnosis model adopts the ISTOA-CNN-based HVDC system fault diagnosis model in the ISTOA-CNN-based HVDC system fault diagnosis method according to any one of claims 1-5.

8. A failure diagnosing apparatus characterized by comprising: The fault diagnosis device at least comprises: a data acquisition module in communication connection with a plurality of power equipment terminals in the high-voltage direct current power transmission system; a first processing unit connected with the data acquisition module, the first processing unit being used for extracting fault samples in data information sent by the acquisition module, dividing the fault samples into a training set and a test set, and storing the fault samples to a storage unit; and a data processing unit, which calls data in the storage unit to complete real-time optimization of a data processing model, and responds to a diagnosis instruction sent by a host computer to perform fault diagnosis on the power equipment terminals; When receiving the diagnosis instruction sent by the host computer, real-time data information in a preset time period at the data acquisition module end is acquired, and the real-time data information is input into the ISTOA-CNN-based HVDC system fault diagnosis model to calculate fault types of each of the power equipment terminals; the ISTOA-CNN-based HVDC system fault diagnosis model adopts the ISTOA-CNN-based HVDC system fault diagnosis model in the ISTOA-CNN-based HVDC system fault diagnosis method according to any one of claims 1-5.

Images