A method and system for identifying faults on a dc transmission line taking into account control response of converter stations

Abstract

This invention provides a method and system for fault identification of DC transmission lines considering converter station control response. The method includes: collecting and preprocessing fault waveforms of DC voltage and current on the rectifier side under fault conditions, as well as seasonal and weather information at the time of the fault; extracting the inherent fault components of the DC voltage and current on the rectifier side; inputting the inherent fault components into a CNN-BiLSTM-SAM model to output a fault cause probability distribution vector based on electrical quantities; simultaneously using external meteorological information as input and calculating the fault cause probability distribution vector using Bayes' theorem; adaptively and dynamically determining the external fault cause fusion weight coefficients by calculating the JS divergence between the two probability distribution vectors, and performing weighted fusion to achieve accurate identification of four fault causes: lightning strikes, wildfires, wind deflection, and foreign objects. This invention eliminates the influence of converter station control response, providing reliable technical support for the operation and maintenance of DC transmission lines.

Description

Technical Field

[0001] This invention relates to the field of power system protection and control, and specifically to a method and system for identifying faults in DC transmission lines that takes into account the control response of converter stations. Background Technology

[0002] With the large-scale grid connection of renewable energy sources such as wind power and photovoltaics, high-voltage direct current (HVDC) transmission, with its advantages of high transmission efficiency and large transmission capacity, has become the core carrier for large-capacity, long-distance, inter-regional power transmission in my country. As a core component of the HVDC transmission system, HVDC transmission lines have long transmission distances and wide-ranging corridors, and are constantly exposed to complex outdoor environments. They are highly susceptible to failures due to external factors such as lightning strikes, wildfires, wind deflection, and foreign object contact, making them the link with the highest failure rate in the HVDC transmission system. Therefore, timely and accurate identification of the causes of HVDC transmission line failures can guide maintenance personnel to quickly develop targeted repair and handling plans, shorten line downtime, and facilitate rapid system power restoration after a failure. This is of paramount importance for ensuring the safe and stable operation of the HVDC transmission system and the entire power grid.

[0003] Domestic and international scholars have conducted extensive research on fault identification in transmission lines. Currently, fault identification methods for DC transmission lines mainly rely on the changes in electrical quantities such as voltage and current after a fault occurs. In actual DC transmission systems, after a transmission line fault occurs, the converter station control system reacts rapidly, resulting in the superposition of control response components in the DC voltage and current on the rectifier side. Because the control response is coupled with the electromagnetic transient process, directly using fault recordings as input to the identification model reduces the distinguishability between different fault causes, affecting the accuracy of fault cause identification.

[0004] In recent years, the application of artificial intelligence algorithms in fault identification has provided new solutions for intelligent operation and maintenance of power systems. Utilizing deep learning methods to automatically extract features from fault waveforms has become an important means of improving identification performance. However, if the input data still contains additional components caused by converter station control responses, deep learning-based fault cause identification models for DC transmission lines may still face problems such as insufficient feature extraction specificity, limited model generalization ability, and weak robustness under complex operating conditions.

[0005] In summary, how to eliminate the impact of converter station control response during faults and propose identification methods for DC transmission system faults have become urgent problems for those skilled in the art. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention proposes a fault identification method for DC transmission lines that considers the control response of converter stations. This method includes:

[0007] S101. Collect fault samples under DC transmission line faults, including DC voltage and DC current fault waveforms on the rectifier side, as well as seasonal and weather information when the fault occurs.

[0008] S102. Calculate the control-related additional components of the rectifier-side DC voltage and DC current, and extract the fault-related inherent components of the rectifier-side DC voltage and DC current.

[0009] S103. Input the inherent components of DC voltage and DC current faults on the rectifier side into the fault cause identification model of the DC transmission system, and output the fault cause probability distribution vector based on electrical quantities; wherein the fault cause identification model of the DC transmission system adopts a convolution-bidirectional long short-term memory neural network-spatial attention mechanism;

[0010] S104. Calculate the probability distribution vector of fault causes based on meteorological information using Bayes' theorem.

[0011] S105. Calculate the JS divergence between the fault cause probability distribution vector based on electrical quantities and the fault cause probability distribution vector based on external meteorological information, and calculate the external fault cause fusion coefficient based on the JS divergence.

[0012] S106. Determine the fault cause identification results of DC transmission lines based on the external fault cause fusion coefficient.

[0013] A fault identification system for DC transmission lines that takes into account the control response of converter stations, the system comprising: a data acquisition module, a preprocessing module, a first calculation module, a second calculation module, a third calculation module, a fourth calculation module, a fifth calculation module, and a comparison module;

[0014] The acquisition module is used to synchronously acquire fault waveform data on the rectifier side and seasonal and weather information corresponding to the time of the fault.

[0015] The preprocessing module is used to uniformly sample the rectifier-side fault recording data to 10kHz and normalize it, and to perform word segmentation and one-hot encoding on external information to generate standardized input data.

[0016] The first calculation module is used to calculate the control-related additional components of the rectifier-side DC current and voltage, and extract the fault-related inherent components of the rectifier-side DC current and voltage;

[0017] The second calculation module is used to calculate the probability distribution vector of fault causes based on fault electrical quantities according to the inherent components of DC current and voltage faults on the rectifier side. ;

[0018] The third calculation module is used to calculate the probability distribution vector of fault causes based on external meteorological information. ;

[0019] The fourth calculation module is used to calculate the JS divergence between the fault cause probability distribution vector based on the fault electrical quantity and the fault cause probability distribution vector based on external meteorological information, and to determine the external fault cause fusion coefficient. ;

[0020] The fifth calculation module is used to perform weighted fusion of the output results of the two identification models based on the external fault cause fusion coefficient, and calculate the probability distribution vector of fault causes in DC transmission lines. ;

[0021] The comparison module is used to compare the fault cause probability distribution vector generated after dynamic fusion. The system compares the results, extracts the fault cause label corresponding to the highest probability value, and outputs the fault cause identification results for DC transmission lines.

[0022] The beneficial effects of this invention are:

[0023] 1. Existing transmission line fault identification technologies are geared towards AC transmission systems and cannot adequately account for the influence of converter station control responses in DC transmission systems. This results in the weakening of fault characteristics caused by different reasons and insufficient identification accuracy. This invention extracts the inherent components of DC voltage and current faults on the rectifier side, which are determined solely by the physical boundary conditions of the fault point, by analytically calculating the additional components of converter station control. This effectively eliminates interference from converter station control and significantly improves the distinctiveness and identifiability of various fault characteristics.

[0024] 2. Current transmission line fault identification technologies mostly use fixed fusion weights, which have limited identification accuracy. Furthermore, data-driven fault identification models are prone to outputting erroneous results when dealing with extreme or unbalanced samples. This invention introduces seasonal and weather information at the time of the fault and proposes an adaptive dynamic weighted fusion mechanism to dynamically adjust the fusion weights, thereby improving the accuracy of DC transmission line fault identification under complex and severe operating conditions. Attached Figure Description

[0025] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0026] Figure 1 This is a flowchart of a DC transmission line fault identification method that considers converter station control response, as disclosed in this invention.

[0027] Figure 2 This is the test result of a test sample set for a DC transmission line fault identification method that considers the control response of converter stations, according to an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] This invention discloses a method for fault identification of DC transmission lines that considers the control response of converter stations. The specific process is as follows: Figure 1 As shown, it includes the following steps:

[0030] S101. Collect fault samples under DC transmission line faults, including DC voltage and DC current fault waveforms on the rectifier side, as well as seasonal and weather information when the fault occurs.

[0031] S102. Calculate the control-related additional components of the rectifier-side DC voltage and DC current, and extract the fault-related inherent components of the rectifier-side DC voltage and DC current.

[0032] S103. Input the inherent components of DC voltage and DC current faults on the rectifier side into the fault cause identification model of the DC transmission system, and output the fault cause probability distribution vector based on electrical quantities; wherein the fault cause identification model of the DC transmission system adopts a convolution-bidirectional long short-term memory neural network-spatial attention mechanism;

[0033] S104. Calculate the probability distribution vector of fault causes based on meteorological information using Bayes' theorem.

[0034] S105. Calculate the JS divergence between the fault cause probability distribution vector based on electrical quantities and the fault cause probability distribution vector based on external meteorological information, and calculate the external fault cause fusion coefficient based on the JS divergence.

[0035] S106. Determine the fault cause identification results of DC transmission lines based on the external fault cause fusion coefficient.

[0036] In this embodiment of the invention, in step S101, the fault recording data on the rectifier side includes the instantaneous values ​​of voltage and current from 10ms before the fault to 20ms after the fault.

[0037] In specific implementation, in step S101, the fault samples are preprocessed according to the following method:

[0038] S1011. The sampling frequency of the rectifier-side fault recording data is unified to 10kHz, and it is normalized to map the values ​​to the interval [-1,1] to eliminate differences in units and ranges. The calculation formula is as follows:

[0039]

[0040] in, This represents the data after normalization. Represents the original data. This represents the upper limit of the original data. This represents the lower limit of the original data.

[0041] S1012. For unstructured text information describing the season and weather conditions of the fault, word segmentation technology and keyword matching are used to extract discrete feature words that represent the causes of the fault. The discrete feature words are then mapped into numerical feature vectors using a one-hot coding algorithm to construct a set of external fault cause features for Bayesian probability model calculation.

[0042] In specific implementation, in step S102, the inherent component of the DC current fault on the rectifier side is extracted in the following manner:

[0043]

[0044] in, This is the measured value of the DC current on the rectifier side; The additional component for rectifier-side DC current control is determined as follows:

[0045]

[0046] in, This is the commanded value of the rectifier-side DC current during normal operation. The DC current command value on the rectifier side during a fault is determined as follows:

[0047]

[0048] in, This is the DC voltage on the inverter side; and These are the upper and lower limit threshold values ​​for low-voltage current limiting control current; and These are the upper and lower threshold values ​​for the low-voltage current limiting control voltage; parameters and Determine as follows:

[0049]

[0050] In specific implementation, in step S102, the inherent component of the DC voltage fault on the rectifier side is extracted in the following manner:

[0051]

[0052] in, This is the measured value of the DC current on the rectifier side; The additional component for rectifier-side DC current control is calculated as follows:

[0053]

[0054] in, and These are the proportional and integral coefficients for constant current control, respectively; k is the proportional coefficient for trigger angle linearization. This refers to the effective value of the AC bus voltage at the LCC rectifier station. The number of 6-pulse converter bridges per pole of the converter station; For converter transformer turns ratio; This refers to the leakage reactance of the converter transformer.

[0055] In specific implementation, in step S102, the firing angle linearization scaling factor k is a linear function of the firing angle and the rectifier-side firing angle. Within the range of firing angle variation on the rectifier side The objective is to minimize the error within the specified range. The objective function is constructed as follows:

[0056]

[0057]

[0058] in, The rectifier-side firing angle before the fault; The rectifier-side firing angle after a fault;

[0059] According to the Karush-Kuhn-Tucker (KKT) conditions, the partial derivatives should be equal to 0 when the objective function F is minimized:

[0060]

[0061] The linearization scaling factor k is calculated as follows:

[0062]

[0063] In specific implementation, the fault cause identification model of DC transmission system based on CNN-BiLSTM-SAM in step S103 is determined as follows:

[0064] S1301. Multi-scale local features of the inherent components of DC voltage and current faults on the rectifier side are extracted using the convolutional and pooling layers of a CNN network. The calculation formula is as follows:

[0065]

[0066] in, This is the feature output of the (k-1)th layer; This is the feature output of the k-th layer; For convolution calculation; The weight matrix of the convolution kernel; The bias of the j-th convolutional kernel; This is the ReLU activation function.

[0067] S1302. Using SAM to weight and enhance the feature map output by the CNN, extracting the regional features that play a major role in identifying the cause of the fault, and outputting a weighted feature matrix. The weighting calculation process is as follows:

[0068]

[0069] in, Input feature maps to the CNN model, including the inherent component waveforms of rectifier-side DC voltage and current faults; and This is the weight matrix; This is the softmax normalization function.

[0070] S1303. Input the weighted feature matrix into the BiLSTM network to extract bidirectional temporal correlation features, and output the bidirectional temporal feature sequence. The calculation formula is as follows:

[0071]

[0072] in, For traditional LSTM network computing units; The output state of the forward hidden layer at time t; The output state of the hidden layer at time t; and These are the weight matrix and the bias, respectively.

[0073] S1304, The bidirectional time-series feature sequence The input is linearly mapped to a fully connected layer and normalized using a softmax function, outputting a probability distribution vector of fault causes based on fault electrical quantities. .

[0074] In specific implementation, in step S104, the probability distribution vector of fault causes based on meteorological information is determined in the following way:

[0075] Define the set of fault causes ,in ~ These respectively represent lightning strikes, wildfires, wind deflection, and foreign objects; external meteorological data sets. ,in These indicate the season of the malfunction and the weather at the location of the malfunction, respectively.

[0076] S1401. Calculate the cause of failure A under the given failure season and weather conditions at the failure point. i The probability of is calculated using the following formula:

[0077]

[0078] S1402. Concatenate the probabilities of each cause of failure in sequence under the given failure season and weather conditions at the failure point to form a probability distribution vector of failure causes based on external meteorological information. :

[0079]

[0080] In specific implementation, the external fault cause fusion coefficient in step S105 is determined as follows:

[0081] S1051. Calculate the probability distribution vector of fault causes based on fault electrical quantities. Compared with the probability distribution vector of fault causes based on external meteorological information JS divergence:

[0082] Where M represents the total number of fault cause categories for DC transmission lines.

[0083] S1052. Calculate the fusion coefficient of external fault causes based on JS divergence results:

[0084]

[0085] In specific implementation, the result of the fault cause identification of the DC transmission line in step S106 is determined in the following manner:

[0086]

[0087] To verify the effectiveness of this invention, 800 sets of DC transmission line fault samples (200 samples for each type of fault) were used from a power grid for four typical fault causes to verify the effectiveness of the proposed method. 70% of the samples from each fault cause were randomly selected as the training set, and 30% as the test set for model training and testing. Fault waveform data from 10 ms before the fault to 20 ms after the fault were selected, and the sampling frequency was standardized to 10 kHz. Because the waveform data differs in units and voltage levels, data features with small electrical quantities are difficult to extract when used as model input; therefore, the data was normalized to the interval [-1, 1].

[0088] During training, the model's loss function was chosen as cross-entropy loss, and the Adam algorithm was selected as the optimizer. The optimal learning rate was ultimately determined to be 0.01, and the optimal batch size was 16. Under these parameter settings, the fault cause identification model achieved an accuracy of 93.38% in identifying the causes of four types of DC transmission line faults. The confusion matrix for fault identification is as follows: Figure 2 As shown.

[0089] To verify the effectiveness of the proposed method for extracting inherent components of rectifier-side faults, a comparative experiment based on the input data type was conducted. Under the same network parameter settings, the original rectifier-side DC fault electrical quantities and the extracted inherent components of rectifier-side faults were used as model inputs, and the identification accuracy is shown in Table 1. Experimental results show that when the original fault waveform without control response interference is directly input, the model identification accuracy is only 89.38%; however, when the extracted inherent components of the fault are used as input, the model identification accuracy increases to 93.38%. This indicates that the inherent components of the fault can effectively weaken the influence of the converter station control response on the fault electrical quantities, enhance the differences between different fault causes, and thus improve the accuracy of fault cause identification in DC transmission lines.

[0090] Table 1. Accuracy of fault cause identification for different inputs

[0091]

[0092] A fault identification system for DC transmission lines that considers the control response of converter stations, the system comprising: a data acquisition module, a preprocessing module, a first calculation module, a second calculation module, a third calculation module, a fourth calculation module, a fifth calculation module, and a comparison module;

[0093] The acquisition module is used to synchronously acquire fault waveform data on the rectifier side and seasonal and weather information corresponding to the time of the fault.

[0094] The preprocessing module is used to uniformly sample the rectifier-side fault recording data to 10kHz and normalize it, and to perform word segmentation and one-hot encoding on external information to generate standardized input data.

[0095] The first calculation module is used to calculate the control-related additional components of the rectifier-side DC current and voltage, and extract the fault-related inherent components of the rectifier-side DC current and voltage;

[0096] The second calculation module is used to calculate the probability distribution vector of fault causes based on fault electrical quantities according to the inherent components of DC current and voltage faults on the rectifier side. ;

[0097] The third calculation module is used to calculate the probability distribution vector of fault causes based on external meteorological information. ;

[0098] The fourth calculation module is used to calculate the JS divergence between the fault cause probability distribution vector based on the fault electrical quantity and the fault cause probability distribution vector based on external meteorological information, and to determine the external fault cause fusion coefficient. ;

[0099] The fifth calculation module is used to perform weighted fusion of the output results of the two identification models based on the external fault cause fusion coefficient, and calculate the probability distribution vector of fault causes in DC transmission lines. ;

[0100] The comparison module is used to compare the fault cause probability distribution vector generated after dynamic fusion. The system compares the results, extracts the fault cause label corresponding to the highest probability value, and outputs the fault cause identification results for DC transmission lines.

[0101] The system implementation method of the present invention is the same as that of the invention.

[0102] The above-described embodiments further illustrate the purpose, technical solution, and advantages of the present invention. It should be understood that the above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made to the present invention within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for fault identification of DC transmission lines considering converter station control response, characterized in that, include: S101. Collect fault samples under DC transmission line faults, including DC voltage and DC current fault waveforms on the rectifier side, as well as seasonal and weather information when the fault occurs. S102. Calculate the control-related additional components of the rectifier-side DC voltage and DC current, and extract the fault-related inherent components of the rectifier-side DC voltage and DC current. S103. Input the inherent components of DC voltage and DC current faults on the rectifier side into the fault cause identification model of the DC transmission system, and output the fault cause probability distribution vector based on electrical quantities; wherein the fault cause identification model of the DC transmission system adopts a convolution-bidirectional long short-term memory neural network-spatial attention mechanism; S104. Calculate the probability distribution vector of fault causes based on meteorological information using Bayes' theorem. S105. Calculate the JS divergence between the fault cause probability distribution vector based on electrical quantities and the fault cause probability distribution vector based on external meteorological information, and calculate the external fault cause fusion coefficient based on the JS divergence. S106. Determine the fault cause identification results of DC transmission lines based on the external fault cause fusion coefficient.

2. The method for fault identification of DC transmission lines considering converter station control response according to claim 1, characterized in that, The fault recording on the rectifier side includes the instantaneous values ​​of voltage and current from 10ms before the fault to 20ms after the fault. The fault samples are preprocessed, specifically including: the sampling frequency of the rectifier side fault recording data is unified to 10kHz and normalized so that the values ​​are mapped to the range of [-1,1]; the unstructured text information describing the fault season and weather conditions is processed by word segmentation and keyword matching to extract discrete feature words that characterize the fault causes, and the discrete feature words are mapped to numerical feature vectors using the one-hot coding algorithm to construct an external fault cause feature set for Bayesian probability model calculation.

3. The method for fault identification of DC transmission lines considering converter station control response according to claim 1, characterized in that, The additional component for rectifier-side DC current control is: ； ； ； in, This is the commanded value of the rectifier-side DC current during normal operation. This is the commanded DC current value on the rectifier side during the fault. This is the DC voltage on the inverter side; and These are the upper and lower limit threshold values ​​for low-voltage current limiting control current; and These are the upper and lower threshold values ​​for the low-voltage current limiting control voltage; and All are control parameters; The additional control component of the rectifier-side DC voltage is: ； in, and These are the proportional and integral coefficients for constant current control, respectively; k is the proportional coefficient for trigger angle linearization. This refers to the effective value of the AC bus voltage at the LCC rectifier station. The number of 6-pulse converter bridges per pole of the converter station; For converter transformer turns ratio; This refers to the leakage reactance of the converter transformer.

4. The method for fault identification of DC transmission lines considering converter station control response according to claim 3, characterized in that, The inherent component of DC current fault on the rectifier side is: ； in, This is the measured value of the DC current on the rectifier side; Adds an additional component to the DC current control on the rectifier side; The inherent component of DC voltage fault on the rectifier side is: ； in, This is the measured value of the DC voltage on the rectifier side; An additional component is added to the DC voltage control on the rectifier side.

5. The method for fault identification of DC transmission lines considering converter station control response according to claim 3, characterized in that, The firing angle linearization scaling factor is calculated as follows: it is a linear function of the firing angle and the firing angle on the rectifier side. Within the range of firing angle variation on the rectifier side The objective is to minimize the fitting error within the range. The optimal value is obtained by constructing a fitting error optimization objective function using the least squares method, where the optimal solution is the trigger angle linearization scaling factor, expressed as: ； in, The rectifier-side firing angle before the fault. This is the firing angle on the rectifier side.

6. The method for fault identification of DC transmission lines considering converter station control response according to claim 1, characterized in that, The fault cause identification model for DC transmission systems processes the inherent components of DC voltage and DC current faults on the rectifier side, including: S1031. Utilize the convolutional and pooling layers of a CNN network to extract multi-scale local features of the inherent components of DC voltage and current faults on the rectifier side. S1032. Use SAM to weight and enhance the feature map output by CNN, extract the regional features that play a major role in identifying the cause of the fault, and output the weighted feature matrix. S1033. Input the weighted feature matrix into the BiLSTM network to extract bidirectional temporal correlation features, and output the bidirectional temporal feature sequence. ; S1034, The bidirectional time-series feature sequence The input is linearly mapped to a fully connected layer and normalized using a softmax function, outputting a probability distribution vector of fault causes based on fault electrical quantities. .

7. The method for fault identification of DC transmission lines considering converter station control response according to claim 1, characterized in that, Calculating the probability distribution vector of fault causes based on meteorological information includes: calculating the fault cause A under the given fault season and weather conditions at the fault location. i The probability of each cause of failure is calculated by concatenating the probabilities of each cause in sequence under given weather conditions at the failure season and location, forming a probability distribution vector of failure causes based on external meteorological information. .

8. The method for fault identification of DC transmission lines considering converter station control response according to claim 1, characterized in that, The fusion coefficient of external failure causes includes: ； ； in, The fusion coefficient of external fault causes. Let JS divergence be the metric. This is a probability distribution vector of fault causes based on external meteorological information. Let M be the probability distribution vector of fault causes based on the fault electrical quantities, where M is the total number of fault cause categories in the DC transmission line. Given the fault season and the weather conditions at the fault location, this represents the probability of the fault cause.

9. The method for fault identification of DC transmission lines considering converter station control response according to claim 1, characterized in that, The results of the fault cause identification for DC transmission lines include: ； in, The fusion coefficient of external fault causes. This is a probability distribution vector of fault causes based on external meteorological information. This is a probability distribution vector of fault causes based on fault electrical quantities.

10. A DC transmission line fault identification system considering converter station control response, the system being used to execute the DC transmission line fault identification method considering converter station control response as described in any one of claims 1 to 9, characterized in that, The system includes: a data acquisition module, a preprocessing module, a first calculation module, a second calculation module, a third calculation module, a fourth calculation module, a fifth calculation module, and a comparison module; The acquisition module is used to synchronously acquire fault waveform data on the rectifier side and seasonal and weather information corresponding to the time of the fault. The preprocessing module is used to uniformly sample the rectifier-side fault recording data to 10kHz and normalize it, and to perform word segmentation and one-hot encoding on external information to generate standardized input data. The first calculation module is used to calculate the control-related additional components of the rectifier-side DC current and voltage, and extract the fault-related inherent components of the rectifier-side DC current and voltage; The second calculation module is used to calculate the probability distribution vector of fault causes based on fault electrical quantities according to the inherent components of DC current and voltage faults on the rectifier side. ; The third calculation module is used to calculate the probability distribution vector of fault causes based on external meteorological information. ; The fourth calculation module is used to calculate the JS divergence between the fault cause probability distribution vector based on the fault electrical quantity and the fault cause probability distribution vector based on external meteorological information, and to determine the external fault cause fusion coefficient. ; The fifth calculation module is used to perform weighted fusion of the output results of the two identification models based on the external fault cause fusion coefficient, and calculate the probability distribution vector of fault causes in DC transmission lines. ; The comparison module is used to compare the fault cause probability distribution vector generated after dynamic fusion. The system compares the results, extracts the fault cause label corresponding to the highest probability value, and outputs the fault cause identification results for DC transmission lines.

Images