Fault diagnosis method and system for HVDC transmission system based on K-means algorithm

Abstract

The present invention provides a high-voltage direct current (HVDC) system fault diagnosis method and system based on the K-means algorithm. Based on fault data obtained from HVDC system fault recordings, the fault data is subjected to zero-mean normalization processing by channel to form an original data set. The fault sample data in the original data set is labeled, and a KM fault diagnosis model is established. The number of initial cluster centers of the model is set. After the model is verified and qualified using the training and test sets, an optimal KM fault diagnosis model is obtained. The optimal KM fault diagnosis model is used to diagnose real-time HVDC system faults and generate visual diagnostic results. The present invention utilizes a high-voltage direct current (HVDC) system fault diagnosis method based on the K-means algorithm to improve the efficiency and accuracy of fault diagnosis, helping engineers quickly resolve problems and reduce the impact of faults.

Description

Technical Field

[0001] The present application relates to the technical field of power system protection, and in particular to a high-voltage direct current transmission system fault diagnosis method and system based on a K-means algorithm. Background Art

[0002] In recent years, high-voltage direct current (HVDC) technology has developed rapidly. It not only offers long-distance, high-capacity, and low-loss power transmission capabilities, but also features low line costs and excellent controllability. Due to its unique advantages, HVDC technology has become a crucial tool for power-rich nations to address high-voltage, high-capacity, long-distance power transmission and grid interconnection. With the national "carbon peak and carbon neutrality" policy, higher requirements have been placed on the environmental and energy-saving performance of power systems. Furthermore, as power systems expand in size due to rising electricity demand, transmission power and distances are also increasing. AC transmission technology has shown many limitations, such as the significant losses it incurs over long transmission distances. DC transmission technology can compensate for these shortcomings. Therefore, HVDC technology is playing an increasingly important role in power grids.

[0003] Improving the safety and reliability of system operation has also become a pressing issue. Due to the complexity and widespread distribution of DC systems, the interactions between AC and DC systems are even more complex. Complex systems are prone to failure, which can easily cause relay protection devices and control circuits to malfunction, leading to interruptions in DC power transmission and impacting the stable operation of the entire AC and DC system. This poses a significant risk of failures in the power system and can lead to even more serious power accidents. HVDC systems often serve as the link between major power grids, and their reliability is crucial to the national economy. Failures and power outages can halt production, disrupt operations, and even endanger personnel and equipment.

[0004] Therefore, fault diagnosis is the core of HVDC technology and the guarantee of HVDC system power supply reliability. Therefore, it is very urgent to conduct research on fault diagnosis of HVDC system. Summary of the Invention

[0005] This application provides a high-voltage direct current (HVDC) system fault diagnosis method and system based on the K-means algorithm. To address the serious impact of current HVDC system faults and the need to improve power supply reliability, this application proposes using the K-means (KM) algorithm to diagnose HVDC system faults. This provides a practical and highly accurate diagnostic model, which is of great significance for ensuring the long-term safe operation of HVDC systems.

[0006] Specifically, the HVDC system fault diagnosis method based on the K-means algorithm includes:

[0007] S1: Based on the fault data obtained from the HVDC system fault recording, the fault data is processed with zero mean normalization by channel to form the original data set. By performing zero mean normalization on the HVDC system fault recording data, the noise interference of the data can be reduced and the training and diagnosis effects of the model can be improved.

[0008] S2: labeling the fault sample data in the original data set and dividing it into a training set and a test set;

[0009] S3: Establishing a KM fault diagnosis model, setting the number of initial cluster centers of the model, and using the training set and the test set to complete the model verification and obtain the optimal KM fault diagnosis model;

[0010] S4: diagnose the real-time HVDC system fault using the optimal KM fault diagnosis model and generate a visual diagnosis result.

[0011] This invention reduces data noise by performing zero-mean normalization on HVDC system fault recording data, improving model training and diagnostic effectiveness. Using the optimal KM fault diagnosis model, real-time HVDC system faults are diagnosed, generating visual diagnostic results. These results can be displayed in charts, scatter plots, and other formats, helping engineers intuitively understand fault conditions and quickly and accurately locate and address them.

[0012] In some implementations, establishing the KM fault diagnosis model specifically includes:

[0013] S31: Randomly select k points from the original data set, each point initially represents the cluster center of each cluster;

[0014] S32: Calculate the distance between each remaining sample and the cluster center, and assign it to the nearest cluster;

[0015] S33: recalculate the mean of each cluster;

[0016] Repeat S31-S33 until the degree of change between two adjacent adjustments is less than a threshold, which indicates that the clusters formed by data clustering have converged.

[0017] This process can be viewed as an iterative process of the K-means algorithm, where the optimal clustering result is found by continuously updating the cluster centers and redistributing the data. Clustering convergence indicates that the cluster division has become stable and the desired clustering effect has been achieved.

[0018] In some implementations, establishing the KM fault diagnosis model further includes:

[0019] In each iteration, the classification of each sample is judged to be correct. If not, adjustments are made. After all samples are adjusted, the cluster center is modified and the next iteration is entered; otherwise, the iteration continues.

[0020] The iteration termination condition is set as:

[0021] No objects were reassigned to a different cluster;

[0022] Or if the cluster center changes again, it can be judged by comparing the difference between the cluster centers in two adjacent iterations. If the difference is less than the preset threshold, the algorithm can stop iterating.

[0023] Or the sum of squared errors reaches a local minimum threshold. The sum of squared errors (SSE) can be used to measure the effectiveness of clustering. In each iteration, the sum of the squares of the distances from all sample points to the cluster center is calculated. If this value is less than the preset minimum threshold, the algorithm can stop iterating.

[0024] In summary, when any of the termination conditions are met, the K-means algorithm stops iterating and outputs the final clustering result. These termination conditions are based on the different convergence and importance of the algorithm, and appropriate judgment criteria are selected to achieve satisfactory clustering results within a reasonable number of iterations.

[0025] The distance between the remaining samples and the cluster center is calculated using the formula:

[0026]

[0027] Where, ω j is the sample data set; x i is the position of the i-th sample point; m j is the position of the jth cluster center; E is the distance between each object and the cluster center; k is the number of initial cluster centers.

[0028] In some implementations, establishing the KM fault diagnosis model further includes:

[0029] The formula for calculating the sum of squared errors (SSE) for the entire data set is:

[0030]

[0031] Select the initial cluster centers: Based on the relationship between the number of clusters k and the error sum squared (SSE), and according to the changing trend of the error sum squared (SSE) for different k values, randomly select k initial cluster centers for one iteration and calculate the clustering measure value after clustering:

[0032]

[0033] In the formula, the size of SSE indicates the quality of clustering results; k is the number of clusters; c i is the cluster center of the i-th class; x ij is the jth data object in the i-th class; n i is the number of data objects in the i-th class; k is the number of clusters; and n is the size of the dataset.

[0034] In some implementations, establishing the KM fault diagnosis model further includes:

[0035] It is determined whether the sum of squared errors SSE changes after each iteration. If so, the process returns to step S32 and recalculates. Otherwise, the clustering is terminated and the final clustering result is output.

[0036] The establishment of the KM fault diagnosis model further includes: finding a new cluster center after each iteration, specifically:

[0037] After each iteration, the cluster with the largest radius among all clusters is found, and the two sample points with the farthest distance between them are selected as new cluster centers. The iteration is repeated with other cluster centers, and the ratio of clustering measure values is calculated after each iteration.

[0038] The present invention provides a supplementary explanation of the iterative process of establishing a KM fault diagnosis model. During the iterative process, by continuously adjusting the cluster centers, recalculating the cluster distribution and the sum of squared errors, and evaluating the clustering performance indicators, a stable clustering result is finally obtained for fault diagnosis and classification.

[0039] Based on the same inventive concept, the present invention also provides a high-voltage direct current transmission system fault diagnosis system based on the K-means algorithm, the diagnostic system comprising:

[0040] A data collection unit is used to process the fault data obtained from the HVDC system fault recording by zero mean value per channel to form an original data set;

[0041] A preprocessing unit, which performs labeling processing on the fault sample data in the original data set and divides the data into a training set and a test set;

[0042] The data processing unit diagnoses real-time HVDC system faults through the pre-established KM fault diagnosis model and generates visual diagnosis results.

[0043] This method uses the K-means algorithm to cluster HVDC system fault data. By analyzing and visualizing the clustering results, faults can be quickly and accurately diagnosed and located. Visualizing the diagnostic results helps engineers more intuitively understand the fault situation and improves the efficiency of troubleshooting.

[0044] Wherein, the data processing unit further includes:

[0045] The model building module is used to establish a KM fault diagnosis model, and obtain the optimal KM fault diagnosis model by setting the number of initial cluster centers of the model and completing the verification of the model using the training set and the test set;

[0046] A model optimization module is used to calculate the distance between each remaining sample and the cluster center, assign it to the nearest cluster, and iterate the calculation until the degree of change between two consecutive adjustments is less than a threshold;

[0047] Iterative adjustment module determines whether the classification of each sample is correct in each iteration. If not, it makes adjustments. After all samples have been adjusted, the cluster center is modified and the next iteration begins; otherwise, the iteration continues.

[0048] The iteration termination module determines whether the sum of squared errors (SSE) changes after each iteration. If so, it returns to recalculate the distance between cluster centers; otherwise, it terminates the clustering and outputs the final clustering result.

[0049] Furthermore, the diagnostic system also includes: an updating unit, which is used to find a new cluster center after each iteration, find the cluster with the largest radius among all clusters after each iteration, select the two sample points with the farthest distance between them as the new cluster center, and re-iterate with other cluster centers, and calculate the ratio of cluster measurement values after each iteration.

[0050] The present invention has the following beneficial effects compared to the prior art:

[0051] 1. The present invention adopts a high-voltage direct current transmission system fault diagnosis method based on the K-means algorithm to improve the efficiency and accuracy of fault diagnosis, help engineers solve problems quickly, and reduce the impact of faults.

[0052] 2. The present invention can meet the fault diagnosis requirements of the HVDC system in terms of diagnostic speed and diagnostic accuracy, and effectively improves the safety, stability and reliability of the HVDC system. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a method for fault diagnosis of a high-voltage direct current transmission system based on the K-means algorithm according to an embodiment of the present application.

[0054] Figure 2 This is a flow chart structure diagram of another HVDC system fault diagnosis method based on the K-means algorithm shown in one embodiment of the present application.

[0055] Figure 3 This is a flow chart structure diagram of another HVDC system fault diagnosis method based on the K-means algorithm shown in one embodiment of the present application.

[0056] Figure 4 A schematic diagram of a high-voltage direct current transmission system fault diagnosis system based on the K-means algorithm is shown for one embodiment of the present application.

[0057] Figure 5 A schematic diagram of a data processing unit is shown for one embodiment of the present application. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0059] Please refer to Figure 1-3 , which is a flow chart and principle diagram of a high-voltage direct current transmission system fault diagnosis method based on the K-means algorithm provided in the first embodiment of the present application.

[0060] The high-voltage direct current transmission system fault diagnosis method based on the K-means algorithm includes:

[0061] S1: Based on the fault data obtained from the HVDC system fault recording, the fault data is processed by zero mean normalization by channel to form the original data set;

[0062] S2: labeling the fault sample data in the original data set and dividing it into a training set and a test set;

[0063] S3: Establishing a KM fault diagnosis model, setting the number of initial cluster centers of the model, and using the training set and the test set to complete the model verification and obtain the optimal KM fault diagnosis model;

[0064] S4: diagnose the real-time HVDC system fault using the optimal KM fault diagnosis model and generate a visual diagnosis result.

[0065] This invention reduces data noise by performing zero-mean normalization on HVDC system fault recording data, improving model training and diagnostic effectiveness. Using the optimal KM fault diagnosis model, real-time HVDC system faults are diagnosed, generating visual diagnostic results. These results can be displayed in charts, scatter plots, and other formats, helping engineers intuitively understand fault conditions and quickly and accurately locate and address them.

[0066] In some implementations, in S1: collect fault data, aggregate all data, and establish an original data set. The specific meaning of each channel is shown in Table 1:

[0067] Table 1 Channel names and their meanings

[0068]

[0069] The fault data is normalized using zero mean (Z-Mean, ZM) as follows:

[0070] Collect HVDC system fault recording data and ensure that the data includes records of each channel.

[0071] For each channel's data, calculate the average value of the channel's data.

[0072] The mean value of each channel was subtracted from the data of the corresponding channel to obtain data with zero mean.

[0073]

[0074] Where x represents the fault data; y represents the normalized fault data; μ represents the mean; and σ represents the variance.

[0075] Among them, this method standardizes the parameters to normally distributed data with a mean of 0 and a variance of 1, indicating the distance the original data deviates from the mean, and the standard for measuring this distance is the standard deviation.

[0076] By performing zero-mean normalization on a per-channel basis, we can eliminate bias between channels and make the data more comparable. The resulting raw dataset can then be used as input for subsequent fault diagnosis models, training and testing their accuracy and effectiveness. This processing helps improve the performance and practicality of fault diagnosis models.

[0077] In some implementations, in S2: labeling process:

[0078] For the fault sample data in the original dataset, in order to conduct supervised learning and model training, it is necessary to label each sample accordingly. The specific label processing method can be defined according to the actual situation. The fault samples can be processed according to the following labels:

[0079] (1) Label 1: AC failure;

[0080] (2) Label 2: DC fault;

[0081] (3) Label 3: rectifier short circuit;

[0082] (4) Label 4: Inverter failure.

[0083] Each fault sample is classified into a corresponding label based on the actual fault type. This provides labeled training data for the model, enabling it to learn the characteristics of different fault types and thus achieve accurate fault classification and diagnosis.

[0084] And divided into training set and test set: 70% of the sample data are randomly selected as training data, and the remaining 30% of the sample data are used as test data.

[0085] In some implementations, the KM algorithm works as follows:

[0086] First, k points are randomly selected from the data set. Each point initially represents the cluster center of each cluster. Then, the distance from each remaining sample to the cluster center is calculated and assigned to the nearest cluster. Then, the average value of each cluster is recalculated. The whole process is repeated. If there is no obvious change between two adjacent adjustments, it means that the clusters formed by data clustering have converged.

[0087] That is, the establishment of the KM fault diagnosis model is specifically as follows:

[0088] S31: Randomly select k points from the original data set, each point initially represents the cluster center of each cluster;

[0089] S32: Calculate the distance between each remaining sample and the cluster center, and assign it to the nearest cluster;

[0090] S33: recalculate the mean of each cluster;

[0091] Repeat S31-S33 until the degree of change between two adjacent adjustments is less than a threshold, which indicates that the clusters formed by data clustering have converged.

[0092] For example, the specific model parameter settings are as follows:

[0093] In the model, the number of initial cluster centers of the KM fault diagnosis model is set to 4, and the remaining data objects and cluster centers C are calculated. i(1≤i≤4) ED, find the cluster center C closest to the target data object i , and assign the data objects to the cluster center C i The corresponding cluster.

[0094] convergence.

[0095] The K-means algorithm, through a continuous iterative process, divides data points into k clusters, ensuring that the data points within each cluster are as similar as possible and that the differences between clusters are large. The KM model is trained using the training dataset to ensure that the model fits the sample well. During training, the model's performance and progress are monitored and analyzed. The test dataset is used to ultimately test the model's generalization to the detection task, thereby verifying the feasibility of the KM fault diagnosis model.

[0096] The characteristic of this algorithm is that it checks the classification of each sample in each iteration. If it is incorrect, it is adjusted. After all samples have been adjusted, the cluster center is modified and the next iteration begins. This process will be repeated until a termination condition is met, which can be any of the following:

[0097] No objects were reassigned to a different cluster;

[0098] Or the cluster center changes again;

[0099] Or the sum of squared errors locally reaches the minimum threshold.

[0100] The above termination conditions can be used independently or in combination. The specific judgment criteria to be selected depends on the specific requirements of the problem and the characteristics of the data, and are not limited thereto.

[0101] This process can be viewed as an iterative process of the K-means algorithm, where the optimal clustering result is found by continuously updating the cluster centers and redistributing the data. Clustering convergence indicates that the cluster division has become stable and the desired clustering effect has been achieved.

[0102] To further explain the KM fault diagnosis model in detail, the specific steps of the KM algorithm in the present invention are as follows:

[0103] ① Randomly select k objects from n data objects as the initial cluster centers so that the obtained clusters satisfy the following conditions: objects in the same cluster have high similarity, while objects in different clusters have low similarity. Cluster similarity is calculated by obtaining a "center object" using the mean of the objects in each cluster;

[0104] ② According to the mean of each cluster object, calculate the distance between each object and these central objects, and re-divide the corresponding objects according to the minimum distance;

[0105] ③ Recalculate the mean of each cluster until the cluster center no longer changes; this division minimizes the following formula:

[0106]

[0107] Where, ω j is the sample data set; x i is the position of the i-th sample point; m j is the position of the jth cluster center; E is the distance between each object and these central objects; k is the number of initial cluster centers.

[0108] ④ Repeat steps ② and ③ until each cluster no longer changes.

[0109] There are three commonly used distance functions between data objects and cluster centers in space:

[0110] ①Manhattan distance:

[0111]

[0112] In the formula, x is the data object; C i is the i-th cluster center; m is the dimension of the data object; x j , C ij x and C respectively i The j-th attribute value of .

[0113] ② Mahalanobis distance:

[0114]

[0115] In the formula, x is the data object; C i is the i-th cluster center; A represents x, C i The covariance matrix of .

[0116] ③Euclidean Distance (ED):

[0117]

[0118] In the formula, x is the data object; C i is the i-th cluster center; m is the dimension of the data object; x j , C ij x and C respectively i The j-th attribute value of .

[0119] Preferably, the ED Euclidean distance is adopted in the present invention.

[0120] The formula for calculating the sum of squared errors (SSE) for the entire data set is:

[0121]

[0122] Select the initial cluster centers: Based on the relationship between the number of clusters k and the error sum squared (SSE), and according to the changing trend of the error sum squared (SSE) for different k values, randomly select k initial cluster centers for one iteration and calculate the clustering measure value after clustering:

[0123]

[0124] In the formula, the size of SSE indicates the quality of clustering results; k is the number of clusters; c i is the cluster center of the i-th class; x ij is the jth data object in the i-th class; n i is the number of data objects in the i-th class; k is the number of clusters; and n is the size of the dataset.

[0125] In some implementations, establishing the KM fault diagnosis model further includes:

[0126] It is determined whether the sum of squared errors SSE changes after each iteration. If so, the process returns to step S32 and recalculates. Otherwise, the clustering is terminated and the final clustering result is output.

[0127] The establishment of the KM fault diagnosis model further includes: finding a new cluster center after each iteration, specifically:

[0128] After each iteration, the cluster with the largest radius is found. Within this cluster, the two sample points with the greatest distance between them are selected as new cluster centers. The algorithm then iterates again with the remaining cluster centers. After each iteration, the ratio of clustering metrics is calculated. Clustering metric ratios can be used to evaluate clustering performance, with common examples including the silhouette coefficient and the Davies-Bouldin index. By comparing the ratios of clustering metrics across iterations, the quality and stability of clustering results can be assessed.

[0129] For example, the ratio of clustering measure values is calculated using the formula:

[0130]

[0131] Where t is the number of iterations.

[0132] In this invention, the core idea of the KM algorithm is: first randomly select k initial cluster centers C from the data set i (1≤i≤k), calculate the remaining data objects and cluster center C i Euclidean distance, find the cluster center C closest to the target data object i , and assign the data objects to the cluster center C iThen calculate the average value of the data objects in each cluster as the new cluster center, and perform the next iteration until the cluster center no longer changes or the maximum number of iterations is reached.

[0133] After the model is built, the test data is substituted into the model to verify the effectiveness and accuracy of the KM, and the diagnosis results are visualized. The specific implementation method is to draw a confusion matrix based on the final diagnosis results.

[0134] Example 2:

[0135] like Figure 4-5 As shown, the present invention also provides a high-voltage direct current transmission system fault diagnosis system based on the K-means algorithm, the diagnosis system comprising:

[0136] A data collection unit is used to process the fault data obtained from the HVDC system fault recording by zero mean value per channel to form an original data set;

[0137] A preprocessing unit, which performs labeling processing on the fault sample data in the original data set and divides the data into a training set and a test set;

[0138] The data processing unit diagnoses real-time HVDC system faults through the pre-established KM fault diagnosis model and generates visual diagnosis results.

[0139] This method uses the K-means algorithm to cluster HVDC system fault data. By analyzing and visualizing the clustering results, faults can be quickly and accurately diagnosed and located. Visualizing the diagnostic results helps engineers more intuitively understand the fault situation and improves the efficiency of troubleshooting.

[0140] Wherein, the data processing unit further includes:

[0141] The model building module is used to establish a KM fault diagnosis model, and obtain the optimal KM fault diagnosis model by setting the number of initial cluster centers of the model and completing the verification of the model using the training set and the test set;

[0142] A model optimization module is used to calculate the distance between each remaining sample and the cluster center, assign it to the nearest cluster, and iterate the calculation until the degree of change between two consecutive adjustments is less than a threshold;

[0143] Iterative adjustment module determines whether the classification of each sample is correct in each iteration. If not, it makes adjustments. After all samples have been adjusted, the cluster center is modified and the next iteration begins; otherwise, the iteration continues.

[0144] The iteration termination module determines whether the sum of squared errors (SSE) changes after each iteration. If so, it returns to recalculate the distance between cluster centers; otherwise, it terminates the clustering and outputs the final clustering result.

[0145] Furthermore, the diagnostic system also includes: an updating unit, which is used to find a new cluster center after each iteration, find the cluster with the largest radius among all clusters after each iteration, select the two sample points with the farthest distance between them as the new cluster center, and re-iterate with other cluster centers, and calculate the ratio of cluster measurement values after each iteration.

[0146] Through the above specific process, the final YOLOv4 fault identification network model is determined based on the detection results and used for HVDC system fault diagnosis, realizing automated fault diagnosis and processing. This can improve the accuracy and efficiency of fault diagnosis and help ensure the stable operation and safety of the power system.

[0147] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application by those skilled in the art should be included within the scope of protection of this application.

Claims

1. A fault diagnosis method for a high-voltage direct current transmission system based on a K-means algorithm, characterized in that: The method comprises: S1: Based on the fault data obtained from the HVDC system fault recording, the fault data is processed by zero mean normalization by channel to form the original data set; S2: labeling the fault sample data in the original data set and dividing it into a training set and a test set, wherein the labeling process includes: For the fault sample data in the original data set, each sample is labeled accordingly, and the fault samples are processed according to the following labels: Label 1: AC failure; Label 2: DC fault; Label 3: rectifier short circuit; Label 4: Inverter failure; According to the actual fault type, each fault sample is classified into the corresponding label; S3: Establishing a KM fault diagnosis model, setting the number of initial cluster centers of the model, and using the training set and the test set to complete the model verification and obtain the optimal KM fault diagnosis model; The establishment of the KM fault diagnosis model specifically includes: S31: Randomly select k points from the original data set, each point initially represents the cluster center of each cluster; S32: Calculate the distance between each remaining sample and the cluster center, and assign it to the nearest cluster; S33: recalculate the mean of each cluster; Repeat S31-S33 until the change between two consecutive adjustments is less than a threshold, which means that the clusters formed by data clustering have converged; The establishment of the KM fault diagnosis model also includes: In each iteration, the classification of each sample is judged to be correct. If not, adjustments are made. After all samples are adjusted, the cluster center is modified and the next iteration is entered; otherwise, the iteration continues. The iteration termination condition is set as: No objects were reassigned to a different cluster; Or the cluster center changes again; or the sum of squared errors reaches the minimum threshold locally; The establishment of the KM fault diagnosis model further includes: finding a new cluster center after each iteration, specifically: finding the cluster with the largest radius among all clusters after each iteration, selecting two sample points with the farthest distance from each other in the cluster as new cluster centers, and iterating again with the other cluster centers, and calculating the ratio of cluster measure values after each iteration; S4: Using the optimal KM fault diagnosis model to diagnose real-time HVDC system faults and generate visual diagnosis results.

2. A method for fault diagnosis of a high voltage direct current transmission system based on a K-means algorithm according to claim 1, characterized in that: The distance between the remaining samples and the cluster center is calculated using the formula: Where, ω j is the sample data set; x i is the position of the i-th sample point; m j is the position of the jth cluster center; E is the distance between each object and the cluster center; k is the number of initial cluster centers.

3. The method for fault diagnosis of a high-voltage direct current transmission system based on the K-means algorithm according to claim 2, characterized in that: The establishment of the KM fault diagnosis model also includes: The formula for calculating the sum of squared errors (SSE) for the entire data set is: Select the initial cluster centers: Based on the relationship between the number of clusters k and the error sum squared (SSE), and according to the changing trend of the error sum squared (SSE) for different k values, randomly select k initial cluster centers for one iteration and calculate the clustering measure value after clustering: In the formula, the size of SSE indicates the quality of clustering results; k is the number of clusters; c i is the cluster center of the i-th class; x ij is the jth data object in the i-th class; n i is the number of data objects in the i-th class; k is the number of clusters; and n is the size of the dataset.

4. The method for fault diagnosis of a high-voltage direct current transmission system based on the K-means algorithm according to claim 3, characterized in that: The establishment of the KM fault diagnosis model also includes: It is determined whether the sum of squared errors SSE changes after each iteration. If so, the process returns to step S32 and recalculates. Otherwise, the clustering is terminated and the final clustering result is output.

5. A high voltage direct current transmission system fault diagnosis system based on K-means algorithm, characterized in that: The system is used to execute a high-voltage direct current transmission system fault diagnosis method based on a K-means algorithm according to any one of claims 1 to 4, wherein the diagnostic system comprises: A data collection unit is used to perform zero-mean normalization processing on the fault data obtained from the HVDC system fault recording according to the channel to form an original data set; A preprocessing unit, which performs labeling processing on the fault sample data in the original data set and divides the data into a training set and a test set; The data processing unit diagnoses real-time HVDC system faults through the pre-established KM fault diagnosis model and generates visual diagnosis results.

6. The diagnostic system according to claim 5, characterized in that The data processing unit further includes: a model construction module for establishing a KM fault diagnosis model, setting the number of initial cluster centers of the model, and obtaining an optimal KM fault diagnosis model after the model is verified to be qualified using the training set and the test set; a model optimization module for calculating the distance between each remaining sample and the cluster center, assigning it to the nearest cluster, and iterating the calculation until the degree of change between two adjacent adjustments is less than a threshold; The iterative adjustment module determines whether the classification of each sample is correct in each iteration. If not, it will be adjusted. After all samples are adjusted, the cluster center is modified and the next iteration is entered; otherwise, the iteration continues; the iterative termination module determines whether the sum of squared errors (SSE) changes after each iteration. If it changes, it returns to recalculate the distance to the cluster center; otherwise, the clustering is terminated and the final clustering result is output.

7. The diagnostic system according to claim 6, characterized in that The diagnostic system also includes: an updating unit, which is used to find a new cluster center after each iteration, find the cluster with the largest radius among all clusters after each iteration, select the two sample points with the farthest distance between them as the new cluster center, and iterate again with the other cluster centers, and calculate the ratio of the cluster measurement values after each iteration.

Images