Power grid fault event identification method and device based on data mining, terminal equipment and computer readable storage medium

Abstract

The invention discloses a power grid fault event identification method and device based on data mining, terminal equipment and a computer readable storage medium, and belongs to the field of data analysis, and the method comprises the steps: obtaining a transaction set in a to-be-processed time period; wherein each transaction comprises a plurality of pieces of operation alarm information; performing association rule mining on all the operation alarm information, and identifying a plurality of alarm information combinations; and for each alarm information combination, performing feature extraction on each operation alarm information in the alarm information combination to obtain a feature vector of each operation alarm information in the alarm information combination, inputting the feature vector into the trained fault event identification model, generating a power grid fault event corresponding to the alarm information combination, and performing fault identification on the power grid. By implementing the method and the device, the problem that monitoring personnel are easy to misjudge or miss important information when facing massive power grid operation alarm information in the prior art can be solved, and the event identification accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and in particular to a method, device, terminal equipment and computer-readable storage medium for identifying power grid fault events based on data mining. Background Art

[0002] Existing power grid monitoring services rely on a one-by-one response monitoring method. As the scale of the power grid continues to expand and the level of intelligence increases, the number of substations, power grid equipment, and alarm information connected to the control agency for centralized monitoring continues to increase. When faced with massive power grid operation alarm information, monitoring personnel are prone to misjudgment or omission of important information. Summary of the Invention

[0003] The embodiments of the present invention provide a method, apparatus, terminal device and computer-readable storage medium for identifying power grid fault events based on data mining, which can improve the efficiency of event identification and thereby solve the problem in the prior art that monitoring personnel are prone to misjudging or missing important information when faced with massive power grid operation alarm information.

[0004] An embodiment of the present invention provides a method for identifying power grid fault events based on data mining, comprising:

[0005] Obtain a set of transactions in a pending time period; wherein each transaction includes: a number of operation alarm information;

[0006] Perform association rule mining on all operational alarm information to identify several alarm information combinations;

[0007] For each alarm information combination, extract the features of each operation alarm information in the alarm information combination to obtain the feature vector of each operation alarm information in the alarm information combination;

[0008] The feature vector of each operation alarm information in the alarm information combination is input into the trained fault event recognition model, so that the fault event recognition model generates the power grid fault event corresponding to the alarm information combination according to the feature vector of each operation alarm information in the alarm information combination, and performs fault recognition on the power grid.

[0009] Furthermore, feature extraction is performed on each operation alarm information in the alarm information combination to obtain a feature vector of each operation alarm information in the alarm information combination, including:

[0010] Cluster all operation alarm information to obtain several alarm categories;

[0011] For each operation alarm message in the alarm information combination, extract the key features used to characterize the system operation status and obtain the key feature vector of each operation alarm message; the key features include: voltage fluctuation frequency, current peak value, number of circuit breaker operations, and number of protection device triggering;

[0012] Normalize the key feature vector of each operation alarm information to obtain a standardized key feature vector;

[0013] Determine the covariance matrix based on the normalized key eigenvectors of each operational alarm information;

[0014] Perform principal component analysis on the covariance matrix to obtain the reconstruction matrix;

[0015] The key eigenvector of each operation alarm information is projected based on the reconstruction matrix to obtain the final eigenvector of each operation alarm information in the alarm information combination.

[0016] Furthermore, the alarm categories include: equipment alarm category, frequency alarm category, voltage alarm category, and load alarm category.

[0017] Furthermore, before clustering all the operation alarm information, the following steps are also included:

[0018] All operation alarm information is cleaned, denoised and format converted to obtain pre-processed operation alarm information.

[0019] Furthermore, association rules are mined for all operational alarm information to identify several alarm information combinations, including:

[0020] For each alarm category, all operational alarm information in the alarm category is mined through association rules to identify several internal alarm information combinations;

[0021] Select the operation alarm information corresponding to the cluster center of each alarm category as the representative information;

[0022] Perform association rule mining on all representative information to identify several representative information combinations;

[0023] Determine the corresponding alarm category combination based on a combination of several representative information;

[0024] For each alarm category combination, association rules are mined between each operation alarm information of different alarm categories in the alarm category combination to identify several cross-category alarm information combinations;

[0025] All internal alarm information combinations and all cross-category alarm information combinations are used as the final alarm information combination.

[0026] Furthermore, after generating the power grid fault event corresponding to the alarm information combination, the following is further included:

[0027] Based on all alarm information combinations and corresponding power grid fault events, a knowledge graph is constructed so that when the power grid fault event identification process is subsequently executed, the power grid fault event corresponding to the operation alarm information can be directly determined based on the knowledge graph;

[0028] According to the knowledge graph, an event subgraph of each power grid fault event is constructed to determine the impact scope of the identified power grid fault event based on the event subgraph corresponding to the identified power grid fault event.

[0029] Furthermore, the fault event identification model is determined in the following way:

[0030] Acquire a number of training samples; each training sample includes: a number of feature vectors of operation alarm information and a corresponding actual fault event;

[0031] Several training samples are input into the fault event recognition model to be trained, so that the fault event recognition model to be trained takes the feature vectors of several operation alarm information in a training sample as input and the predicted fault events corresponding to the training sample as output for training. During the training process, the loss function is calculated according to the predicted fault events of the training samples and the corresponding actual fault events, and the network parameters of the fault event recognition model are adjusted according to the loss function until the loss function converges to obtain a trained fault event recognition model.

[0032] Based on the above method embodiment, the present invention provides a corresponding device embodiment, including: an alarm information acquisition module, an alarm information association module, a feature extraction module and a fault event identification module;

[0033] The alarm information acquisition module is used to obtain a set of transactions in a waiting period; wherein each transaction includes: a number of operation alarm information;

[0034] The alarm information association module is used to mine association rules for all operation alarm information and identify several alarm information combinations;

[0035] A feature extraction module is used to extract features of each operation alarm information in each alarm information combination to obtain a feature vector of each operation alarm information in the alarm information combination;

[0036] The fault event recognition module is used to input the feature vector of each operation alarm information in the alarm information combination into the trained fault event recognition model, so that the fault event recognition model generates the power grid fault event corresponding to the alarm information combination according to the feature vector of each operation alarm information in the alarm information combination, and performs fault recognition on the power grid.

[0037] Based on the above-mentioned method embodiment, the present invention provides a corresponding terminal device embodiment, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps of the power grid fault event identification method based on data mining as described in the present invention.

[0038] Based on the above-mentioned method embodiment, the present invention provides a corresponding computer-readable storage medium embodiment, including: a stored computer program, which, when the computer program is running, controls the device where the computer-readable storage medium is located to execute the steps of the power grid fault event identification method based on data mining as described in the present invention.

[0039] Compared with the prior art, the beneficial effects of the embodiment of this solution are:

[0040] The present invention obtains a set of transactions under a waiting period, wherein each transaction includes a number of operation alarm information, performs association rule mining on all operation alarm information, and discovers potential associations between the operation alarm information. By identifying these alarm information combinations, it is possible to more comprehensively understand the relationships and patterns between the operation alarm information, thereby reducing misjudgments and missed judgments caused by information fragmentation. For each alarm information combination, feature extraction is performed to obtain the feature vectors of each operation alarm information in the alarm information combination. By extracting features, the subtle differences and similarities between the operation alarm information can be better captured. Finally, the feature vectors of each operation alarm information in the alarm information combination are input into a trained fault event recognition model. Based on the model analysis of the feature vectors, the corresponding power grid fault events are accurately matched and generated, reducing the subjectivity and uncertainty of manual judgment, while improving the accuracy and efficiency of event recognition. The present invention not only improves the accuracy and efficiency of event recognition through the above-mentioned automated event recognition process, but also significantly reduces manual misjudgments or missed judgments. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 1 is a flow chart of a method for identifying power grid fault events based on data mining provided by an embodiment of the present invention;

[0042] Figure 2 It is a structural diagram of a power grid fault event identification device based on data mining provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0044] like Figure 1 As shown, in order to solve the problem in the prior art that monitoring personnel are prone to misjudging or missing important information when faced with massive power grid operation alarm information, an embodiment of the present invention provides a power grid fault event identification method based on data mining, which includes at least the following steps:

[0045] Step S1: Obtain a set of transactions in a pending time period; wherein each transaction includes: a number of operation alarm information;

[0046] For step S1, the transaction set D and the operation alarm information in the pending period T are obtained. The pending period includes several periods, that is, T = {T1, T2, T3, ..., T n}, each transaction is the operation alarm information received in different time periods, that is, D = {D(T1), D(T2), D(T3), ..., D(T n )}. In this embodiment, the operation alarm information includes: system short circuit, system overtemperature, system overload or incorrect switch operation; for example: D(T1)={system short circuit, system overtemperature} indicates that the operation alarm information of system short circuit and system overtemperature occurs in the T1 time period, D(T2)={system overload, incorrect switch operation} indicates that the operation alarm information of system overload and incorrect switch operation occurs in the T2 time period, D(T3)={system overtemperature, incorrect switch operation} indicates that the operation alarm information of system overtemperature and incorrect switch operation occurs in the T3 time period, ..., D(T n )={system short circuit, system over temperature, wrong switch operation} represents T n Operation alarm information such as system short circuit, system overtemperature and incorrect switch operation appears in the time period.

[0047] It should be noted that "several" is a quantifier used to indicate an unspecified number, typically referring to two or more (including two). In this embodiment, the number of operation alarm information included in each transaction can be one or more, and must be an integer. These operation alarm information include, but are not limited to, system short circuit, system overtemperature, system overload, or improper switch operation.

[0048] Step S2: Perform association rule mining on all operation alarm information to identify several alarm information combinations;

[0049] In a preferred embodiment, association rule mining is performed on all operation alarm information to identify several alarm information combinations, including:

[0050] Cluster all operation alarm information to obtain several alarm categories;

[0051] For each alarm category, all operational alarm information in the alarm category is mined through association rules to identify several internal alarm information combinations;

[0052] Select the operation alarm information corresponding to the cluster center of each alarm category as the representative information;

[0053] Perform association rule mining on all representative information to identify several representative information combinations;

[0054] Determine the corresponding alarm category combination based on a combination of several representative information;

[0055] For each alarm category combination, association rules are mined between each operation alarm information of different alarm categories in the alarm category combination to identify several cross-category alarm information combinations;

[0056] All internal alarm information combinations and all cross-category alarm information combinations are used as the final alarm information combination.

[0057] In a preferred embodiment, the alarm categories include: equipment alarm category, frequency alarm category, voltage alarm category, and load alarm category.

[0058] In step S2, since there may be a massive amount of alarm information throughout the entire power grid, directly mining association rules for all operational alarm information would result in excessive computational complexity and low efficiency. Therefore, clustering is used to group alarm information of the same type. Mining association rules within a group can reveal specific internal patterns, while cross-cluster analysis may reveal dependencies or chain reactions between information. Furthermore, clustering significantly reduces the amount of data within each group. Since each category only contains operational alarm information of similar types, association rule mining can process data more efficiently, reduce computing resource consumption, and improve mining efficiency.

[0059] In this embodiment, the K-Means clustering method is adopted, and the number of cluster centers K is selected by the elbow method to classify all operation alarm information into K major categories, such as equipment alarm category, frequency alarm category, voltage alarm category, and load alarm category.

[0060] Then, based on the cluster classification results, association rule mining is performed on the frequent patterns of operation alarm information within or between clusters to find the potential associations between the operation alarm information.

[0061] Specifically, for the association mining within the cluster, association rule mining is performed on all the operation alarm information in each alarm category. In this embodiment, the Apriori algorithm and the FP-Growth algorithm are used to perform association rule mining. In the application of the Apriori algorithm, frequent 1-item sets are first found, that is, those operation alarm information whose individual occurrence times exceed the preset threshold. Then, frequent item sets are searched layer by layer, and longer frequent item sets are gradually generated until no more item sets that meet the conditions can be found. In the process of generating frequent item sets, support (that is, the frequency of item sets appearing in all transactions) is used as a screening criterion to ensure that only those truly frequent item sets are retained. Finally, several strongly associated internal alarm information combinations are extracted from these frequent item sets. The combination represents the potential association between the various operation alarm information within the alarm category; in comparison, the FP-Growth algorithm provides a more efficient method to mine frequent item sets. Unlike the Apriori algorithm, the FP-Growth algorithm avoids the problem of frequent database scanning. It stores compressed transaction information by constructing an FP tree data structure. The FP tree uses frequent items as nodes, and the connections between nodes represent the co-occurrence relationship between items. After constructing the FP tree, the algorithm mines frequent item sets by traversing the tree. Since the FP tree has compressed and reorganized the transaction information, this process is more efficient than searching for frequent item sets layer by layer. Ultimately, the FP-Growth algorithm can also extract strongly associated operation alarm information combinations in each alarm category.

[0062] For association mining between clusters, within the clustering results for each alarm category, the operational alarm information corresponding to the cluster center is selected as the representative information for that alarm category. The cluster center typically represents the core features of the information within that cluster, so selecting the alarm information corresponding to the cluster center can more accurately reflect the main characteristics of that alarm category. Next, the selected representative information set is used as input, and an association rule mining algorithm (also using the Apriori algorithm or FP-Growth algorithm) is applied to identify several representative information combinations. These combinations represent potential associations between different alarm categories. For these identified associated representative information combinations, their corresponding alarm category combinations can be further traced, thereby performing cross-category association mining on operational alarm information for these associated alarm categories. This means that association rule mining only needs to be performed on associated alarm category combinations, without having to process the associations between all pairs of operational alarm information, greatly reducing the amount of computation.

[0063] After determining the alarm category combination, association rule mining algorithms (Apriori or FP-Growth) are used to mine association rules between each operational alarm message of different alarm categories within each alarm category combination, identifying several cross-category alarm message combinations. For example, in a power grid, voltage anomalies fall into the voltage alarm category, while current overloads fall into the load alarm category. By analyzing the operational alarm data for these two alarm categories, it is possible to identify the frequent simultaneous occurrence of voltage anomalies and current overloads. Therefore, these two operational alarm messages can be considered a strongly correlated alarm message combination. This allows for further analysis of their relationship with power grid events, helping operations and maintenance personnel take preventative measures in advance.

[0064] Finally, all internal alarm information combinations and all cross-category alarm information combinations are used as the final alarm information combination.

[0065] Preferably, before clustering all the operation alarm information, the method further includes:

[0066] All operation alarm information is cleaned, denoised and format converted to obtain pre-processed operation alarm information.

[0067] Specifically, to improve data quality, accuracy, and processability, all operational alarm information undergoes data cleaning, data denoising, and format conversion. Data cleaning involves removing abnormal data from alarm information, including missing values, outliers, and duplicate values. Missing values ​​refer to missing values ​​of parameters like current due to data transmission interruptions or sensor failures. Outliers refer to excessively high or low temperatures, voltages, and currents caused by data entry errors, measurement errors, data processing errors, or real-world extremes. Duplicate values ​​refer to repeated recordings of switch alarm signals.

[0068] In power grid alarm information, denoising is an important step to improve data quality, which can be achieved through a variety of methods. Specifically, statistical methods such as mean filtering, median filtering and standard deviation filtering identify and remove outliers and noise by calculating the statistical characteristics of the data; machine learning methods such as decision trees, Bayesian decision trees and autoencoders identify and filter noisy data by training models; association analysis methods identify irrelevant or erroneous alarm information by associating alarm information with other data sources such as maintenance records or telemetry values; signal processing methods such as wavelet transform denoising and filter technology are suitable for processing noise in time series data, and remove noise components by decomposing and reconstructing signals. The present invention can combine these denoising methods according to the specific circumstances of the alarm information to effectively reduce the noise in the power grid alarm information, improve the accuracy and reliability of the data, and thus provide more reliable data support for fault detection and accident handling of the power grid.

[0069] Finally, the alarm information data of different sources and formats are converted into a unified format. Specifically, the data of different formats (such as strings, integers, floating-point numbers, etc.) are converted into a unified data type, and the redundant spaces in the text data are removed.

[0070] Step S3: For each alarm information combination, extract the features of each operation alarm information in the alarm information combination to obtain the feature vector of each operation alarm information in the alarm information combination;

[0071] In a preferred embodiment, feature extraction is performed on each operation alarm information in the alarm information combination to obtain a feature vector of each operation alarm information in the alarm information combination, including:

[0072] For each operation alarm message in the alarm information combination, extract the key features used to characterize the system operation status and obtain the key feature vector of each operation alarm message; the key features include: voltage fluctuation frequency, current peak value, number of circuit breaker operations, and number of protection device triggering;

[0073] Normalize the key feature vector of each operation alarm information to obtain a standardized key feature vector;

[0074] Determine the covariance matrix based on the normalized key eigenvectors of each operational alarm information;

[0075] Perform principal component analysis on the covariance matrix to obtain the reconstruction matrix;

[0076] The key eigenvector of each operation alarm information is projected based on the reconstruction matrix to obtain the final eigenvector of each operation alarm information in the alarm information combination.

[0077] For step S3, the key features of the system operating status are extracted from the operation alarm information, and these features are converted into feature vectors for subsequent analysis and processing. Specifically, feature extraction algorithms such as principal component analysis (PCA) or neighborhood attribute importance algorithm are used to extract key features representing the system operating status from the alarm data. For example, in the power grid system, key features include voltage fluctuation frequency, current peak value, number of circuit breaker operations, and number of protection device triggering. These features can reflect the stability of the power grid and the operating status of the equipment, and help identify potential fault risks and abnormal events. By extracting these key features, important data support can be provided for fault diagnosis and prevention of the power grid.

[0078] The extracted key feature vectors may have different dimensions and numerical ranges, which will affect subsequent analysis and processing. Therefore, it is necessary to standardize the key feature vectors of each operational alarm information to have a unified dimension and numerical range. This not only improves the convergence speed and accuracy of the algorithm, but also prevents certain features from having too large a weight in subsequent calculations.

[0079] After standardization, the covariance matrix is ​​determined based on the standardized key eigenvectors of each operational alarm. The covariance matrix reflects the correlation and linear relationship between eigenvectors. Principal component analysis (PCA) is performed on the covariance matrix to obtain a reconstructed matrix. PCA maps the original features into a new low-dimensional space through linear transformation. These new features (i.e., principal components) are linear combinations of the original features and are mutually independent. By selecting principal components with the highest variance contribution, the data structure can be simplified while retaining important information.

[0080] Step S4: Input the feature vector of each operation alarm information in the alarm information combination into the trained fault event recognition model, so that the fault event recognition model generates a power grid fault event corresponding to the alarm information combination according to the feature vector of each operation alarm information in the alarm information combination, and performs fault recognition on the power grid.

[0081] For step S4, the feature vector of each operation alarm information in the alarm information combination obtained in step S3 is input into the trained fault event recognition model. The model uses advanced machine learning algorithms, such as support vector machines (SVM), neural networks or decision trees, etc., which can automatically learn and recognize different fault event patterns based on the input feature vectors.

[0082] Preferably, the fault event identification model is determined by:

[0083] Acquire a number of training samples; each training sample includes: a number of feature vectors of operation alarm information and a corresponding actual fault event;

[0084] Several training samples are input into the fault event recognition model to be trained, so that the fault event recognition model to be trained takes the feature vectors of several operation alarm information in a training sample as input and the predicted fault events corresponding to the training sample as output for training. During the training process, the loss function is calculated according to the predicted fault events of the training samples and the corresponding actual fault events, and the network parameters of the fault event recognition model are adjusted according to the loss function until the loss function converges to obtain a trained fault event recognition model.

[0085] Specifically, during the model training phase, a large amount of historical alarm data and known fault event labels are used for training, enabling the model to learn the mapping relationship between feature vectors and fault events. This allows the model to quickly and accurately generate corresponding power grid fault events based on the input alarm information feature vectors when it is put into practical use.

[0086] This invention uses an association rule mining algorithm to identify strongly correlated alarm combinations. It then converts each operational alarm in these combinations into feature vectors, which are then fed into a fault event recognition model. The model then activates its powerful recognition mechanism to quickly and accurately identify the corresponding fault event based on the input feature vectors. This automated event recognition process demonstrates significant advantages over traditional manual recognition methods, significantly improving the accuracy and efficiency of event recognition while also effectively reducing misjudgments and missed detections due to human error.

[0087] In addition, in order to further improve the intelligent level of power grid operation and maintenance and the efficiency of fault handling, preferably, after generating the power grid fault event corresponding to the alarm information combination, the following is also included:

[0088] Based on all alarm information combinations and corresponding power grid fault events, a knowledge graph is constructed so that when the power grid fault event identification process is subsequently executed, the power grid fault event corresponding to the operation alarm information can be directly determined based on the knowledge graph;

[0089] According to the knowledge graph, an event subgraph of each power grid fault event is constructed to determine the impact scope of the identified power grid fault event based on the event subgraph corresponding to the identified power grid fault event.

[0090] Specifically, after generating grid fault events corresponding to alarm combinations, a detailed knowledge graph is constructed based on all alarm combinations and their corresponding grid fault events for easy storage and retrieval. Furthermore, during the subsequent grid fault event identification process, the grid fault events corresponding to the operational alarms can be directly determined based on the knowledge graph. This knowledge graph not only encompasses the associations between alarms and fault events, but also encompasses complex relationships between fault events, such as causal relationships and temporal sequences. Through the knowledge graph's visualization and intelligent analysis capabilities, operations and maintenance personnel can more intuitively understand the grid's operational status and fault conditions, enabling them to make faster and more accurate fault handling decisions. For example, if the fault event identification model identifies a "system short circuit alarm + system overtemperature alarm" as a "transformer overheating fault," a rule can be generated: "If both a system short circuit alarm and a system overtemperature alarm occur simultaneously, the transformer overheating fault is inferred."

[0091] By building a comprehensive knowledge graph through a large number of alarm information combinations and corresponding power grid fault events, operators can make full use of the rule relationships therein to directly deduce the current operating status of the power grid, potential fault risks, and the response measures to be taken, thereby quickly accessing and applying them in the event reasoning process, providing strong decision-making support for operators' judgment.

[0092] Next, based on the knowledge graph, an event subgraph is constructed for each power grid fault event. The event subgraph is the part related to the specific power grid fault event extracted from the knowledge graph. By constructing the event subgraph, the impact scope of the fault can be quickly confirmed when a power grid fault occurs, providing strong support for subsequent fault detection and repair work.

[0093] It should be noted that in the present invention, the constructed knowledge graph or the event subgraph of each power grid fault event can be used for power grid fault event identification, that is, based on the obtained operation alarm information in the time period to be processed, the corresponding power grid fault event can be determined through the constructed knowledge graph or the event subgraph of each power grid fault event, thereby improving the reasoning efficiency and accuracy, and supporting the system's emergency response and decision-making.

[0094] like Figure 2 As shown, based on the above method embodiment, a corresponding device embodiment is provided;

[0095] An embodiment of the present invention provides a power grid fault event identification device based on data mining, comprising: an alarm information acquisition module, an alarm information association module, a feature extraction module, and a fault event identification module;

[0096] The alarm information acquisition module is used to obtain a set of transactions in a waiting period; wherein each transaction includes: a number of operation alarm information;

[0097] The alarm information association module is used to mine association rules for all operation alarm information and identify several alarm information combinations;

[0098] A feature extraction module is used to extract features of each operation alarm information in each alarm information combination to obtain a feature vector of each operation alarm information in the alarm information combination;

[0099] The fault event recognition module is used to input the feature vector of each operation alarm information in the alarm information combination into the trained fault event recognition model, so that the fault event recognition model generates the power grid fault event corresponding to the alarm information combination according to the feature vector of each operation alarm information in the alarm information combination, and performs fault recognition on the power grid.

[0100] It can be understood that the above-mentioned device embodiment corresponds to the method embodiment of the present invention, which can implement the data mining-based power grid fault event identification method provided by any of the above-mentioned method embodiments of the present invention.

[0101] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. Furthermore, in the drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which may be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement the present invention without inventive effort.

[0102] Based on the above-mentioned embodiment of the method for identifying power grid fault events based on data mining, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for identifying power grid fault events based on data mining according to any embodiment of the present invention is implemented.

[0103] For example, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more module elements may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0104] The terminal device may be a computing device such as a desktop computer, a notebook computer, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0105] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0106] Based on the above method embodiment, another embodiment is provided: another embodiment of the present invention provides a computer-readable storage medium, including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the data mining-based power grid fault event identification method described in any one of the above method embodiments of the present invention.

[0107] In particular, if the module / unit integrated into the data mining-based power grid fault event identification device / terminal equipment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. In particular, the computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.

[0108] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying power grid fault events based on data mining, characterized in that: include: Obtain a set of transactions in a pending time period; wherein each transaction includes: a number of operation alarm information; Perform association rule mining on all operational alarm information to identify several alarm information combinations; For each alarm information combination, extract the features of each operation alarm information in the alarm information combination to obtain the feature vector of each operation alarm information in the alarm information combination; The feature vector of each operation alarm information in the alarm information combination is input into the trained fault event recognition model, so that the fault event recognition model generates a power grid fault event corresponding to the alarm information combination according to the feature vector of each operation alarm information in the alarm information combination, and performs fault recognition on the power grid.

2. The method for identifying power grid fault events based on data mining according to claim 1, characterized in that: Perform association rule mining on all operational alarm information and identify several alarm information combinations, including: Cluster all operation alarm information to obtain several alarm categories; For each alarm category, all operational alarm information in the alarm category is mined through association rules to identify several internal alarm information combinations; Select the operation alarm information corresponding to the cluster center of each alarm category as the representative information; Perform association rule mining on all representative information to identify several representative information combinations; Determining a corresponding alarm category combination based on the plurality of representative information combinations; For each alarm category combination, association rules are mined between each operation alarm information of different alarm categories in the alarm category combination to identify several cross-category alarm information combinations; All internal alarm information combinations and all cross-category alarm information combinations are used as the final alarm information combination.

3. The method for identifying power grid fault events based on data mining according to claim 2, characterized in that: The alarm categories include: equipment alarm category, frequency alarm category, voltage alarm category and load alarm category.

4. The method for identifying power grid fault events based on data mining according to claim 3, characterized in that: Before clustering all operation alarm information, the following steps are also included: All operation alarm information is cleaned, denoised and format converted to obtain pre-processed operation alarm information.

5. The method for identifying power grid fault events based on data mining according to claim 1, characterized in that: Perform feature extraction on each operation alarm information in the alarm information combination to obtain a feature vector of each operation alarm information in the alarm information combination, including: For each operation alarm message in the alarm information combination, extract key features used to characterize the system operation status to obtain a key feature vector for each operation alarm message; the key features include: voltage fluctuation frequency, current peak value, number of circuit breaker operations, and number of protection device triggering times; Normalize the key feature vector of each operation alarm information to obtain a standardized key feature vector; Determine the covariance matrix based on the normalized key eigenvectors of each operational alarm information; Performing principal component analysis on the covariance matrix to obtain a reconstruction matrix; The key eigenvector of each operation alarm information is projected based on the reconstruction matrix to obtain the final eigenvector of each operation alarm information in the alarm information combination.

6. The method for identifying power grid fault events based on data mining according to claim 1, characterized in that: After generating the grid fault event corresponding to the alarm information combination, it also includes: Based on all alarm information combinations and corresponding power grid fault events, a knowledge graph is constructed so that when the power grid fault event identification process is subsequently executed, the power grid fault event corresponding to the operation alarm information can be directly determined based on the knowledge graph; According to the knowledge graph, an event subgraph of each power grid fault event is constructed to determine the impact scope of the identified power grid fault event based on the event subgraph corresponding to the identified power grid fault event.

7. The method for identifying power grid fault events based on data mining according to claim 1, characterized in that: The fault event identification model is determined in the following way: Acquire a number of training samples; each training sample includes: a number of feature vectors of operation alarm information and a corresponding actual fault event; Several training samples are input into the fault event recognition model to be trained, so that the fault event recognition model to be trained takes the feature vectors of several operation alarm information in a training sample as input and the predicted fault events corresponding to the training sample as output for training. During the training process, the loss function is calculated according to the predicted fault events of the training samples and the corresponding actual fault events, and the network parameters of the fault event recognition model are adjusted according to the loss function until the loss function converges to obtain a trained fault event recognition model.

8. A power grid fault event identification device based on data mining, characterized in that: include: Alarm information acquisition module, alarm information association module, feature extraction module and fault event identification module; The alarm information acquisition module is used to obtain a set of transactions in a waiting period; wherein each transaction includes: a number of operation alarm information; The alarm information association module is used to perform association rule mining on all operation alarm information and identify several alarm information combinations; The feature extraction module is used to extract features of each operation alarm information in each alarm information combination to obtain a feature vector of each operation alarm information in the alarm information combination; The fault event identification module is used to input the feature vector of each operation alarm information in the alarm information combination into a trained fault event identification model, so that the fault event identification model generates a power grid fault event corresponding to the alarm information combination based on the feature vector of each operation alarm information in the alarm information combination, and performs fault identification on the power grid.

9. A terminal device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for identifying power grid fault events based on data mining according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that include: A stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the power grid fault event identification method based on data mining according to any one of claims 1 to 7.

Images