Power grid fault handling method and system

Abstract

The invention provides a power grid fault handling method and system, and the method comprises the steps: obtaining multi-source tripping fault data based on a preset collection strategy; constructing a feature extraction model, and constructing a key tripping feature vector based on the feature extraction model and the multi-source tripping fault data; constructing an intelligent diagnosis model, and inputting the key tripping feature vector into the intelligent diagnosis model to obtain a primary tripping fault result; isolating the current tripping fault equipment based on the primary tripping fault result to obtain a secondary tripping fault result; and generating a fault handling strategy based on the secondary trip fault result, and sending the fault handling strategy to the operation and maintenance end to handle the power grid fault. According to the invention, the problem of poor operation stability of the power grid caused by fault omission or fault misjudgment due to the failure of accurately positioning the fault for a complex power grid structure in the prior art is solved. The method achieves the precise positioning of a fault condition, generates a fault processing strategy according to the fault condition, and transmits the strategy to the operation and maintenance end, thereby improving the operation stability of a power grid.

Description

Technical Field

[0001] The present invention relates to the field of power grid operation and maintenance, and in particular to a power grid fault handling method and system. Background Art

[0002] In the increasingly complex power grid environment, the design of power grid structure has also become complicated. When a power grid trips, how to quickly and accurately locate the fault and provide a solution has become a major technical challenge.

[0003] At present, the existing technology uses traditional machine learning models, which can only locate and process basic faults in low-voltage distribution networks once. When faced with a power grid with a large number of devices and complex connection relationships, faults are easily missed. Fault inspections are usually judged by different operators, and manual inspections are easily affected by subjective factors, resulting in fault misjudgments and affecting the efficiency of power grid fault recovery. It can be seen that the existing technology cannot accurately locate faults in the face of complex power grid structures, resulting in fault omissions or fault misjudgments, resulting in poor power grid operation stability. Summary of the Invention

[0004] In order to solve the above problems, the present invention proposes a power grid fault handling method and system, which can accurately locate the fault situation and generate a fault handling strategy based on the fault situation and send it to the operation and maintenance end, thereby improving the stability of power grid operation.

[0005] To achieve the above-mentioned objectives, an embodiment of the present invention provides a power grid fault handling method, comprising: acquiring multi-source tripping fault data based on a preset acquisition strategy; constructing a feature extraction model, and constructing a key tripping feature vector based on the feature extraction model and the multi-source tripping fault data; constructing an intelligent diagnosis model, inputting the key tripping feature vector into the intelligent diagnosis model, and obtaining a first-level tripping fault result; isolating the current tripping fault device based on the first-level tripping fault result, and obtaining a second-level tripping fault result; generating a fault handling strategy based on the second-level tripping fault result, and sending the fault handling strategy to the operation and maintenance end to handle the power grid fault.

[0006] An embodiment of the present invention proposes a power grid fault handling method, which generates multi-source tripping fault data with a preset acquisition strategy to ensure comprehensive acquisition of fault information, constructs a feature extraction model to perform feature extraction on the multi-source tripping fault data, eliminates redundant data to obtain key tripping feature vectors, and then constructs an intelligent diagnosis model to perform fault analysis on the key tripping feature vectors to obtain a first-level tripping fault result. The first-level tripping fault result is used to locate the current tripping fault device and perform isolation monitoring to obtain a second-level tripping fault result. Finally, a fault handling strategy is generated based on the second-level tripping fault result, and the fault handling strategy is sent to the operation and maintenance end to handle the power grid fault. In this way, by comprehensively collecting fault data, reducing redundant data through feature extraction, and performing two-level analysis on tripping faults, the cause of the fault is analyzed more comprehensively, the reliability of the fault handling strategy is guaranteed, and then the fault situation is accurately located and a fault handling strategy is generated based on the fault situation and sent to the operation and maintenance end, thereby improving the stability of power grid operation.

[0007] Furthermore, based on the preset acquisition strategy, multi-source tripping fault data is obtained, including: based on the preset acquisition strategy, obtaining data acquisition frequency and data acquisition priority; based on the data acquisition frequency and data acquisition priority, collecting protection device action information, fault recording data and multi-system related data; based on the protection device action information, fault recording data and multi-system related data, obtaining multi-source tripping fault data.

[0008] In the above scheme, a collection strategy is set to achieve comprehensive, timely and orderly collection of multi-source tripping fault data according to different data collection frequencies and data collection priorities, providing an accurate and reliable data basis for subsequent analysis of multi-source tripping fault data, so as to provide support for subsequent fault diagnosis and fault handling strategy generation, thereby achieving accurate positioning of fault conditions and generating fault handling strategies based on the fault conditions and sending them to the operation and maintenance end, thereby improving the stability of power grid operation.

[0009] Furthermore, a feature extraction model is constructed, and based on the feature extraction model and the multi-source tripping fault data, a key tripping feature vector is constructed, including: arranging each data in the multi-source tripping fault data based on a preset arrangement order, and generating an initial tripping feature vector according to the data arrangement result; constructing a feature extraction model based on the principal component analysis algorithm and the mutual information algorithm; inputting the initial tripping feature vector into the feature extraction model to obtain a key tripping feature vector corresponding to the initial tripping feature vector.

[0010] In the above scheme, multi-source tripping fault data are sorted and an initial tripping feature vector is constructed so that the initial tripping feature vector contains information related to various types of tripping faults. Then, combined with principal component analysis and mutual information algorithm, the most valuable features for tripping fault diagnosis and the essential features that can prominently reflect the tripping fault are screened out, providing data support for subsequent more accurate diagnosis of tripping faults, achieving accurate positioning of fault conditions and generating fault handling strategies based on the fault conditions and sending them to the operation and maintenance end, thereby improving the stability of power grid operation.

[0011] Furthermore, the initial tripping feature vector is input into the feature extraction model to obtain the key tripping feature vector corresponding to the initial tripping feature vector, including: based on the principal component analysis algorithm, the initial tripping feature vector is reduced in dimension, and the principal component is extracted from the reduced initial tripping feature vector to obtain several key feature components; based on the mutual information algorithm, the mutual information between several key feature components is calculated, and the key feature components that meet the preset value requirements are screened to obtain the key tripping feature vector.

[0012] In the above scheme, although the initial tripping feature vector contains various types of tripping fault information that can preliminarily reflect the fault characteristics, in order to make the subsequent fault diagnosis analysis and fault handling strategy formulation more accurate, the principal component analysis method is used to reduce the dimension of the high-dimensional initial tripping feature vector that may contain redundant information, and then retain the essential characteristics that can prominently reflect the tripping fault according to the mutual information algorithm. This provides data support for subsequent more accurate diagnosis of tripping faults, accurately locates the fault situation, and generates a fault handling strategy based on the fault situation and sends it to the operation and maintenance end, thereby improving the stability of power grid operation.

[0013] Furthermore, an intelligent diagnosis model is constructed, and the key tripping feature vector is input into the intelligent diagnosis model to obtain a first-level tripping fault result, including: constructing an intelligent diagnosis model based on a long short-term memory network and a support vector machine; inputting the key tripping feature vector into the intelligent diagnosis model, capturing the long-term data dependency of the key tripping feature vector through the long short-term memory network, and obtaining the output features of the long short-term memory network; inputting the output features of the long short-term memory network into the support vector machine, performing data classification on the output features of the long short-term memory network, and obtaining a first-level tripping fault result.

[0014] In the above scheme, a long short-term memory network is used to capture the long-term dependencies in the key tripping feature vector data, and an in-depth analysis of the change pattern of fault characteristics over time is conducted to provide support for the accurate extraction of key information related to power grid faults. Furthermore, a support vector machine is used to effectively distinguish the feature data of different fault types, thereby improving the accuracy and reliability of power grid fault diagnosis, facilitating more accurate analysis of complex power grid faults in the future, and quickly determining the type and location of the tripping fault to generate a fault handling strategy, thereby improving the stability of power grid operation.

[0015] Furthermore, the key tripping feature vector is input into the intelligent diagnosis model, and the long-term dependency of the data of the key tripping feature vector is captured through the long short-term memory network to obtain the output feature of the long short-term memory network, including: inputting the key tripping feature vector into the intelligent diagnosis model, eliminating the key tripping feature vector that does not meet the dependency requirements through the forgetting gate of the long short-term memory network to obtain the forgetting gate output feature; performing memory feature selection on the key tripping feature vector through the input gate of the long short-term memory network to obtain candidate memory units and input gate output features; updating the memory units of the long short-term memory network based on the forgetting gate output features, the candidate memory units and the input gate output features to obtain the memory units at the current moment; inputting the memory units at the current moment into the output gate of the long short-term memory network to obtain the output features of the long short-term memory network.

[0016] In the above scheme, the forget gate of the long short-term memory network is used to eliminate key feature vectors that do not meet the long-term dependency requirements, generate forget gate output features to filter out irrelevant information, then select memory features of the feature vector through the input gate, output candidate memory units and input gate output features, filter out key time series data, and then update the memory unit according to the forget gate output features, candidate memory units and input gate output features to obtain the current moment memory unit, retaining the long-term fault dependency, and finally input the memory unit into the output gate to generate the long short-term memory network output features. Thus, the change law of the fault characteristics over time is deeply analyzed, which provides support for the accurate extraction of key information related to the power grid fault, facilitates the subsequent more accurate analysis of complex power grid faults, quickly determines the type and location of the tripping fault to generate a fault handling strategy, and improves the stability of power grid operation.

[0017] Furthermore, the output features of the long short-term memory network are input into a support vector machine, and data classification is performed on the output features of the long short-term memory network to obtain a first-level tripping fault result, including: constructing an optimization problem and constraints based on the support vector machine; solving the optimization problem and setting corresponding fault labels for the output features of the long short-term memory network; and generating a first-level tripping fault result based on the fault label.

[0018] In the above scheme, support vector machines are used to classify the output features of the long short-term memory network, and optimization problems and constraints are constructed to find the optimal classification hyperplane. The characteristic data of different fault types are effectively distinguished, thereby accurately classifying the features output by the long short-term memory network and judging the fault label of the current tripping fault device. This significantly improves the accuracy and reliability of power grid fault diagnosis, accurately locates the fault situation, and generates a fault handling strategy based on the fault situation and sends it to the operation and maintenance end, thereby improving the stability of power grid operation.

[0019] Furthermore, based on the first-level tripping fault result, the current tripping fault device is isolated to obtain the second-level tripping fault result; including: based on the first-level tripping fault result, locating the current tripping fault device; collecting the operating parameters of the current tripping fault device, and isolating and monitoring the operating parameters of the current tripping fault device according to the preset Internet of Things technology to obtain the current tripping fault device structure data; based on the preset historical fault data and the current tripping fault device structure data, obtaining the fault type and fault location of the current tripping fault device to obtain the second-level tripping fault result.

[0020] In the above scheme, the first-level tripping fault result is used to quickly locate the current tripping fault device, and the current tripping fault device is isolated and monitored. The operating parameters of the current tripping fault device are analyzed to obtain the current tripping fault device structure data. Combined with the preset historical fault data, the specific type and location of the tripping fault are further analyzed to obtain a more accurate second-level tripping fault result. Thus, a two-level analysis of the tripping fault is performed, and the cause of the fault is analyzed more comprehensively to ensure the reliability of the fault handling strategy, thereby achieving accurate positioning of the fault situation and generating a fault handling strategy based on the fault situation and sending it to the operation and maintenance end, thereby improving the stability of power grid operation.

[0021] Furthermore, based on the secondary tripping fault result, a fault handling strategy is generated, and the fault handling strategy is sent to the operation and maintenance end to handle the power grid fault, including: based on the preset knowledge graph technology, the secondary tripping fault result is rule mapped to generate an intelligent decision-making knowledge base; based on the preset intelligent decision-making algorithm, the intelligent decision-making knowledge base is activated and the fault handling logic is constructed to generate a fault handling strategy, and the fault handling strategy is sent to the operation and maintenance end to handle the power grid fault.

[0022] In the above scheme, the preset knowledge graph technology is used to construct the rule mapping of the secondary tripping fault results to obtain the intelligent decision-making knowledge base, and then combined with the preset intelligent decision-making algorithm to activate the intelligent decision-making knowledge base and construct the fault handling logic, generate the fault handling strategy, and realize the automation from fault diagnosis to handling strategy sending, avoiding the subjective problems existing in manual decision-making, improving the efficiency and accuracy of fault handling, and realizing the precise positioning of fault conditions and generating fault handling strategies based on the fault conditions and sending them to the operation and maintenance end, thereby improving the stability of power grid operation.

[0023] An embodiment of the present invention also provides a power grid fault handling system, including: a fault data acquisition module, a feature extraction module, a first-level diagnosis module, a second-level diagnosis module and a fault handling module; the fault data acquisition module is used to obtain multi-source tripping fault data based on a preset acquisition strategy; the feature extraction module is used to construct a feature extraction model, and based on the feature extraction model and the multi-source tripping fault data, construct a key tripping feature vector; the first-level diagnosis module is used to construct an intelligent diagnosis model, input the key tripping feature vector into the intelligent diagnosis model, and obtain a first-level tripping fault result; the second-level diagnosis module is used to isolate the current tripping fault device based on the first-level tripping fault result, and obtain a second-level tripping fault result; the fault handling module is used to generate a fault handling strategy based on the second-level tripping fault result, and send the fault handling strategy to the operation and maintenance end to handle the power grid fault.

[0024] An embodiment of the present invention proposes a power grid fault handling system, in which a fault data acquisition module generates multi-source tripping fault data using a preset acquisition strategy to ensure comprehensive acquisition of fault information, a feature extraction module constructs a feature extraction model to perform feature extraction on the multi-source tripping fault data, and eliminates redundant data to obtain a key tripping feature vector, a first-level diagnosis module then constructs an intelligent diagnosis model to perform fault analysis on the key tripping feature vector to obtain a first-level tripping fault result, a second-level diagnosis module locates the current tripping fault device using the first-level tripping fault result and performs isolation monitoring to obtain a second-level tripping fault result, and finally a fault handling module generates a fault handling strategy based on the second-level tripping fault result and sends the fault handling strategy to the operation and maintenance end to handle the power grid fault. Thus, by comprehensively collecting fault data, reducing redundant data through feature extraction, and performing two-level analysis on the tripping fault, the cause of the fault is analyzed more comprehensively, the reliability of the fault handling strategy is guaranteed, and the fault situation is accurately located and a fault handling strategy is generated based on the fault situation and sent to the operation and maintenance end, thereby improving the stability of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic flow chart of the steps of a power grid fault handling method provided in one embodiment of the present invention;

[0026] Figure 2 A schematic diagram of the module structure of a power grid fault handling system provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0027] 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] In an embodiment of the present invention, in order to further illustrate the technical solution of the present invention, a power grid fault handling method proposed in an embodiment of the present invention is integrated into a secondary intelligent operation and maintenance integrated cockpit for execution, wherein the secondary intelligent operation and maintenance integrated cockpit is set as a collection point for information of multiple professional master stations. When a power grid fault occurs, a power grid fault handling method proposed in the present invention can be executed to comprehensively analyze and protect the tripping and action conditions of multiple professional master stations such as the professional master station, the safety professional master station, and the distribution network automation master station, and promptly determine the cause of the power grid fault and tripping; therefore, in this embodiment, the power grid fault handling method provided by the embodiment of the present invention will be explained by executing the secondary intelligent operation and maintenance integrated cockpit, which will not be repeated below. The specific implementation method is as follows:

[0029] Example 1

[0030] See also Figure 1 , Figure 1 A schematic flow chart of the steps of a power grid fault handling method provided in one embodiment of the present invention; Figure 1 As shown, the embodiment of the present invention provides a method for handling power grid faults, including steps 101 to 105, each of which is specifically as follows:

[0031] Step 101, acquiring multi-source trip fault data based on a preset acquisition strategy;

[0032] Step 102: construct a feature extraction model, and construct a key trip feature vector based on the feature extraction model and multi-source trip fault data;

[0033] Step 103: construct an intelligent diagnosis model, input the key tripping feature vector into the intelligent diagnosis model, and obtain a first-level tripping fault result;

[0034] Step 104: Isolate the current tripped fault device based on the primary trip fault result to obtain a secondary trip fault result;

[0035] Step 105: Generate a fault handling strategy based on the secondary trip fault result, and send the fault handling strategy to the operation and maintenance end to handle the power grid fault.

[0036] A specific implementation method is that the secondary intelligent operation and maintenance integrated cockpit collects and summarizes multi-source tripping fault data according to the data collection rules pre-set by the preset collection strategy, wherein the preset collection strategy includes the collection frequency and priority of the multi-source tripping fault data. The collection frequency refers to the data collection time interval. The reasonable setting of the collection frequency can ensure the timeliness of the data. The collection priority determines the collection order. Prioritizing the collection of key data can provide rapid support for tripping fault analysis. By clarifying the data collection frequency and priority, the collection strategy can ensure that the collected multi-source data is comprehensive, timely and orderly, which is conducive to the subsequent analysis and processing of power grid faults; multi-source tripping fault data includes protection device action information, fault recording data and multi-system related data. More specifically, the protection device action information includes protection action type, protection action time, protection action value, protection device number and related switch information. The protection device action information is the action signal generated by the protection device when the tripping fault occurs, which can be used to analyze the rationality of the protection action and quickly locate the tripping fault point and the affected area through the protection device number and related switch information; the fault recording data includes the waveform data of voltage and current before and after the tripping fault occurs, the tripping fault duration, waveform distortion characteristics, The timestamp and sampling frequency of the recording device, and the waveform data of voltage and current before and after the tripping fault are waveform data of the changes in electrical quantities in a period of time before and after the tripping fault occurs. They can intuitively present the changes in parameters such as voltage and current, and the duration of the tripping fault and the waveform distortion characteristics can assist in judging the severity of the tripping fault. The timestamp and sampling frequency of the recording device can ensure the accuracy and reliability of the data and provide an accurate basis for tripping fault analysis; multi-system related data include primary wiring diagrams, equipment ledgers, lightning location data, and real-time operation modes. Among them, the primary wiring diagram helps to understand the impact range of the tripping fault. Based on the relationship between the equipment and the power grid topology, the equipment record can assist in analyzing the historical status of the equipment and determine whether the tripping fault is caused by equipment aging. The lightning location data can be used to determine the correlation between lightning strikes and tripping faults. Based on the real-time operation mode and the real-time working conditions of the power grid, the cause of the tripping fault can be accurately analyzed, making the fault diagnosis more comprehensive and accurate, providing support for the formulation of subsequent disposal strategies, and improving the efficiency of power grid tripping fault processing and power supply stability. Then, a feature extraction model is constructed by fusing the principal component analysis algorithm and the mutual information algorithm. The feature extraction model is used to extract features from multi-source tripping fault data to obtain key tripping feature vectors.The intelligent diagnosis model is then constructed by integrating the long short-term memory (LSTM) network and the support vector machine (SVM). The LSTM network excels at processing time series data and can remember long-term dependencies. Through the collaborative work of the forget gate, input gate, memory unit, and output gate, it effectively processes and memorizes the input data. The support vector machine (SVM) is mainly used for classification and regression analysis. By finding an optimal classification hyperplane, different categories of data are separated. The key tripping feature vector is then input into the intelligent diagnosis model to obtain the first-level tripping fault result. Based on the first-level tripping fault result, the current tripping fault device is quickly located. The tripping fault device can be a switch body, cable, motor, etc. The current tripping fault device is isolated and monitored based on the preset Internet of Things technology to obtain the second-level tripping fault result. Finally, based on the second-level tripping fault result, a fault handling strategy is generated by combining knowledge graph technology and intelligent decision-making algorithms. The fault handling strategy is sent to the operation and maintenance end to handle the power grid fault.

[0037] An embodiment of the present invention proposes a power grid fault handling method, which generates multi-source tripping fault data with a preset acquisition strategy to ensure comprehensive acquisition of fault information, constructs a feature extraction model to perform feature extraction on the multi-source tripping fault data, eliminates redundant data to obtain key tripping feature vectors, and then constructs an intelligent diagnosis model to perform fault analysis on the key tripping feature vectors to obtain a first-level tripping fault result. The first-level tripping fault result is used to locate the current tripping fault device and perform isolation monitoring to obtain a second-level tripping fault result. Finally, a fault handling strategy is generated based on the second-level tripping fault result, and the fault handling strategy is sent to the operation and maintenance end to handle the power grid fault. In this way, by comprehensively collecting fault data, reducing redundant data through feature extraction, and performing two-level analysis on tripping faults, the cause of the fault is analyzed more comprehensively, the reliability of the fault handling strategy is guaranteed, and then the fault situation is accurately located and a fault handling strategy is generated based on the fault situation and sent to the operation and maintenance end, thereby improving the stability of power grid operation.

[0038] A preferred solution obtains multi-source tripping fault data based on a preset acquisition strategy, including: obtaining data acquisition frequency and data acquisition priority based on the preset acquisition strategy; collecting protection device action information, fault recording data and multi-system correlation data based on the data acquisition frequency and data acquisition priority; obtaining multi-source tripping fault data based on the protection device action information, fault recording data and multi-system correlation data.

[0039] In an implementable embodiment of a preferred solution, the secondary intelligent operation and maintenance integrated cockpit collects and summarizes multi-source tripping fault data such as protection device action information, fault recording data and multi-system related data according to the collection frequency and priority pre-set by the preset collection strategy, wherein the protection device action information includes protection action type, protection action time, protection action value, protection device number and related switch information, which is the action signal generated by the protection device in the event of a tripping fault; the fault recording data includes the waveform data of voltage and current before and after the tripping fault, the duration of the tripping fault, the waveform distortion characteristics, the timestamp and sampling frequency of the recording device, which is the waveform data of the electrical quantity changes in a period of time before and after the tripping fault occurs, which can intuitively present the changes in parameters such as voltage and current; the multi-system related data includes a primary wiring diagram, equipment ledger, lightning location data and a specific example of real-time operation mode, Taking the actual operation of a regional power grid, a 110kV line switch at a 220kV substation suddenly tripped, causing power outages in some areas, as an example, when a tripping fault occurred, the secondary intelligent operation and maintenance integrated cockpit collected protection device action information and fault recording data in real time through the high-speed communication network. The protection device action information indicates that the protection device has overcurrent protection action and the action time is 0.15 seconds. The fault recording data indicates that the A-phase current suddenly rises to 3 times the rated value and the voltage suddenly drops to 0.6 times the rated value. In addition, the secondary intelligent operation and maintenance integrated cockpit can automatically retrieve multi-system related data, including the primary wiring diagram of the tripped line, the lightning location system records of the day, and the equipment ledger. In the scenario of this embodiment, the lightning location system records of the day indicate that there was a lightning strike within 2 kilometers at the time of the tripping fault, and the equipment ledger indicates that the line cable has been in operation for 12 years and is close to the life limit.

[0040] In the above scheme, a collection strategy is set to achieve comprehensive, timely and orderly collection of multi-source tripping fault data according to different data collection frequencies and data collection priorities, providing an accurate and reliable data basis for subsequent analysis of multi-source tripping fault data, so as to provide support for subsequent fault diagnosis and fault handling strategy generation, thereby achieving accurate positioning of fault conditions and generating fault handling strategies based on the fault conditions and sending them to the operation and maintenance end, thereby improving the stability of power grid operation.

[0041] A preferred solution is to construct a feature extraction model and, based on the feature extraction model and multi-source trip fault data, construct a key trip feature vector, including: arranging each data in the multi-source trip fault data based on a preset arrangement order, and generating an initial trip feature vector according to the data arrangement result; constructing a feature extraction model based on a principal component analysis algorithm and a mutual information algorithm; inputting the initial trip feature vector into the feature extraction model to obtain a key trip feature vector corresponding to the initial trip feature vector.

[0042] In an implementation method of a preferred solution, before using multi-source trip fault data to construct a key trip feature vector, it is usually necessary to pre-process the multi-source trip fault data. For example, the multi-source trip fault data is input into a deep learning data denoising autoencoder to remove abnormal fluctuation data caused by electromagnetic interference in the multi-source data. Then, each data in the multi-source trip fault data is arranged in a preset arrangement order, and an initial trip feature vector is generated according to the data arrangement result. The initial trip feature vector contains various types of information related to the trip fault. Although the initial trip feature vector can fuse various types of trip fault information and preliminarily reflect the fault characteristics, its data dimension is high and may contain redundant information. Therefore, it can be used A feature extraction model that integrates the principal component analysis algorithm and the mutual information algorithm is used to process the initial tripping feature vector. The principal component analysis algorithm is a method for data dimensionality reduction, which can extract key features from multiple features and remove redundant information. The mutual information algorithm is used to measure the correlation between two variables. By calculating the mutual information between each key feature component, the most valuable features for tripping fault diagnosis are screened out. The feature extraction model is then used to extract features from the multi-source tripping fault data to obtain the key tripping feature vector. Compared with the initial tripping feature vector, the key tripping feature vector obtained after feature extraction can prominently reflect the essential characteristics of the tripping fault, facilitating a more accurate subsequent diagnosis of the tripping fault.

[0043] In the above scheme, multi-source tripping fault data are sorted and an initial tripping feature vector is constructed so that the initial tripping feature vector contains information related to various types of tripping faults. Then, combined with principal component analysis and mutual information algorithm, the most valuable features for tripping fault diagnosis and the essential features that can prominently reflect the tripping fault are screened out, providing data support for subsequent more accurate diagnosis of tripping faults, achieving accurate positioning of fault conditions and generating fault handling strategies based on the fault conditions and sending them to the operation and maintenance end, thereby improving the stability of power grid operation.

[0044] A preferred solution is to input the initial tripping feature vector into a feature extraction model to obtain a key tripping feature vector corresponding to the initial tripping feature vector, including: based on a principal component analysis algorithm, reducing the dimension of the initial tripping feature vector, and extracting the principal component of the reduced initial tripping feature vector to obtain several key feature components; based on a mutual information algorithm, calculating the mutual information between several key feature components, and screening the key feature components that meet the preset value requirements to obtain the key tripping feature vector.

[0045] An implementable method of a preferred scheme is explained by taking the scenario in which a 110kV line switch of a 220kV substation suddenly trips during the actual operation of a regional power grid, resulting in power outages in some areas as an example. The collected multi-source data are input into a data denoising autoencoder for preprocessing and arranged in a preset order. The initial tripping feature vector containing 12 parameters such as protection type, action time, waveform characteristics and equipment status is obtained by integration. The 12 parameters are then input into a feature extraction model. The initial tripping feature vector is subjected to dimensionality reduction processing by a principal component analysis algorithm to extract 8 key feature components. The 5 features with the strongest correlation with the tripping fault are then screened out by a mutual information algorithm to generate a key tripping feature vector focusing on the essence of the tripping fault.

[0046] In the above scheme, although the initial tripping feature vector contains various types of tripping fault information that can preliminarily reflect the fault characteristics, in order to make the subsequent fault diagnosis analysis and fault handling strategy formulation more accurate, the principal component analysis method is used to reduce the dimension of the high-dimensional initial tripping feature vector that may contain redundant information, and then retain the essential characteristics that can prominently reflect the tripping fault according to the mutual information algorithm. This provides data support for subsequent more accurate diagnosis of tripping faults, accurately locates the fault situation, and generates a fault handling strategy based on the fault situation and sends it to the operation and maintenance end, thereby improving the stability of power grid operation.

[0047] A preferred solution is to construct an intelligent diagnostic model, input the key tripping feature vector into the intelligent diagnostic model, and obtain a first-level tripping fault result, including: constructing an intelligent diagnostic model based on a long short-term memory network and a support vector machine; inputting the key tripping feature vector into the intelligent diagnostic model, capturing the long-term data dependency of the key tripping feature vector through the long short-term memory network, and obtaining the long short-term memory network output feature; inputting the long short-term memory network output feature into the support vector machine, performing data classification on the long short-term memory network output feature, and obtaining a first-level tripping fault result.

[0048] An implementable method of a preferred solution is to build an intelligent diagnosis model through long short-term memory network and support vector machine, and input the key trip feature vector into the intelligent diagnosis model. The long short-term memory network LSTM is good at processing time series data and can remember long-term dependencies. Through the collaborative work of forget gate, input gate, memory unit and output gate, the forget gate can selectively forget previous information to avoid interference from irrelevant historical information. The input gate can control the input of new information, the memory unit can save important long-term information, and the output gate can generate output according to the content of the memory unit and the current input, so as to effectively process and memorize the input data. The support vector machine SVM is mainly used for classification and regression analysis. By finding an optimal classification Hyperplane separates data of different categories. The support vector machine can accurately classify the features output by the long short-term memory network, determine the fault label of the current tripping fault device, and then input the key tripping feature vector into the intelligent diagnosis model to obtain the first-level tripping fault result. Taking the actual operation of a regional power grid, a 220kV substation suddenly tripped a 110kV line switch, resulting in power outages in some areas as an example, after inputting the key tripping feature vector into the intelligent diagnosis model, the intelligent diagnosis model combines the line operation mode and historical fault cases to quickly output the first-level tripping fault result, and preliminarily determines that the single-phase grounding fault is caused by aging of the cable insulation, and the tripping fault location is located in the interval between tower 3 and tower 4 of the line.

[0049] In the above scheme, a long short-term memory network is used to capture the long-term dependencies in the key tripping feature vector data, and an in-depth analysis of the change pattern of fault characteristics over time is conducted to provide support for the accurate extraction of key information related to power grid faults. Furthermore, a support vector machine is used to effectively distinguish the feature data of different fault types, thereby improving the accuracy and reliability of power grid fault diagnosis, facilitating more accurate analysis of complex power grid faults in the future, and quickly determining the type and location of the tripping fault to generate a fault handling strategy, thereby improving the stability of power grid operation.

[0050] A preferred solution is to input the key tripping feature vector into the intelligent diagnosis model, capture the long-term data dependency of the key tripping feature vector through a long short-term memory network, and obtain the output feature of the long short-term memory network, including: inputting the key tripping feature vector into the intelligent diagnosis model, eliminating the key tripping feature vector that does not meet the dependency requirements through the forgetting gate of the long short-term memory network, and obtaining the forgetting gate output feature; selecting the memory feature of the key tripping feature vector through the input gate of the long short-term memory network, and obtaining the candidate memory unit and the input gate output feature; updating the memory unit of the long short-term memory network based on the forgetting gate output feature, the candidate memory unit and the input gate output feature, and obtaining the memory unit at the current moment; inputting the memory unit at the current moment into the output gate of the long short-term memory network, and obtaining the output feature of the long short-term memory network.

[0051] A preferred embodiment of the present invention is to input the key tripping feature vector into the intelligent diagnosis model. First, the long-term dependency of the key tripping feature vector is captured through the long short-term memory network (LSTM). The long short-term memory network (LSTM) is good at processing time series data and can remember long-term dependencies. Through the collaborative work of the forget gate, input gate, memory unit, and output gate, the input data is effectively processed and memorized. The specific settings of the forget gate, input gate, memory unit, and output gate of the long short-term memory network are as follows: The calculation formula of the forget gate of the long short-term memory network is as follows:

[0052] g t =σ(W g ×[h t-1 ,y t ]+B g );

[0053] Where g t Represents the output feature of the forget gate at the current time t; W g Indicates the weight corresponding to the forget gate; h t-1 represents the hidden state at the previous moment t-1; y t represents the key tripping feature vector input at the current time t; [h t-1 ,y t ] represents the hidden state h t-1 and the key tripping characteristic vector y t The vector composed of B g represents the bias term corresponding to the forget gate; σ represents the activation function;

[0054] The calculation formula of the input gate of the long short-term memory network is as follows:

[0055] i t =σ(W i ×[h t-1 ,y t ]+B i );

[0056]

[0057] Where i t Represents the output features of the input gate at the current time t; represents the candidate memory unit at the current time t; W i and W C Represent the weights corresponding to the input gate and candidate memory unit respectively; h t-1 represents the hidden state at the previous moment t-1; y t represents the key tripping feature vector input at the current time t; [h t-1 ,y t] represents the hidden state h t-1 and the key tripping characteristic vector y t The vector composed of B i and B C Represent the bias terms corresponding to the input gate and the candidate memory unit respectively; σ represents the activation function; tanh represents the hyperbolic tangent function;

[0058] The calculation formula for the memory unit update of the long short-term memory network is as follows:

[0059]

[0060] Where C t Represents the memory unit at the current time t; g t Represents the output feature of the forget gate at the current time t; C t-1 Represents the memory unit of the previous moment t-1; i t Represents the output features of the input gate at the current time t; Represents the candidate memory unit at the current time t;

[0061] The calculation formula of the output gate of the long short-term memory network is as follows:

[0062] o t =σ(W o ×[h t-1 ,y t ]+B o );

[0063] h t =o t ×tanh(C t );

[0064] In the formula, o t represents the output feature of the output gate at the current time t; σ represents the activation function; W o Indicates the weight corresponding to the output gate; h t-1 represents the hidden state at the previous moment t-1; y t represents the key tripping feature vector input at the current time t; [h t-1 ,y t ] represents the hidden state h t-1 and the key tripping characteristic vector y t The vector composed of B o Represents the bias term corresponding to the output gate; h t represents the hidden state at the current time t; tanh represents the hyperbolic tangent function; C t Represents the memory unit at the current time t.

[0065] This demonstrates the interplay between the forget gate, input gate, memory unit, and output gate of the LSTM network. The forget gate selectively forgets previous information, preventing interference from irrelevant historical information; the input gate controls the input of new information; the memory unit stores important long-term information; and the output gate generates output based on the contents of the memory unit and the current input. This mechanism enables the LSTM network to better capture long-term dependencies in data, deeply analyze how fault characteristics change over time, and accurately extract key information related to power grid faults.

[0066] In the above scheme, the forget gate of the long short-term memory network is used to eliminate key feature vectors that do not meet the long-term dependency requirements, generate forget gate output features to filter out irrelevant information, then select memory features of the feature vector through the input gate, output candidate memory units and input gate output features, filter out key time series data, and then update the memory unit according to the forget gate output features, candidate memory units and input gate output features to obtain the current moment memory unit, retaining the long-term fault dependency, and finally input the memory unit into the output gate to generate the long short-term memory network output features. Thus, the change law of the fault characteristics over time is deeply analyzed, which provides support for the accurate extraction of key information related to the power grid fault, facilitates the subsequent more accurate analysis of complex power grid faults, quickly determines the type and location of the tripping fault to generate a fault handling strategy, and improves the stability of power grid operation.

[0067] A preferred solution is to input the output features of the long short-term memory network into a support vector machine, perform data classification on the output features of the long short-term memory network, and obtain a first-level tripping fault result, including: constructing an optimization problem and constraint conditions based on the support vector machine; solving the optimization problem and setting corresponding fault labels for the output features of the long short-term memory network; and generating a first-level tripping fault result based on the fault label.

[0068] One possible implementation method of a preferred solution is to use the output features of the long short-term memory network as the input features of the support vector machine, perform classification and regression analysis through the support vector machine (SVM), and find an optimal classification hyperplane to separate data of different categories. The specific calculation method is as follows: When the input features of the support vector machine are classified, the corresponding optimization problem is as follows:

[0069]

[0070] Where w u represents the weight vector of the u-th support vector machine, which is used to determine the direction of the classification hyperplane; b u represents the bias term of the u-th support vector machine, which is used to determine the distance between the classification hyperplane and the origin; ξ uvrepresents the slack variable when the vth input feature is linearly inseparable in the uth support vector machine; Q represents the regularization parameter, which is used to control the trade-off between classification interval and classification error; R represents the number of input features; min represents the minimum operation;

[0071] The constraints corresponding to the optimization problem are as follows:

[0072] p v (w u *o v +b u )≥1-ξ uv ,p v =u;

[0073] p v (w u *o v +b u )≤ξ uv -1,p v ≠u;

[0074]

[0075] Where p v Indicates the feature category label corresponding to the vth input feature; o v represents the vth input feature; w u represents the weight vector of the u-th support vector machine, which is used to determine the direction of the classification hyperplane; b u represents the bias term of the u-th support vector machine, which is used to determine the distance between the classification hyperplane and the origin; ξ uv represents the slack variable when the vth input feature is linearly inseparable in the uth support vector machine; R represents the number of input features;

[0076] The feature category label corresponding to the input feature is the fault label of the current trip fault device, and the first-level trip fault result is generated according to the fault label;

[0077] Therefore, in the process of supporting vector machines classifying the output features of the long short-term memory network, the characteristic data of different fault types can be effectively distinguished by finding the optimal classification hyperplane. Among them, the weight vector and bias term determine the direction and position of the classification hyperplane, the regularization parameter is used to balance the classification interval and classification error, and the slack variable is used to deal with linearly inseparable data.

[0078] In the above scheme, support vector machines are used to classify the output features of the long short-term memory network, and optimization problems and constraints are constructed to find the optimal classification hyperplane. The characteristic data of different fault types are effectively distinguished, thereby accurately classifying the features output by the long short-term memory network and judging the fault label of the current tripping fault device. This significantly improves the accuracy and reliability of power grid fault diagnosis, accurately locates the fault situation, and generates a fault handling strategy based on the fault situation and sends it to the operation and maintenance end, thereby improving the stability of power grid operation.

[0079] A preferred solution is to isolate the current tripping fault device based on the first-level tripping fault result to obtain the second-level tripping fault result; including: locating the current tripping fault device based on the first-level tripping fault result; collecting the operating parameters of the current tripping fault device, and isolating and monitoring the operating parameters of the current tripping fault device according to the preset Internet of Things technology to obtain the current tripping fault device structure data; based on the preset historical fault data and the current tripping fault device structure data, obtaining the fault type and fault location of the current tripping fault device to obtain the second-level tripping fault result.

[0080] An implementable method of a preferred solution is to obtain a first-level tripping fault result, quickly locate the current tripping fault device according to the first-level tripping fault result, collect the operating parameters of the current tripping fault device through the intelligent sensors pre-set on each device, such as temperature, pressure and current, and then realize data transmission between the intelligent sensors and the monitoring center through the Internet of Things technology, establish an information interaction channel to isolate and monitor the current tripping fault device, and then combine the structural characteristics of the current tripping fault device and historical fault data to further analyze and determine the specific type and location of the tripping fault, such as short circuit and open circuit, to obtain a more accurate second-level tripping fault result; wherein, the historical fault data is the data obtained by comprehensively analyzing the type, cause and treatment of the tripping fault of the current tripping fault device in the past. In this embodiment, in the actual operation of a regional power grid, a 110kV line switch of a 220kV substation suddenly tripped, causing part of the area Taking a power outage scenario as an example, after obtaining the first-level trip fault result of "preliminary determination of a single-phase ground fault caused by cable insulation aging, located between towers 3 and 4 on the line," the cable device is automatically identified as the current trip fault device. Smart sensors pre-installed on the cable body and connectors in this section collect operating parameters in real time and transmit them to a monitoring center via the IoT channel. The monitoring center then isolates and monitors the current trip fault device. Combined with the cable's structural design parameters, such as insulation thickness and installation method, and historical fault records, the center determines that the insulation failure rate for this type of cable has increased by 23% after 10 years of operation. Ultimately, the insulation resistance at the cable connector on tower 3 has dropped to 0.8 MΩ, below the normal threshold of 10 MΩ. This confirms that the trip fault type is insulation breakdown due to aging of the cable connector. A second-level trip fault result, including the specific fault point and damage level, is then generated.

[0081] In the above scheme, the first-level tripping fault result is used to quickly locate the current tripping fault device, and the current tripping fault device is isolated and monitored. The operating parameters of the current tripping fault device are analyzed to obtain the current tripping fault device structure data. Combined with the preset historical fault data, the specific type and location of the tripping fault are further analyzed to obtain a more accurate second-level tripping fault result. Thus, a two-level analysis of the tripping fault is performed, and the cause of the fault is analyzed more comprehensively to ensure the reliability of the fault handling strategy, thereby achieving accurate positioning of the fault situation and generating a fault handling strategy based on the fault situation and sending it to the operation and maintenance end, thereby improving the stability of power grid operation.

[0082] A preferred solution generates a fault handling strategy based on the secondary tripping fault result, and sends the fault handling strategy to the operation and maintenance end to handle the power grid fault, including: based on the preset knowledge graph technology, the secondary tripping fault result is rule-mapped to generate an intelligent decision-making knowledge base; based on the preset intelligent decision-making algorithm, the intelligent decision-making knowledge base is activated and the fault handling logic is constructed to generate a fault handling strategy, and the fault handling strategy is sent to the operation and maintenance end to handle the power grid fault.

[0083] In one preferred embodiment, after obtaining the secondary trip fault result, the secondary intelligent operation and maintenance integrated cockpit can perform a full-factor analysis of the secondary trip fault result, integrate knowledge in the fields of grid equipment association, historical fault cases, and handling rules through knowledge graph technology, construct a semantic network model of fault characteristics and handling strategies, and perform multi-dimensional mapping of fault phenomena with equipment ledgers and protection configuration files to form an intelligent decision-making knowledge base containing a fault cause reasoning chain, a handling rule set, and a resource demand matrix. Then, the intelligent decision-making algorithm can activate the handling logic in the intelligent decision-making knowledge base through the rule engine, perform multi-objective optimization operations based on the fault feature vector, and generate a structured fault handling strategy, that is, an actual fault solution, such as clarifying the fault isolation scope, generating maintenance step priorities, and scheduling operation and maintenance resources, and send the fault handling strategy to the operation and maintenance end to handle the grid fault. For example, the actual grid trip fault situation and the corresponding handling strategy can be sent to the operation and maintenance personnel in real time through the communication channel. More specifically, in the cable insulation aging fault scenario, instructions for isolating the fault section and replacing the joint insulation sleeve can be automatically generated, and the operation and maintenance personnel can be simultaneously notified to bring the corresponding materials to the site to ensure the actual operation of the power grid in a certain area. In the example, a 110kV line switch at a 220kV substation suddenly tripped, causing power outages in some areas. The result of the secondary trip fault was "the insulation resistance at the cable joint of the No. 3 tower dropped to 0.8MΩ, which is lower than the normal threshold of 10MΩ. It can be confirmed that the trip fault type is the aging and breakdown of the cable joint insulation." The secondary intelligent operation and maintenance integrated cockpit can use the knowledge graph technology to associate equipment maintenance strategies based on the secondary trip fault result, automatically generate a disposal plan, and immediately isolate the faulty cable section. The operation and maintenance team is dispatched to the scene with insulation repair materials, and it is recommended to repair the same Cables that have been in operation for more than 10 years are subjected to comprehensive insulation testing. In addition, after the operation and maintenance personnel arrive, they can use an infrared imager to remeasure the joint temperature to confirm whether there is a temperature abnormality, and replace the insulating sleeve in time to restore power supply. It is worth mentioning that the power grid fault handling method proposed in an embodiment of the present invention takes a total of 52 minutes from the occurrence of the fault to the restoration of power supply, which is nearly 2 hours shorter than the traditional manual troubleshooting method. It also accurately avoids misjudgment of device faults and significantly improves the power supply restoration efficiency under complex faults, further verifying the reliability and efficiency of the present invention in actual power grid fault handling.

[0084] In the above scheme, the preset knowledge graph technology is used to construct the rule mapping of the secondary tripping fault results to obtain the intelligent decision-making knowledge base, and then combined with the preset intelligent decision-making algorithm to activate the intelligent decision-making knowledge base and construct the fault handling logic, generate the fault handling strategy, and realize the automation from fault diagnosis to handling strategy sending, avoiding the subjective problems existing in manual decision-making, improving the efficiency and accuracy of fault handling, and realizing the precise positioning of fault conditions and generating fault handling strategies based on the fault conditions and sending them to the operation and maintenance end, thereby improving the stability of power grid operation.

[0085] Example 2

[0086] See also Figure 2 , Figure 2 A schematic diagram of the module structure of a power grid fault handling system provided by an embodiment of the present invention. Figure 2 As shown, an embodiment of the present invention further provides a power grid fault handling system, including: a fault data acquisition module 201, a feature extraction module 202, a first-level diagnosis module 203, a second-level diagnosis module 204 and a fault handling module 205; the fault data acquisition module 201 is used to obtain multi-source tripping fault data based on a preset acquisition strategy; the feature extraction module 202 is used to construct a feature extraction model, and construct a key tripping feature vector based on the feature extraction model and the multi-source tripping fault data; the first-level diagnosis module 203 is used to construct an intelligent diagnosis model, input the key tripping feature vector into the intelligent diagnosis model, and obtain a first-level tripping fault result; the second-level diagnosis module 204 is used to isolate the current tripping fault device based on the first-level tripping fault result, and obtain a second-level tripping fault result; the fault handling module 205 is used to generate a fault handling strategy based on the second-level tripping fault result, and send the fault handling strategy to the operation and maintenance end to handle the power grid fault.

[0087] A specific possible implementation method is to integrate a power grid fault handling system provided by an embodiment of the present invention into a secondary intelligent operation and maintenance integrated cockpit, and collect and summarize multi-source tripping fault data through the fault data collection module 201 according to the data collection rules pre-set by the preset collection strategy, wherein the preset collection strategy includes the collection frequency and priority of the multi-source tripping fault data, and the multi-source tripping fault data includes protection device action information, fault recording data and multi-system related data. More specifically, the protection device action information includes protection action type, protection action time, protection action value, protection device number and related switch information, which is the action signal generated by the protection device when a tripping fault occurs; the fault recording data includes the waveform data of voltage and current before and after the tripping fault occurs, the duration of the tripping fault, the waveform distortion characteristics, the timestamp and sampling frequency of the recording device, which is the waveform data of the electrical quantity changes in a period of time before and after the tripping fault occurs, which can intuitively present the changes in parameters such as voltage and current; the multi-system related data includes a primary wiring diagram, equipment ledger, lightning location data and real-time operation mode; then the feature extraction module 202 constructs a feature extraction model by fusing the principal component analysis algorithm and the mutual information algorithm The first-level diagnosis module 203 then integrates the long short-term memory network (LSTM) and the support vector machine (SVM) to construct an intelligent diagnosis model. The long short-term memory network (LSTM) is good at processing time series data and can remember long-term dependencies. The input data is effectively processed and memorized through the collaborative work of the forget gate, input gate, memory unit, and output gate. The support vector machine (SVM) is mainly used for classification and regression analysis. By finding an optimal classification hyperplane, data of different categories are separated. The key trip feature vector is then input into the intelligent diagnosis model to obtain a first-level trip fault result. The second-level diagnosis module 204 quickly locates the current trip fault device based on the first-level trip fault result. The trip fault device can be a switch body, cable, motor, etc., and isolates and monitors the current trip fault device based on the preset Internet of Things technology to obtain a second-level trip fault result. Finally, the fault handling module 205 generates a fault handling strategy based on the second-level trip fault result by combining knowledge graph technology and intelligent decision-making algorithm, and sends the fault handling strategy to the operation and maintenance end to handle the power grid fault.

[0088] An embodiment of the present invention proposes a power grid fault handling system, in which a fault data acquisition module generates multi-source tripping fault data using a preset acquisition strategy to ensure comprehensive acquisition of fault information, a feature extraction module constructs a feature extraction model to perform feature extraction on the multi-source tripping fault data, and eliminates redundant data to obtain a key tripping feature vector, a first-level diagnosis module then constructs an intelligent diagnosis model to perform fault analysis on the key tripping feature vector to obtain a first-level tripping fault result, a second-level diagnosis module locates the current tripping fault device using the first-level tripping fault result and performs isolation monitoring to obtain a second-level tripping fault result, and finally a fault handling module generates a fault handling strategy based on the second-level tripping fault result and sends the fault handling strategy to the operation and maintenance end to handle the power grid fault. Thus, by comprehensively collecting fault data, reducing redundant data through feature extraction, and performing two-level analysis on the tripping fault, the cause of the fault is analyzed more comprehensively, the reliability of the fault handling strategy is guaranteed, and the fault situation is accurately located and a fault handling strategy is generated based on the fault situation and sent to the operation and maintenance end, thereby improving the stability of power grid operation.

[0089] An embodiment of the present invention further proposes an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the steps of a power grid fault handling method proposed in an embodiment of the present invention; specifically, the electronic device comprises: a processor, a memory and a computer program; wherein the memory is used to store the computer program, and the memory may also be a flash memory; the computer program is, for example, an application program and a functional module that implements the above method; the processor is used to execute the computer program stored in the memory to implement the above method, in which the device executes each step of a power grid fault handling method proposed in an embodiment of the present invention, for details, please refer to the relevant description in the previous embodiment 1; optionally, the memory may be independent or integrated with the processor; when the memory is a device independent of the processor, the device may further include: a bus for connecting the memory and the processor.

[0090] An embodiment of the present invention further provides a readable storage medium, wherein the readable storage medium stores a computer program. When the computer program is executed by a processor, the computer program is used to implement the steps of the power grid fault handling method provided in the embodiment of the present invention. The readable storage medium may be a computer storage medium or a communication medium. The communication medium includes any medium that facilitates the transmission of a computer program from one location to another. The computer storage medium may be any available medium that can be accessed by a general-purpose or special-purpose computer. For example, the readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium may also be an integral part of the processor. The processor and the readable storage medium may be located in an application-specific integrated circuit (ASIC). In addition, the ASIC may be located in a user device. Of course, the processor and the readable storage medium may also exist as discrete components in a communication device. The readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0091] An embodiment of the present invention further provides a program product, which includes execution instructions stored in a readable storage medium. At least one processor of a device can read the execution instructions from the readable storage medium. At least one processor executes the execution instructions so that the device implements the method provided in Example 1 of the present invention. In the above-mentioned electronic device, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), and application-specific integrated circuits (ASICs). The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention can be directly implemented as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor.

[0092] The above is only 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 technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0093] In the description of this specification, the reference terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0094] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, features specified as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of this application, "plurality" means two or more, unless otherwise specifically defined.

Claims

1. A method for handling power grid faults, characterized in that: include: Based on the preset acquisition strategy, multi-source trip fault data is obtained; Constructing a feature extraction model, and constructing a key trip feature vector based on the feature extraction model and the multi-source trip fault data; Constructing an intelligent diagnosis model, inputting the key tripping feature vector into the intelligent diagnosis model, and obtaining a first-level tripping fault result; Isolating the current tripped fault device based on the primary tripping fault result to obtain a secondary tripping fault result; Based on the secondary trip fault result, a fault handling strategy is generated, and the fault handling strategy is sent to the operation and maintenance end to handle the power grid fault.

2. A method for handling power grid faults according to claim 1, characterized in that: The method of acquiring multi-source tripping fault data based on a preset acquisition strategy includes: Based on the preset collection strategy, obtaining data collection frequency and data collection priority; Based on the data collection frequency and data collection priority, collecting protection device action information, fault recording data and multi-system correlation data; Multi-source tripping fault data is obtained based on the protection device action information, fault recording data and multi-system correlation data.

3. A method for handling power grid faults according to claim 1, characterized in that: The constructing of a feature extraction model and constructing a key trip feature vector based on the feature extraction model and the multi-source trip fault data includes: Arranging each data in the multi-source trip fault data based on a preset arrangement order, and generating an initial trip feature vector according to the data arrangement result; Based on principal component analysis algorithm and mutual information algorithm, a feature extraction model is constructed; The initial tripping feature vector is input into the feature extraction model to obtain a key tripping feature vector corresponding to the initial tripping feature vector.

4. A method for handling power grid faults according to claim 3, characterized in that: Inputting the initial tripping feature vector into the feature extraction model to obtain a key tripping feature vector corresponding to the initial tripping feature vector includes: Based on the principal component analysis algorithm, the initial tripping feature vector is reduced in dimension, and the principal component extraction is performed on the initial tripping feature vector after the dimension reduction to obtain several key characteristic components; Based on the mutual information algorithm, the mutual information between the key characteristic components is calculated, and the key characteristic components that meet the preset value requirements are screened to obtain the key tripping characteristic vector.

5. A method for handling power grid faults according to claim 1, characterized in that: The constructing of the intelligent diagnosis model, inputting the key tripping feature vector into the intelligent diagnosis model, and obtaining a first-level tripping fault result includes: Build an intelligent diagnosis model based on long short-term memory network and support vector machine; Inputting the key tripping feature vector into the intelligent diagnosis model, capturing the long-term data dependency of the key tripping feature vector through the long short-term memory network, and obtaining the long short-term memory network output feature; The output features of the long short-term memory network are input into the support vector machine, and data classification is performed on the output features of the long short-term memory network to obtain a first-level tripping fault result.

6. A method for handling power grid faults according to claim 5, characterized in that: The key tripping feature vector is input into the intelligent diagnosis model, and the long-term data dependency of the key tripping feature vector is captured by the long short-term memory network to obtain the long short-term memory network output feature, including: Inputting the key tripping feature vector into the intelligent diagnosis model, eliminating the key tripping feature vector that does not meet the dependency requirements through the forget gate of the long short-term memory network, and obtaining a forget gate output feature; Performing memory feature selection on the key trip feature vector through the input gate of the long short-term memory network to obtain candidate memory units and input gate output features; Based on the forget gate output feature, the candidate memory unit and the input gate output feature, updating the memory unit of the long short-term memory network to obtain the current memory unit; The current memory unit is input into the output gate of the long short-term memory network to obtain the long short-term memory network output feature.

7. A method for handling power grid faults according to claim 5 or 6, characterized in that: Inputting the output features of the long short-term memory network into the support vector machine, performing data classification on the output features of the long short-term memory network, and obtaining a first-level trip fault result, including: Based on the support vector machine, construct an optimization problem and constraints; Solving the optimization problem and setting corresponding fault labels for the long short-term memory network output features; A primary trip fault result is generated based on the fault tag.

8. A method for handling power grid faults according to claim 1, characterized in that: Isolating the current tripped fault device based on the first-level tripping fault result to obtain the second-level tripping fault result; including: Based on the first-level trip fault result, locate the current trip fault device; Collecting the operating parameters of the currently tripped faulty device, and isolating and monitoring the operating parameters of the currently tripped faulty device according to the preset Internet of Things technology to obtain the structural data of the currently tripped faulty device; Based on the preset historical fault data and the current tripping fault device structure data, the fault type and fault location of the current tripping fault device are obtained to obtain a secondary tripping fault result.

9. A method for handling power grid faults according to claim 1, characterized in that: Based on the secondary trip fault result, a fault handling strategy is generated, and the fault handling strategy is sent to the operation and maintenance end to handle the power grid fault, including: Based on the preset knowledge graph technology, the secondary trip fault result is mapped to rules to generate an intelligent decision-making knowledge base; Based on a preset intelligent decision-making algorithm, the intelligent decision-making knowledge base is activated and a fault handling logic is constructed to generate a fault handling strategy, which is then sent to the operation and maintenance end to handle the power grid fault.

10. A power grid fault handling system, characterized in that: include: Fault data acquisition module, feature extraction module, primary diagnosis module, secondary diagnosis module and fault handling module; The fault data acquisition module is used to acquire multi-source trip fault data based on a preset acquisition strategy; The feature extraction module is used to construct a feature extraction model, and to construct a key trip feature vector based on the feature extraction model and the multi-source trip fault data; The primary diagnosis module is used to construct an intelligent diagnosis model, input the key tripping feature vector into the intelligent diagnosis model, and obtain a primary tripping fault result; The secondary diagnosis module is used to isolate the current trip fault device based on the primary trip fault result to obtain a secondary trip fault result; The fault handling module is used to generate a fault handling strategy based on the secondary tripping fault result, and send the fault handling strategy to the operation and maintenance end to handle the power grid fault.

Images