A Substation Equipment Fault Early Warning System and Method for Fusion Defect Detection

Abstract

The present invention discloses a substation equipment fault early warning system and method integrating defect detection, which relates to the technical field of fault early warning. The method includes: obtaining multiple operation modes of the substation equipment and performing mode switching, collecting multiple real-time operation data; through defect extraction, outputting multiple groups of defect feature vectors; identifying fault events according to the multiple groups of defect feature vectors, outputting matching fault events, and respectively calculating corresponding multiple risk indicators; fusing the multiple risk indicators, and generating a fault early warning signal according to the fused risk indicator for fault early warning. The present invention solves the technical problem that the existing substation equipment fault early warning methods only detect single operation modes or specific defects, and the comprehensiveness and accuracy of fault early warning are poor, and achieves the technical effect of improving the comprehensiveness and accuracy of fault early warning by fusing real-time operation data under multiple operation modes, performing defect extraction and fault event identification.

Description

Technical Field

[0001] The present invention relates to the technical field of fault early warning, and particularly relates to a substation equipment fault early warning system and method integrating defect detection. Background Art

[0002] In modern power systems, substation equipment, as an important part of the power grid, its safe and stable operation is crucial. With the continuous increase of power load and the complexity of equipment operation environment, the fault risks faced by substation equipment are also increasing day by day. Therefore, timely and accurately identifying potential faults and taking preventive measures have become the key to ensuring the reliability of power systems.

[0003] However, the existing substation equipment fault early warning methods often only detect for a single operation mode or specific defects, lacking comprehensive consideration of different operation modes and various defect characteristics, and the comprehensiveness and accuracy of fault early warning are relatively poor. Summary of the Invention

[0004] This application provides a substation equipment fault early warning system and method integrating defect detection, which is used to solve the technical problem that the existing substation equipment fault early warning methods only detect for a single operation mode or specific defects, and the comprehensiveness and accuracy of fault early warning are relatively poor.

[0005] In the first aspect of this application, a substation equipment fault early warning method integrating defect detection is provided. The method includes: obtaining multiple operation modes of the substation equipment; performing mode switching on the substation equipment according to the multiple operation modes, and collecting multiple real-time operation data under the multiple operation modes; extracting defects from the multiple real-time operation data, and outputting multiple groups of defect feature vectors, where each operation mode corresponds to a group of defect feature vectors; identifying fault events of the substation equipment according to the multiple groups of defect feature vectors, and outputting matching fault events; under the condition of the matching fault events, respectively calculating multiple risk indicators corresponding to the multiple groups of defect feature vectors; fusing the multiple risk indicators to generate a fused risk indicator, and generating a fault early warning signal according to the fused risk indicator, and sending it to the upper computer of the substation equipment for fault early warning.

[0006] In a second aspect of the present application, a substation equipment fault early warning system integrating defect detection is provided. The system includes: an equipment operation mode acquisition module for acquiring multiple operation modes of the substation equipment; a real-time operation data acquisition module for performing mode switching on the substation equipment according to the multiple operation modes and acquiring multiple real-time operation data under the multiple operation modes; a defect feature extraction module for extracting defects from the multiple real-time operation data and outputting multiple sets of defect feature vectors, where each operation mode corresponds to a set of defect feature vectors; a fault event recognition module for recognizing fault events of the substation equipment according to the multiple sets of defect feature vectors and outputting matching fault events; a risk index calculation module for calculating multiple risk indexes corresponding to the multiple sets of defect feature vectors respectively under the condition of the matching fault events; and a fault early warning module for fusing the multiple risk indexes to generate a fused risk index, generating a fault early warning signal according to the fused risk index, and sending it to the upper computer of the substation equipment for fault early warning.

[0007] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0008] A substation equipment fault early warning system and method integrating defect detection provided by the present application relate to the technical field of fault early warning. By acquiring and switching multiple operation modes of the equipment to collect real-time operation data, then extracting defects from these data to generate multiple sets of defect feature vectors, analyzing these feature vectors to identify fault events and calculate corresponding risk indexes, fusing multiple risk indexes to generate a comprehensive risk index, and issuing a fault early warning signal based on this index, it ensures timely response to potential faults, solves the technical problem that the existing substation equipment fault early warning methods only detect single operation modes or specific defects, and the comprehensiveness and accuracy of fault early warning are poor, and realizes the technical effect of improving the comprehensiveness and accuracy of fault early warning by fusing real-time operation data under multiple operation modes for defect extraction and fault event recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0010] Figure 1Schematic flowchart of a substation equipment fault warning method integrating defect detection provided by an embodiment of the present application;

[0011] Figure 2 Schematic structural diagram of a substation equipment fault warning system integrating defect detection provided by an embodiment of the present application.

[0012] Explanation of reference numerals: Equipment operation mode acquisition module 11, real-time operation data acquisition module 12, defect feature extraction module 13, fault event recognition module 14, risk index calculation module 15, fault warning module 16. Detailed implementation manners

[0013] The present application provides a substation equipment fault warning system and method integrating defect detection, which is used to solve the technical problem that the existing substation equipment fault warning methods only detect single operation modes or specific defects, and the comprehensiveness and accuracy of fault warning are poor.

[0014] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0015] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned accompanying drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices.

[0016] Embodiment 1, as Figure 1 shown, the present application provides a substation equipment fault warning method integrating defect detection, and the method includes:

[0017] P10: Obtain multiple operation modes of the substation equipment.

[0018] Optionally, obtaining multiple operating modes of the power transformation equipment is the primary link of the fault warning method, aiming to comprehensively grasp the operating characteristics of the power transformation equipment under different working conditions. First, by deeply analyzing the design documents, technical manuals, and historical operation records of the power transformation equipment, all possible operating modes of the equipment are identified. These modes may include normal operating mode, emergency operating mode, maintenance mode, etc., and each mode corresponds to different operating parameters and states. During this process, special attention needs to be paid to identifying those special operating modes that are crucial for equipment fault warning, because they can often reveal the behavioral characteristics of the equipment under extreme or abnormal conditions.

[0019] Secondly, using sensor technology and data acquisition systems, the operating state of the power transformation equipment is monitored in real time. These sensors can accurately measure key parameters such as voltage, current, and temperature of the equipment and transmit this data to the central processing unit in real time. By analyzing this data, the operating modes obtained through document analysis can be further verified and refined.

[0020] Then, pattern recognition techniques such as clustering analysis, decision trees, or neural networks are used to process the collected operating data. These techniques can extract representative and discriminatory features from a large amount of data, thereby accurately identifying the operating mode in which the power transformation equipment is currently located. For example, clustering analysis can group similar operating states into the same category to form different operating mode clusters; decision trees determine the operating mode of the equipment step by step through a series of judgment conditions.

[0021] Finally, the identified operating modes are sorted and archived to form an operating mode library for the power transformation equipment. This library not only contains all possible operating modes of the equipment but also records the typical characteristics, parameter ranges, and possible fault types under each mode. It provides a solid foundation for subsequent mode switching, data acquisition, defect extraction, and fault identification.

[0022] P20: Perform mode switching on the power transformation equipment according to the multiple operating modes, and collect multiple real-time operating data under the multiple operating modes.

[0023] Specifically, after obtaining the multiple operating modes, it is necessary to perform dynamic mode switching on the power transformation equipment according to the multiple operating modes. By adjusting the working state of the equipment in real time, real-time operating data under different modes can be collected, thereby providing a comprehensive basis for fault warning.

[0024] First, the implementation of mode switching relies on an automated control system, which typically includes an embedded controller and a monitoring system. These systems can automatically identify the current working state based on set logical conditions and switch to the corresponding operating mode when necessary. For example, when the device detects a load change or temperature anomaly, the system can automatically switch to a fault troubleshooting mode to collect relevant data.

[0025] During the mode switching process, the collected data should cover key operating parameters such as current, voltage, frequency, temperature, oil level, etc. These parameters can reflect the operating state of the device in different modes and contribute to subsequent fault analysis. Data collection can be achieved through sensors and a real-time data acquisition system to obtain real-time operating data for each operating mode, that is, the dynamic data collected when the device is actually running. This data is crucial for timely detection of potential faults and will be stored in a database to form a comprehensive operating archive for subsequent defect extraction and fault identification.

[0026] In summary, through flexible mode switching and efficient data collection, comprehensive monitoring of substation equipment can be achieved, ensuring detailed real-time operating data is collected under different working conditions, providing rich information support for the fault warning system, and making the assessment of the device's health status more accurate and reliable.

[0027] P30: Extract defects from the multiple real-time operating data and output multiple sets of defect feature vectors, where each operating mode corresponds to a set of defect feature vectors.

[0028] It should be understood that after collecting multiple real-time operating data, the next step is to extract defects from these data to generate multiple sets of defect feature vectors. Defect feature vectors are quantitative representations of potential device faults and can reflect the health status of the device in different operating modes.

[0029] First, the process of defect extraction usually relies on signal processing and data analysis techniques. Common methods include time-domain analysis, frequency-domain analysis, and wavelet transform, etc. These techniques can extract representative features from the original real-time data, such as vibration amplitude, frequency components, and waveform features, etc. Through these features, abnormal operating modes of the device can be identified.

[0030] For each operating mode, specific thresholds and algorithms need to be set for the defect extraction process to screen out possible defects from the real-time data. For example, in the normal operating mode, the characteristic values should be maintained within a certain range; if the data exceeds this range, it may indicate potential device faults. In this case, the system will automatically record the corresponding features and form defect feature vectors.

[0031] Each set of defect feature vectors will correspond to a corresponding operating mode. Such an organization not only helps with subsequent fault event identification but also provides rich context information. For example, the feature vectors under a certain operating mode may be closely related to the performance of the device under high load, while another set may reflect the health indicators of the device in the maintenance mode.

[0032] By systematically organizing the defect feature vectors under multiple operating modes, it can provide the necessary data basis for subsequent fault analysis and decision support. This method ensures comprehensive monitoring of the device's health status, helps to promptly identify and handle potential faults, thereby improving the reliability and safety of substation equipment.

[0033] P40: Identify fault events for the substation equipment according to the multiple sets of defect feature vectors and output the matching fault events.

[0034] Furthermore, step P40 of the embodiment of the present application further includes:

[0035] P41: Obtain multiple sets of training data sets, where the multiple sets of training data sets include multiple fault event samples and multiple operation data samples of the substation equipment based on the multiple operating modes under each fault event sample; P42: Perform defect feature convolution on the multiple operation data samples and output multiple defect feature samples; P43: Establish a mapping network between the multiple fault event samples and the multiple defect feature samples; P44: Construct a fault event recognizer according to the mapping network, call the fault event recognizer to perform matching recognition on the multiple sets of defect feature vectors, and output the matching fault events.

[0036] Specifically, after obtaining multiple sets of defect feature vectors of the substation equipment, perform fault event identification, identify the patterns that match the known fault events from these feature vectors, and output the matching fault events.

[0037] First, obtain multiple sets of training data sets. These training data sets are composed of multiple fault event samples, and each fault event sample contains the operation data samples of the substation equipment under multiple operating modes. These operation data samples have been preprocessed and feature-extracted, and can accurately reflect the operating status of the equipment under different fault events. The diversity and representativeness of the training data sets are crucial for constructing an accurate fault event identification model.

[0038] Next, perform defect feature convolution on multiple operation data samples. Using technologies such as convolutional neural networks (CNNs), deeply extract and encode the defect features in the operation data samples. The convolution operation can capture local features in the data, and through multiple layers of convolution and pooling operations, gradually abstract higher-level feature representations. The output is multiple defect feature samples, which contain the feature information of the device under different fault events.

[0039] Then, establish a mapping network between multiple fault event samples and multiple defect feature samples. This step constructs the mapping relationship between fault events and defect features through machine learning algorithms such as support vector machines (SVMs), neural networks (NNs), or deep learning models. The mapping network can learn the complex relationship between fault events and features and achieve accurate classification of new feature vectors.

[0040] Finally, construct a fault event recognizer according to the mapping network. The fault event recognizer is a trained model that can perform matching recognition on the input defect feature vectors and output the matching fault events. In practical applications, call the fault event recognizer to process multiple groups of defect feature vectors, and determine the possible current fault events of the device according to the classification results in the mapping network.

[0041] Through the above steps, fault event recognition depends not only on the accuracy of the data but also on effective model construction and feature extraction technologies to ensure accurate identification of the device's fault status under different operating modes.

[0042] Furthermore, when establishing the mapping network between the multiple fault event samples and the multiple defect feature samples, step P43 of the embodiment of the present application further includes:

[0043] P43-1: The mapping network includes a first encoding mapping network and a second association mapping network, and the first encoding mapping network is connected to the second association mapping network; P43-2: Among them, the first encoding mapping network is used to identify the one-to-many encoding mapping relationship between multiple defect feature samples under each fault event sample; P43-3: The second association mapping network is used to identify the association mapping relationship of the association relationship between multiple operation data samples under each fault event sample.

[0044] Optionally, when establishing the mapping network between multiple fault event samples and multiple defect feature samples, first, construct a composite mapping network including a first encoding mapping network and a second association mapping network. The combination of these two networks aims to fully explore the complex relationship between fault events and defect features and improve the overall recognition ability of the system.

[0045] Specifically, the first encoding mapping network is responsible for identifying the one-to-many encoding mapping relationship between multiple defect feature samples under each fault event sample. This means that for each fault event sample, the first encoding mapping network can generate the encoding representations of one or more defect feature samples, and these encoding representations accurately reflect the corresponding relationship between the fault event and the defect feature. This one-to-many mapping relationship reflects the complexity and diversity of fault events, because the same fault event may trigger multiple different defect features. To achieve this mapping, encoder structures in deep learning, such as autoencoders or variational autoencoders, can be used, which can extract low-dimensional and representative feature encodings from high-dimensional data.

[0046] Meanwhile, the second association mapping network focuses on identifying the association relationship between multiple operation data samples under each fault event sample. This association relationship may be manifested as the similarity, dependence or causal relationship between data. By constructing this association mapping, we can better understand how fault events affect the operation state of the device and the internal connection between different operation states. To achieve the association mapping, techniques such as graph neural networks or attention mechanisms can be used, which can capture the complex relationships between data and generate association matrices or attention weights reflecting these relationships.

[0047] In summary, by establishing a composite structure including the first encoding mapping network and the second association mapping network, the deep mapping between fault event samples and defect feature samples can be realized. This complex mapping relationship not only improves the accuracy of fault identification, but also provides a solid technical foundation for subsequent fault prediction and early warning, thus enhancing the safety and reliability of substation equipment.

[0048] Furthermore, step P44 of the embodiment of the present application further includes:

[0049] P44-1: The fault event recognizer further includes an adversarial network, which is obtained by training a generator and a discriminator; P44-2: The generator is used to simulate the generation of samples of multiple defect features of the substation equipment based on the multiple operation modes under each fault event sample; P44-3: The discriminator is used to compare the multiple operation data samples of the multiple operation modes with the multiple defect feature generation samples output by the generator to obtain a comparison loss rate; P44-4: Optimize the generator according to the comparison loss rate until the comparison loss rate is less than a preset comparison loss rate, and the generator and the discriminator converge to output the adversarial network.

[0050] It should be understood that the construction of the fault event recognizer further enhances the intelligence of the system, especially by introducing an adversarial network to improve the accuracy and robustness of fault recognition. The adversarial network consists of two main components: a generator and a discriminator. The generator is responsible for generating samples similar to the real data, while the discriminator is responsible for evaluating the authenticity of these samples. Through this adversarial training method, the generator and the discriminator compete with each other to continuously improve the recognition accuracy.

[0051] Among them, the role of the generator is to simulate multiple defective feature samples generated based on multiple operating modes under each fault event sample. The generator uses a deep learning model, such as a generative adversarial network (GAN), to generate new samples through the input defective features. These samples should not only reflect the potential faults of the device but also simulate the data features under real operating conditions to a certain extent. The discriminator calculates the comparison loss rate by comparing the real operating data samples under multiple operating modes with the defective feature generated samples output by the generator. The comparison loss rate is an index to measure the difference between the generated samples and the real samples, and the smaller it is, the more similar the generated samples are to the real data. The discriminator improves the recognition accuracy of real data by optimizing its own judgment ability.

[0052] According to the calculated comparison loss rate, the generator is optimized. The goal is to continuously adjust the parameters of the generator until the comparison loss rate is lower than the preset threshold, which means that the difference between the samples generated by the generator and the real samples is small enough. In addition, when both the generator and the discriminator reach the convergence state, it indicates that the training of the adversarial network is completed and can effectively generate samples highly similar to the real data.

[0053] By introducing the adversarial network, the fault event recognizer can not only generate diverse defective feature samples but also improve the generalization ability of the model and enhance the recognition ability of unknown faults. This process provides stronger intelligent support for the fault warning system of substation equipment, thus improving the overall operation safety and reliability.

[0054] Furthermore, the adversarial network is obtained by training the generator and the discriminator. Step P44-1 of the embodiment of the present application further includes:

[0055] P44-11: Randomly select an initial alternating step size; P44-12: Perform multiple rounds of alternating training on the generator and the discriminator according to the initial alternating step size to obtain the gradient change of the generator and the gradient change of the discriminator; P44-13: Identify the balance degree of the gradient change of the generator and the gradient change of the discriminator. If the balance degree is less than the preset balance degree, perform iterative optimization on the initial alternating step size according to the balance degree difference.

[0056] In a possible embodiment of the present application, during the process of constructing the adversarial network, the model is optimized by alternately training the generator and the discriminator. Specifically, first, the system randomly selects an initial alternating step size. The alternating step size refers to the frequency and amplitude at which the generator and the discriminator update their parameters respectively during the training process. A reasonable initial alternating step size can effectively promote the stability and convergence speed of the training process.

[0057] Next, according to the initial alternating step size, the generator and the discriminator are alternately trained for multiple rounds. In each round of training, the generator is first updated once to generate new defective feature samples; subsequently, the discriminator updates its parameters to improve its ability to distinguish between real samples and generated samples. During this process, the system records the gradient changes of the generator and the gradient changes of the discriminator. These gradients reflect the learning progress of the model in the current training state and can provide a basis for subsequent optimization.

[0058] Next, the balance degree of the gradient changes of the generator and the discriminator is identified. The balance degree is an index to measure the difference in the learning progress of the two networks. If the learning progress of the generator is too fast and the discriminator cannot keep up, it may lead to a decrease in the quality of the samples generated by the generator, and vice versa. If the balance degree is less than the preset balance degree threshold, it indicates that the current training process is not balanced enough. At this time, according to the difference in the balance degree, the initial alternating step size is iteratively optimized.

[0059] By dynamically adjusting the alternating step size, the coordinated development between the generator and the discriminator can be achieved, ensuring that the two maintain a good learning balance during the training process. This optimization strategy improves the stability and generation ability of the adversarial network, making the finally generated samples more representative, thereby further enhancing the overall performance and accuracy of the fault event recognizer.

[0060] Furthermore, step P40 of the embodiment of the present application further includes:

[0061] P44-5: The fault event recognizer performs a first match according to the multiple groups of defective feature vectors to obtain the first match fault events with a match degree greater than the preset match degree; P44-6: Perform feature correlation analysis on the multiple groups of defective feature vectors and output the feature correlation; P44-7: Perform a second match among the first match fault events according to the feature correlation to obtain the second match fault events; P44-8: Output the event with the highest match degree among the second match fault events as the match fault event.

[0062] Specifically, to further improve the accuracy and reliability of fault event recognition, the fault event recognition machine performs a series of detailed matching and analysis operations. First, the recognition machine conducts a primary match based on multiple groups of defect feature vectors to obtain fault events with a matching degree greater than a preset matching degree. The core of this step is to use a pre-trained model to compare the defect feature vectors with known fault events, thereby screening out potential fault events that meet the conditions. The matching degree is a quantitative indicator that reflects the similarity between the feature vectors and the fault events. A higher matching degree indicates a stronger correlation between the feature vectors and the fault events.

[0063] Next, perform feature correlation analysis on multiple groups of defect feature vectors to explore the mutual relationships between various defect features, and reveal the potential connections between features through statistical analysis methods (such as correlation analysis or principal component analysis). Feature correlation can help identify which features are most critical for the judgment of fault events and provide a reference basis for subsequent matching.

[0064] Furthermore, based on the obtained feature correlation, perform a secondary match among the fault events obtained from the primary match. For the already screened fault events, use the feature correlation to further optimize the matching results. Through more refined comparison, more accurate fault events can be identified, improving the recognition accuracy. Finally, output the event with the highest matching degree in the secondary match as the final matching fault event, ensuring that the recognition machine can provide the most credible fault judgment results for subsequent maintenance and decision-making.

[0065] By implementing these steps, the fault event recognition machine can not only achieve the preliminary screening of potential faults, but also further improve the accuracy and reliability of the recognition results through in-depth analysis and secondary matching, providing strong support for the fault warning of substation equipment.

[0066] P50: Under the condition of the matching fault event, calculate multiple risk indicators corresponding to the multiple groups of defect feature vectors.

[0067] It should be understood that for the identified matching fault events, calculate multiple risk indicators corresponding to multiple groups of defect feature vectors. The goal of this process is to quantify the impact of potential faults on the safe operation of substation equipment, thereby providing a basis for subsequent risk assessment and decision-making.

[0068] First, the calculation of risk indicators is usually based on the key data extracted from the defect feature vectors. These feature vectors may include information such as abnormal values of operating parameters, fault frequencies, and defect severities. Through these data, risk assessment methods in statistics and engineering, such as fault tree analysis (FTA) or failure mode and effects analysis (FMEA), can be used to evaluate the risk levels of each defect feature.

[0069] In this process, the system will compare each set of defect feature vectors with the preset risk criteria to calculate the risk indicators related to the features. For example, some features may indicate overheating during equipment operation, which can be quantified by the deviation of the temperature threshold. Other features may be related to the vibration mode of the equipment. In this case, the corresponding risk can be calculated based on the abnormal values of the vibration frequency. At the same time, the system can also adopt machine learning algorithms, such as regression analysis or decision tree, to predict the risk indicators. By analyzing historical data, a relationship model between defect features and failure risks is established to provide more accurate risk assessment results.

[0070] The calculated multiple risk indicators will provide basic data for subsequent risk fusion. These indicators not only help identify the current equipment status but also provide guidance for operators to formulate corresponding maintenance strategies, perform preventive maintenance in advance, and reduce the probability of failures. Through such calculation of risk indicators, the fault event identification system can comprehensively evaluate the operation safety of substation equipment, provide data support for timely decision-making, and thus effectively improve the overall reliability and safety of the equipment.

[0071] P60: Integrate the multiple risk indicators to generate a fused risk indicator, generate a fault warning signal according to the fused risk indicator, and send it to the upper computer of the substation equipment for fault warning.

[0072] Furthermore, step P60 of the embodiment of the present application further includes:

[0073] P61: Calculate the mode importance of the multiple operation modes respectively; P62: Configure weights according to the mode importance, and integrate the multiple risk indicators according to the configured weights to generate a fused risk indicator.

[0074] Optionally, a comprehensive fused risk indicator is generated by integrating multiple risk indicators, and then a fault warning signal is generated and sent to the upper computer of the substation equipment to ensure timely response to potential faults.

[0075] First of all, based on the fact that multiple risk indicators have been calculated, it can be found that although these risk indicators respectively reflect different aspects of equipment failures, directly adding them or simply averaging them may not accurately reflect the overall risk level. Therefore, the concept of mode importance is introduced, aiming to consider the contribution degree of different operation modes to the equipment failure risk. The system needs to calculate the mode importance of multiple operation modes respectively. The mode importance is an indicator for evaluating the contribution of each operation mode to the overall safety and operation efficiency of the equipment. This process is usually based on historical data analysis and expert knowledge, and the relative importance of each mode in fault warning is determined through quantitative methods (such as weight calculation or sensitivity analysis). For example, some modes may show higher sensitivity under specific failures, so their importance is relatively high.

[0076] Next, weight configuration is performed according to the calculated pattern importance. This means that when fusing risk indicators, risk indicators of different patterns will be assigned different weights to reflect their actual influence in fault warning. In this way, patterns with higher importance will have a greater impact on the final fused risk indicator, while patterns with lower importance will correspondingly reduce their influence.

[0077] Furthermore, the generation process of the fused risk indicator involves weighted summation of each risk indicator according to the configured weights. In this way, the final fused risk indicator not only synthesizes multiple risk factors but also considers the contribution degrees of different patterns to fault warning. This indicator provides a comprehensive assessment of the equipment operation status and can effectively indicate potential fault risks.

[0078] Finally, according to the generated fused risk indicator, a fault warning signal is issued and sent to the upper computer of the substation equipment. This signal not only reminds the operator to pay attention to potential faults but also can trigger automated response measures, such as adjusting equipment operation parameters or starting preventive maintenance procedures. Through this series of steps, the fault warning system can effectively integrate various risk information and provide timely and accurate warnings, greatly improving the safety and operation reliability of substation equipment.

[0079] In summary, the embodiments of the present application have at least the following technical effects:

[0080] The present application obtains multiple operation modes of the equipment and switches them to collect real-time operation data. Then, defect extraction is performed on these data to generate multiple groups of defect feature vectors. By analyzing these feature vectors, fault events are identified and corresponding risk indicators are calculated. Multiple risk indicators are fused to generate a comprehensive risk indicator, and a fault warning signal is issued based on this indicator to ensure timely response to potential faults.

[0081] It achieves the technical effect of improving the comprehensiveness and accuracy of fault warning by fusing real-time operation data under multiple operation modes, performing defect extraction and fault event identification.

[0082] Embodiment 2, based on the same inventive concept as the method for fault warning of substation equipment with integrated defect detection in the foregoing embodiment, as Figure 2 shown, the present application provides a fault warning system for substation equipment with integrated defect detection. The system in the embodiments of the present application and the method embodiments are based on the same inventive concept. Among them, the system includes:

[0083] An equipment operation mode acquisition module 11, which is used to acquire multiple operation modes of the substation equipment.

[0084] The real-time operation data acquisition module 12 is used to switch the mode of the substation equipment according to the multiple operation modes and collect multiple real-time operation data under the multiple operation modes.

[0085] The defect feature extraction module 13 is used to extract defects from the multiple real-time operation data and output multiple groups of defect feature vectors, where each operation mode corresponds to a group of defect feature vectors.

[0086] The fault event recognition module 14 is used to identify fault events of the substation equipment according to the multiple groups of defect feature vectors and output matching fault events.

[0087] The risk index calculation module 15 is used to calculate multiple risk indexes corresponding to the multiple groups of defect feature vectors respectively under the condition of the matching fault event.

[0088] The fault warning module 16 is used to fuse the multiple risk indexes, generate a fused risk index, generate a fault warning signal according to the fused risk index, and send it to the upper computer of the substation equipment for fault warning.

[0089] Further, the fault event recognition module 14 is further used to execute the following steps:

[0090] Obtain multiple groups of training data sets, where the multiple groups of training data sets include multiple fault event samples and multiple operation data samples of the substation equipment based on the multiple operation modes under each fault event sample; perform defect feature convolution on the multiple operation data samples and output multiple defect feature samples; establish a mapping network between the multiple fault event samples and the multiple defect feature samples; construct a fault event recognizer according to the mapping network, and call the fault event recognizer to perform matching recognition on the multiple groups of defect feature vectors and output matching fault events.

[0091] Further, the fault event recognition module 14 is further used to execute the following steps:

[0092] Establish a mapping network between the multiple fault event samples and the multiple defect feature samples. The mapping network includes a first coding mapping network and a second association mapping network, and the first coding mapping network is connected to the second association mapping network; wherein, the first coding mapping network is used to identify the one-to-many coding mapping relationship between the multiple defect feature samples under each fault event sample; the second association mapping network is used to identify the association mapping relationship between the multiple operation data samples under each fault event sample.

[0093] Further, the fault event recognition module 14 is further configured to perform the following steps:

[0094] The fault event recognizer further includes an adversarial network, which is obtained by training a generator and a discriminator; the generator is used to simulate, under each fault event sample, a plurality of defect features of the power transformation equipment based on the plurality of operation modes to generate samples; the discriminator is used to compare a plurality of operation data samples of the plurality of operation modes with the plurality of defect feature generated samples output by the generator to obtain a comparison loss rate; optimize the generator according to the comparison loss rate until the comparison loss rate is less than a preset comparison loss rate, and the generator and the discriminator converge to output the adversarial network.

[0095] Further, the fault event recognition module 14 is further configured to perform the following steps:

[0096] Randomly select an initial alternating step size; perform multiple rounds of alternating training on the generator and the discriminator according to the initial alternating step size to obtain the gradient change of the generator and the gradient change of the discriminator; identify the balance degree of the gradient change of the generator and the gradient change of the discriminator, and if the balance degree is less than a preset balance degree, iteratively optimize the initial alternating step size according to the balance degree difference.

[0097] Further, the fault event recognition module 14 is further configured to perform the following steps:

[0098] Perform a first match according to the multi-group defect feature vectors by the fault event recognizer to obtain a first-match fault event with a match degree greater than a preset match degree; perform feature correlation analysis on the multi-group defect feature vectors to output feature correlation; perform a second match on the first-match fault events according to the feature correlation to obtain a second-match fault event; output the event with the highest match degree in the second-match fault events as the matched fault event.

[0099] Further, the fault warning module 16 is further configured to perform the following steps:

[0100] Calculate the importance of each of the plurality of operation modes respectively; perform weight configuration with the mode importance, and fuse the plurality of risk indicators according to the configured weights to generate a fused risk indicator.

[0101] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0102] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0103] This specification and the drawings are merely exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A fault early warning method for substation equipment integrating defect detection, characterized in that: The method comprises: Obtain multiple operating modes of substation equipment; Switching the mode of the power transformation equipment according to the multiple operation modes, and collecting multiple real-time operation data under the multiple operation modes; Extracting defects from the plurality of real-time operation data and outputting a plurality of groups of defect feature vectors, wherein each operation mode corresponds to a group of defect feature vectors; Identify fault events of the substation equipment according to the multiple groups of defect feature vectors, and output matching fault events; Under the condition of the matching fault event, respectively calculating a plurality of risk indicators corresponding to the plurality of groups of defect feature vectors; The multiple risk indicators are integrated to generate a fused risk indicator, a fault warning signal is generated according to the fused risk indicator, and the signal is sent to the host computer of the power transformation equipment for fault warning.

2. A method for early warning of power transformation equipment failures integrating defect detection as claimed in claim 1, characterized in that: The method includes: identifying fault events of the substation equipment according to the multiple groups of defect feature vectors and outputting matching fault events. Acquire multiple sets of training data sets, wherein the multiple sets of training data sets include multiple fault event samples and multiple operation data samples of the substation equipment based on the multiple operation modes under each fault event sample; Performing defect feature convolution on the multiple operation data samples, and outputting multiple defect feature samples; Establishing a mapping network between the plurality of fault event samples and the plurality of defect feature samples; A fault event recognition machine is constructed according to the mapping network, and the fault event recognition machine is called to match and recognize the multiple groups of defect feature vectors, and a matching fault event is output.

3. A method for early warning of power transformation equipment failure integrating defect detection as claimed in claim 2, characterized in that: Establishing a mapping network between the multiple fault event samples and the multiple defect feature samples, the mapping network comprising a first coding mapping network and a second association mapping network, the first coding mapping network being connected to the second association mapping network; The first coding mapping network is used to identify a one-to-many coding mapping relationship between each fault event sample and multiple defect feature samples; The second association mapping network is used to identify an association mapping relationship of association relationships between multiple operation data samples under each fault event sample.

4. A method for early warning of power transformation equipment failures integrating defect detection as claimed in claim 2, characterized in that: The fault event recognition machine also includes an adversarial network, which is obtained by training a generator and a discriminator; The generator is used to simulate that under each fault event sample, the substation equipment generates samples based on multiple defect characteristics of the multiple operation modes; The discriminator is used to compare the multiple operation data samples of the multiple operation modes with the multiple defect feature generation samples output by the generator to obtain a comparison loss rate; The generator is optimized according to the comparison loss rate until the comparison loss rate is less than a preset comparison loss rate, the generator and the discriminator converge, and the adversarial network is output.

5. A method for early warning of power transformation equipment failure integrating defect detection as claimed in claim 4, characterized in that: The adversarial network is obtained by training a generator and a discriminator, and the method further includes: Randomly select the initial alternation step size; Perform multiple rounds of alternating training on the generator and the discriminator according to the initial alternating step length to obtain the gradient change of the generator and the gradient change of the discriminator; The balance between the gradient change of the generator and the gradient change of the discriminator is identified, and if the balance is less than a preset balance, the initial alternating step size is iteratively optimized according to the balance difference.

6. A method for early warning of power transformation equipment failures integrating defect detection as claimed in claim 2, characterized in that: The fault event identification machine is called to match and identify the multiple groups of defect feature vectors and output matching fault events, the method comprising: According to the fault event identification machine, a match is performed according to the multiple groups of defect feature vectors to obtain a matching fault event with a matching degree greater than a preset matching degree; Performing feature correlation analysis on the multiple groups of defect feature vectors and outputting feature correlation; Performing a secondary match in the primary matching failure event according to the feature correlation to obtain a secondary matching failure event; The event with the highest matching degree among the secondary matching failure events is output as the matching failure event.

7. A method for early warning of power transformation equipment failures integrating defect detection as claimed in claim 1, characterized in that: The multiple risk indicators are integrated to generate a fused risk indicator, the method comprising: respectively calculating the mode importance of the plurality of operating modes; The weights are configured according to the importance of the modes, and the multiple risk indicators are fused according to the configured weights to generate a fused risk indicator.

8. A fault warning system for substation equipment integrating defect detection, characterized in that: The system comprises: A device operation mode acquisition module, wherein the device operation mode acquisition module is used to acquire multiple operation modes of the substation equipment; A real-time operation data acquisition module, the real-time operation data acquisition module is used to switch the mode of the substation equipment according to the multiple operation modes, and collect multiple real-time operation data under the multiple operation modes; A defect feature extraction module, the defect feature extraction module is used to extract defects from the multiple real-time operation data and output multiple groups of defect feature vectors, wherein each operation mode corresponds to a group of defect feature vectors; A fault event identification module, the fault event identification module is used to identify fault events of the substation equipment according to the multiple groups of defect feature vectors, and output matching fault events; A risk indicator calculation module, wherein the risk indicator calculation module is used to respectively calculate a plurality of risk indicators corresponding to the plurality of groups of defect feature vectors under the condition of the matching fault event; A fault warning module is used to fuse the multiple risk indicators to generate a fused risk indicator, generate a fault warning signal according to the fused risk indicator, and send it to the host computer of the substation equipment for fault warning.

Images