Distribution line discharge fault control method and system based on primary and secondary fusion switches

Abstract

The invention discloses a distribution line discharge fault control method and system based on a primary and secondary fusion switch, and the method comprises the steps: S1, collecting a distribution line discharge fault feature signal and at least one auxiliary state signal through a built-in sensing unit of the primary and secondary fusion switch, and forming multi-source time sequence data; s2, performing space-time alignment and standardization processing on the multi-source time sequence data to generate a fusion feature signal; s3, inputting the fusion feature signal into a pre-trained fault identification and severity evaluation model, and generating a comprehensive judgment signal of a fault type, a position and a danger level; s4, a preset control strategy library is matched based on the comprehensive judgment signal, and a precise control execution signal of the primary and secondary fusion switch is generated; and S5, after the control action is executed, executing effect data are collected through a feedback unit, and online self-adaptive optimization is carried out on the model and the control strategy. According to the invention, the problems of delayed discharge fault identification, low control precision and poor adaptability of the distribution line are solved, and rapid and accurate control of the fault is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power distribution line fault control technology and primary-secondary fusion technology, and particularly relates to a power distribution line discharge fault control method and system based on a primary-secondary fusion switch. BACKGROUND

[0002] As an important part of the power system, the power distribution line directly undertakes the key task of power transmission and distribution, and its operation stability directly affects the reliability of power supply. Discharge fault is one of the common fault types of the power distribution line, mainly including arc grounding fault, insulator flashover, cable partial discharge, etc. If such faults cannot be controlled in time and effectively, they may cause line tripping, equipment damage, and even expand into large-area power outage accidents, causing serious economic losses and social impact.

[0003] The traditional discharge fault control method of the power distribution line mainly relies on the cooperation of the relay protection device and the conventional switch equipment. In terms of hardware, the fault signal is detected by configuring overcurrent relays, zero sequence current transformers and other devices, and the traditional mechanical switch is used to perform the tripping operation; in terms of algorithm, the fault judgment logic based on fixed threshold is mostly used, such as detecting whether the line current and voltage exceed the preset threshold to determine the occurrence of fault, and then triggering the switch action. However, these traditional methods have significant limitations: first, the fault detection only relies on a single electrical quantity signal, and the early weak characteristics of the discharge fault are difficult to capture, which is easily affected by factors such as line load fluctuation and harmonic interference, resulting in low fault identification accuracy and high misoperation rate; second, the control strategy uses a fixed configuration mode, which cannot be differentiated according to the fault type, location and severity, and has poor adaptability to complex fault scenarios, often resulting in control action lag or over-control problems; third, it lacks the ability to adapt to changes in the operating state of the equipment, and as the mechanical parts of the primary-secondary fusion switch age and the characteristics of the sensing unit drift, the control accuracy of the traditional method will gradually decrease, requiring frequent manual maintenance and calibration; fourth, the existing technology fails to fully utilize the integrated and intelligent advantages of the primary-secondary fusion switch, and the coordination between sensing data and control execution is insufficient, which cannot realize the deep fusion of fault identification and control decision-making.

[0004] In recent years, the application of intelligent technology in power systems has gradually become popular. Some research attempts to introduce artificial intelligence algorithms into the field of distribution line fault identification. However, existing solutions still have obvious defects: most of them only identify a single fault type and do not consider the multi-dimensional feature changes in the fault evolution process; the linkage between control strategy and fault identification result is weak, and a closed-loop control mechanism of "identification - decision - execution - optimization" has not been formed; at the same time, existing solutions lack deep adaptation to the hardware characteristics of primary and secondary fusion switches, making it difficult to meet the real-time and accuracy requirements of distribution line fault control. Therefore, developing an intelligent control method and system that can fully utilize the multi-source sensing capability of primary and secondary fusion switches, accurately identify discharge faults, dynamically match control strategies, and have self-optimization capability, has become the key to breaking through the bottleneck of existing distribution line discharge fault control. SUMMARY

[0005] In view of the above shortcomings of the prior art, the purpose of the present application is to provide a distribution line discharge fault control method and system based on a primary and secondary fusion switch, to solve the problems of low accuracy of discharge fault identification, poor adaptability of control strategy, and lack of self-optimization capability in traditional methods. Through the multi-source sensing components of the primary and secondary fusion switch, the discharge fault characteristic signals and line auxiliary state signals are synchronously collected, and the fusion feature signals are generated after time-space alignment and standardization processing; the pre-trained AI model is input to extract deep features, and the comprehensive judgment signals of fault type, location and danger level are output; the precise control signals are generated by matching the hierarchical control strategy library, and the switch is driven to execute the action, while the online adaptive updating of the model and the strategy library is realized by using the feedback signals, efficiently solving the fault control problem.

[0006] The present application provides a distribution line discharge fault control method based on a primary and secondary fusion switch, comprising: S1: Collecting discharge fault characteristic signals of the distribution line through the sensing components of the primary and secondary fusion switch, and synchronously collecting at least one auxiliary state signal reflecting the operating state of the line, forming multi-source time series data; S2: Time-space alignment and standardization processing of the discharge fault characteristic signals and auxiliary state signals are performed to eliminate signal dimension difference and time deviation, and fusion feature signals are generated; S3: Inputting the fusion feature signals into a pre-trained fault identification and evaluation model, the model extracts deep features related to discharge faults in the signals, and generates comprehensive judgment signals containing fault type, fault location and danger level; S4: Based on the comprehensive judgment signal, calling a pre-set control strategy library, matching the control logic suitable for the current fault scene, and generating precise control execution signals for the primary and secondary fusion switch; S5: According to the control execution signal, a secondary fusion switch is driven to perform corresponding control actions, and a line state feedback signal after control is collected, and the signal is used for online adaptive updating of a fault identification and evaluation model and a control strategy library, so that the control effect is continuously optimized.

[0007] In an embodiment of the application, the discharge fault characteristic signal in step S1 includes a fault transient current signal collected by a transient current sensor, an ultrasonic signal collected by a partial discharge ultrasonic sensor, and a transient voltage signal collected by a transient voltage sensor, and the auxiliary state signal includes an environmental temperature and humidity signal, a line load current signal, and a switch mechanical state signal. When a discharge fault occurs, significant electrical parameter mutations and physical characteristic changes will be accompanied. The transient current sensor can capture the current mutation waveform at the moment of fault occurrence, including key information such as the initial time of fault occurrence and the development intensity. The partial discharge ultrasonic sensor can detect the ultrasonic signal generated in the discharge process, and can reflect the position and severity of the partial discharge. The transient voltage sensor records the voltage amplitude and phase change before and after the fault, and assists in judging the fault type. The environmental temperature and humidity signal can reflect the influence of the external environment on the insulation performance of the line, such as that high humidity environment is easy to cause flashover of the insulator. The line load current signal can distinguish between fault and normal load fluctuation, and avoid misjudgment. The switch mechanical state signal can monitor the health status of the opening and closing mechanism of the primary and secondary fusion switch, and ensure the reliability of the control action. The synchronous collection of multi-dimensional signals makes the fusion characteristic signal be able to comprehensively cover the electrical, physical and environmental influence factors of fault occurrence, and provides a rich data source for subsequent accurate fault identification.

[0008] In an embodiment of the present application, the spatio-temporal alignment and normalization processing in step S2 specifically includes unifying all time series signals to the same time reference, achieving strict time synchronization of signals with different sampling frequencies through signal interpolation and timestamp calibration, segmenting the synchronized signals using a sliding time window to ensure that the signals within each data segment have spatio-temporal consistency; and performing detrending and normalization conversion on each type of signal within each data segment to eliminate the effects of differences in the dimensions and signal amplitude fluctuations of different sensing components. Due to differences in the sampling frequencies and signal dimensions of different sensing components, such as the MHz-level sampling frequency of transient current sensors and the Hz-level sampling frequency of environmental temperature and humidity sensors, direct fusion can lead to distortion in fault feature extraction. Through timestamp calibration and interpolation algorithms, low sampling frequency signals are adapted to the time axis of high sampling frequency signals to ensure that multi-source signals at the same time can be accurately corresponded; and sliding time window segmentation processing converts continuous time series signals into fixed-length signal segments, facilitating feature extraction by the model. Detrending processing can eliminate slow drift components in the signal, such as the effects of slow temperature changes on sensor output; and normalization conversion maps signals of different dimensions to the same numerical interval through mathematical transformation, avoiding the masking of key fault features by signals with large amplitudes, generating fusion feature signals with uniform scales and prominent features, and providing stable and high-quality input data for fault recognition and evaluation models, effectively improving the accuracy and robustness of model recognition.

[0009] In an embodiment of the present application, the fault identification and assessment model is a hybrid neural network model based on deep learning, including a convolutional neural network module for extracting local fault features, a recurrent neural network module for capturing signal time dependence, and an attention mechanism module for feature fusion and decision making. The fusion feature signal is first input into the convolutional neural network module, which automatically extracts local fault features in the signal, such as transient current mutation peak, ultrasonic signal specific frequency component, etc., through the alternating operation of multiple convolutional layers and pooling layers; the output of the convolutional neural network module is sent to the recurrent neural network module, which remembers the time evolution law of the signal through the gating mechanism and captures the feature change of the whole process from the occurrence, development to stability of the fault, such as the amplitude decay trend of the partial discharge signal with time; the attention mechanism module dynamically weights and fuses the local features extracted by the convolutional neural network and the time sequence features captured by the recurrent neural network, focusing on the feature dimensions that play a key role in fault identification, such as the sudden drop feature of transient voltage in arc grounding fault, and strengthening the influence weight of key information. Finally, through the full connection layer, the fused deep features are classified and regressed, and the comprehensive judgment signal containing fault type accurate classification (such as arc grounding fault, insulator flashover, cable partial discharge, etc.), fault location positioning (accurate to line section or equipment node) and danger level quantitative assessment (such as high, medium and low three levels) is output, realizing the synchronous identification of multi-dimensional information of discharge fault and solving the problem of low accuracy of single feature identification in traditional methods.

[0010] In an embodiment of the present application, the training method of the pre-trained fault identification and evaluation model comprises building a discharge fault simulation platform of distribution lines in a controlled test environment, simulating discharge fault scenes of different types, different positions and different severity, synchronously collecting various fault characteristic signals and auxiliary state signals through the sensing components of the primary and secondary fusion switches, simultaneously using high-precision fault detection equipment to obtain the true labels of fault type, position and danger level, and constructing a large-scale labeled training sample set; taking the true labels as the training target and taking the fusion characteristic signals as the input, the mixed neural network model is supervised trained. The test platform can simulate common discharge fault types of distribution lines, generate fault scenes of different severity by adjusting the wiring mode of the fault occurrence position and the insulation damage degree of the fault point; the high-precision fault detection equipment adopts professional fault locator, partial discharge detector, etc., which can provide millimeter-level fault position measurement and accurate fault level evaluation, ensuring the accuracy of the training sample labels. In the training process, the cross-entropy loss function is used to measure the fault type classification error, the mean square error loss function is used to measure the prediction error of the fault position and danger level, the model parameters are iteratively adjusted by the Adam optimizer, and the prediction error is continuously reduced, so that the model learns the complex mapping relationship between the fusion features and the multi-dimensional fault information. Early stopping mechanism and regularization technology are introduced in the training process to avoid model overfitting, until the identification error of the model on the validation set meets the preset accuracy requirement, ensuring that the model has good generalization ability.

[0011] In an embodiment of the present application, the preset control strategy library in step S4 includes hierarchical control strategies for different fault types, locations and danger levels, which are divided into three levels of emergency control, conventional control and early warning control according to the danger level in the comprehensive judgment signal. The emergency control strategy corresponds to high-danger-level faults, such as serious arc grounding faults and cable phase-to-phase discharge faults. Such faults develop rapidly and are easy to cause equipment explosion or line fire. The control strategy adopts fast tripping + fault isolation + load transfer linkage control logic. After receiving the control signal, the primary and secondary fusion switch needs to complete the tripping action within milliseconds to cut off the connection between the fault line and the system, and at the same time, through the linkage control of adjacent switch devices, the load on the fault line is transferred to the standby line to ensure normal power supply in the non-fault area. The conventional control strategy corresponds to medium-danger-level faults, such as slight insulator flashover and early-stage cable partial discharge faults. The control strategy executes precise tripping and reclosing coordination control. After the switch is tripped, a reclosing operation is performed after a preset time interval. If the reclosing is successful, the line power supply is restored. If the reclosing fails, it is determined as a permanent fault and fault isolation is performed. The early warning control strategy corresponds to low-danger-level faults, such as weak discharge caused by line insulation dampness. The control strategy starts the fault monitoring intensification and operation and maintenance linkage mechanism, increases the sampling frequency of the sensing components, continuously monitors the fault development trend, and at the same time sends early warning information to the operation and maintenance management platform to prompt the operation and maintenance personnel to timely perform line inspection and maintenance. The control strategy library also includes adaptive control parameters for different fault types and locations, such as adjusting the timing of tripping action for different fault locations and optimizing the interval time of reclosing for different fault types, to ensure that the control execution signal can accurately match the fault scene and realize differentiated and efficient management and control of discharge faults.

[0012] In an embodiment of the present application, the online adaptive update in step S5 specifically includes continuously monitoring the control-after line state feedback signal, when detecting that the line is in a stable operation state or a fault complete removal state, correlatively analyzing the current comprehensive judgment signal, the control execution signal and the state feedback signal, calculating the fault identification deviation and the control effect evaluation index; if the evaluation index does not reach the preset optimal threshold, using the correlation analysis data to incrementally train the fault identification and evaluation model, and dynamically adjusting the model parameters to adapt to the line operation state change and the equipment characteristic drift. The line state feedback signal includes the line current after control, the voltage recovery condition, the load operation state and the like, and by analyzing these signals, it can be judged whether the fault is successfully removed or the line is restored to normal operation. The fault identification deviation is calculated by comparing the comprehensive judgment signal with the actual fault condition (such as the on-site verification result of the operation and maintenance personnel), and the control effect evaluation index includes the fault removal time, the non-fault area power outage time, the load transfer success rate and the like. When the evaluation index does not reach the optimal threshold, such as when the fault identification deviation exceeds the allowed range or the control action execution time is too long, the system automatically constructs the current multi-source signal data, the comprehensive judgment signal, the control execution signal and the state feedback signal as an incremental training sample, and uses a small batch gradient descent algorithm to fine-tune the parameters of the fault identification and evaluation model, so that the model can adapt to long-term time-varying factors such as line insulation performance change and switchgear aging. At the same time, based on the control effect feedback, the adaptive parameters in the control strategy library are optimized, such as adjusting the reclosing interval time and optimizing the linkage logic of load transfer, so that the system can continuously adapt to the dynamic changes of the distribution line, maintain efficient fault control performance, and reduce the cost of manual maintenance.

[0013] In an embodiment of the present application, the method further comprises a fault prediction and early warning step. In step S3, the fault identification and evaluation model generates a fault development prediction signal based on the trend changes of the fused feature signals in addition to the comprehensive judgment signal. The signal contains the time window and risk level change trend of the possible escalation of the fault. The system starts the control strategy preloading mechanism in advance according to the fault development prediction signal, automatically triggers the corresponding control action when the fault risk level reaches the preset threshold, and sends early warning information to the operation and maintenance management platform. The fault development prediction signal is generated by analyzing parameters such as the change rate and amplitude growth trend of the fused feature signals. For example, if the amplitude of the partial discharge ultrasonic signal continues to increase and the growth rate accelerates, it indicates that the fault may escalate to a serious discharge fault in a short time, and the model will predict the time window of fault escalation and the corresponding risk level. The control strategy preloading mechanism can load the appropriate control logic to the control unit of the primary and secondary fusion switch in advance, shorten the response time of the control signal, and ensure that control measures are taken in time before the fault escalates. The early warning information contains the current state of the fault, the development trend, and the recommended operation and maintenance measures. Operation and maintenance personnel can prepare for repair in advance based on the early warning information, realize the forward-looking control of discharge faults, reduce the loss caused by the expansion of the fault, and further improve the operation reliability of the distribution line.

[0014] In an embodiment of the present application, the method further comprises a control process safety verification step. After generating the control execution signal in step S4, the system verifies the execution conditions, action timing and safety boundaries of the control signal. The verification content includes line load transfer feasibility, switch mechanical action reliability, and fault isolation range rationality. If the verification is passed, the control execution signal is issued. If the verification is not passed, the control strategy parameters are adjusted to generate the control signal again, or a backup control scheme is triggered, and the verification abnormal information is recorded for subsequent system optimization. The line load transfer feasibility verification analyzes parameters such as the capacity of the backup line and the current load rate to determine whether the backup line can handle the transferred load of the fault line, avoiding overloading of the backup line. The switch mechanical action reliability verification monitors parameters such as the opening and closing mechanism state of the primary and secondary fusion switch and the energy storage power voltage to ensure that the switch can normally execute the control action. The fault isolation range rationality verification analyzes the fault position and line topology to ensure that the control action can accurately isolate the fault area and avoid mis-cutting the non-fault area. If it is found that the backup line capacity is insufficient during the verification process, the system will adjust the load transfer scheme and prioritize the transfer of important loads. If the switch mechanical state is abnormal, the backup switch device will be triggered to execute the control action. The safety verification step ensures the safety and reliability of the control action, avoids secondary faults caused by improper control, and improves the stability of the system operation.

[0015] The application also comprises a power distribution line discharge fault control system based on a primary and secondary fusion switch, comprising a multi-source signal acquisition module, a signal preprocessing and fusion module, a fault identification and control decision module, a control execution and feedback module; the multi-source signal acquisition module is integrated in the primary and secondary fusion switch, comprising a fault characteristic sensing unit and an auxiliary state sensing unit, responsible for real-time acquisition of discharge fault characteristic signals and auxiliary state signals; the signal preprocessing and fusion module is connected to the multi-source signal acquisition module, performs time and space alignment and standardization processing to generate fusion characteristic signals; the fault identification and control decision module internally pre-trains a fault identification and evaluation model and a pre-set control strategy library, receives fusion characteristic signals to generate comprehensive judgment signals and precise control execution signals; the control execution and feedback module drives the primary and secondary fusion switch to execute control actions, acquires line state feedback signals, and supports online adaptive updating of the model and the control strategy library. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0017] Figure 1 The method flowchart of the discharge fault control method of the power distribution line based on the primary and secondary fusion switch; Figure 2 The system architecture diagram of the discharge fault control system of the power distribution line based on the primary and secondary fusion switch. DETAILED DESCRIPTION

[0018] The embodiments of the present application will be described below through specific concrete examples, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the present description. The present application can also be implemented or applied through other different specific embodiments, and each detail in the present description can be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0019] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and the diagrams only show the components related to the present application, not the number, shape and size of the components when actually implemented. The actual implementation of each component may be a random change in type, number and proportion, and the layout pattern of the components may also be more complex.

[0020] In the following description, numerous specific details are discussed in order to provide a thorough explanation of embodiments of the application. However, it will be apparent to one of ordinary skill in the art that embodiments of the application can be practiced without these specific details. In other instances, well-known structures and devices are not described in detail in order to avoid obscuring embodiments of the application.

[0021] See Figure 1 The power distribution line discharge fault control method based on the primary and secondary fusion switch of the application comprises: S1: collecting the discharge fault characteristic signal of the power distribution line through the sensing component of the primary and secondary fusion switch, and synchronously collecting at least one auxiliary state signal reflecting the line operation state to form multi-source time sequence data; S2: performing space-time alignment and standardization processing on the discharge fault characteristic signal and the auxiliary state signal to eliminate signal dimension difference and time deviation, and generating a fusion characteristic signal; S3: inputting the fusion characteristic signal into a pre-trained fault identification and evaluation model, the model extracts deep features related to discharge faults in the signal, and generates a comprehensive judgment signal containing fault type, fault location and danger level; S4: calling a preset control strategy library based on the comprehensive judgment signal, matching the control logic suitable for the current fault scene, and generating a precise control execution signal of the primary and secondary fusion switch; S5: driving the primary and secondary fusion switch to execute corresponding control actions according to the control execution signal, simultaneously collecting the line state feedback signal after control, and using the signal to perform online adaptive update on the fault identification and evaluation model and the control strategy library, continuously optimizing the control effect.

[0022] As Figure 1As shown, the core process of the method revolves around the precise control of distribution line discharge faults. Through five closely linked steps, a complete fault control system is established, realizing a full closed-loop processing from signal acquisition to dynamic optimization. The first step is the acquisition of multi-source time series data. Relying on the sensing components integrated in the primary and secondary fusion switch, two types of key signals are captured synchronously, forming multi-source time series data that comprehensively reflects the line state. As the core carrier of data acquisition, the sensing component integrates detection units for direct fault characteristics and auxiliary detection units reflecting the background state of line operation, ensuring the comprehensiveness and relevance of data acquisition. Among them, the discharge fault characteristic signal is directly related to the core physical and electrical changes during fault occurrence, and is the key basis for identifying faults. The auxiliary state signal provides supplementary information from external environment, line operation load, and equipment state, etc. The synchronous acquisition of the two breaks the limitations of single signal acquisition. In practical application, this synchronous acquisition mechanism is realized through high-precision clock synchronization technology, ensuring that different types of signals strictly correspond in the time dimension, establishing a reliable time and space benchmark for subsequent analysis, and avoiding misjudgment of fault characteristics caused by asynchronous data. Through this step, relevant information can be captured comprehensively from the initial stage of fault occurrence, laying a solid data foundation for subsequent precise processing. Whether it is a sudden arc grounding fault or a slowly developing partial discharge fault, it can be effectively captured through the coordinated acquisition of multi-source signals. Focusing on the preprocessing and fusion of multi-source time series data, the original multi-dimensional signals are converted into unified and standardized fusion feature signals through time and space alignment and standardization processing, eliminating the influence of various interference factors on signal quality. The core goal of time and space alignment processing is to solve the time deviation problem caused by differences in sampling frequency and data format of different sensing units. By establishing a unified time reference, all signals are anchored on the same time axis. Specifically, a high-precision time reference standard is first determined, usually taking the clock of the fault characteristic signal with the highest sampling frequency as the reference. For auxiliary state signals with lower sampling frequency, linear interpolation or spline interpolation algorithms are used for data completion to ensure that there are corresponding multi-source signal data at each time node. At the same time, sliding time window technology is used for segmented processing of the synchronized signals. The length of each time window is set according to the typical time characteristics of fault development, ensuring that it can contain complete fault characteristic segments and avoiding processing delay caused by excessive data volume. After time and space alignment, further standardization processing is carried out. For each type of signal in each time window, detrending and standardization conversion operations are performed. Detrending processing is mainly used to eliminate slow drift components in the signal. These drifts may be caused by sensor aging, slow changes in environmental temperature, etc. By fitting the signal trend with the least squares method and removing it, the mutation characteristics caused by faults can be highlighted.The standardized conversion maps signals of different dimensions and amplitude ranges to a unified numerical interval through mathematical transformation. For example, using Z-score standardization to convert signals to a distribution with a mean of 0 and a standard deviation of 1 effectively solves the problem of the order of magnitude difference between transient current signals (amplitude may reach thousands of amperes) and temperature and humidity signals (amplitude is only a few digits), avoiding the dominance of signals with large amplitudes in subsequent processing, which can mask critical weak fault features. The fusion feature signals generated after this series of processing not only achieve synchronization in time and unity in dimension, but also provide high-quality and reliable input data for subsequent fault identification by strengthening signal features, significantly improving the accuracy and efficiency of subsequent processing steps.

[0023] Further, intelligent identification and comprehensive evaluation of faults are achieved by inputting the fusion feature signals into a pre-trained model to deeply mine and accurately judge multi-dimensional information of faults. The model, as the core intelligent processing unit, has the ability to extract deep and nonlinear features from complex fusion signals. Its design goal is to break through the limitations of traditional recognition based on fixed thresholds or simple linear models and adapt to the complex and variable characteristics of discharge fault features. After receiving the fusion feature signals, the model first extracts local fault features from the signals. These local features may include sudden peak values of transient current, specific frequency components of ultrasonic signals, phase mutations of transient voltage, etc. These features are the key basis for distinguishing different fault types. Based on local feature extraction, the model further mines the time sequence dependence of the signals to capture the whole process change law of faults from occurrence, development to stability, such as the amplitude decay trend of local discharge signals over time and the oscillation characteristics of current signals in arc grounding faults. Through analysis of time sequence features, the development stage and severity of faults can be more accurately judged. At the same time, the model also has the ability of feature fusion and dynamic weighting. It automatically adjusts feature weights according to the importance differences of various features in different fault scenarios, strengthens key features that play a decisive role in fault identification, and suppresses the influence of irrelevant interference features. Through this series of feature extraction and analysis processes, the model finally outputs a comprehensive judgment signal containing three core dimensional information: accurate classification of fault types, which can clearly distinguish different types of discharge faults such as arc grounding fault, insulator flashover, and cable partial discharge, providing a basis for subsequent matching of targeted control strategies; accurate positioning of fault location, which can narrow down the fault range to specific line sections or device nodes, reducing fault troubleshooting time; and quantitative evaluation of danger level, which reflects the severity and potential harm of the fault through numerical means, providing a quantitative standard for the hierarchical execution of control strategies. This step realizes all-around and high-precision judgment of discharge faults through the deep involvement of the intelligent model, completely changing the status quo of low accuracy and single information dimension of traditional fault identification methods, and providing solid decision support for subsequent precise control.

[0024] In an embodiment of the present application, the acquisition link of the multi-source signal achieves comprehensive coverage of fault information through rich signal dimensions. Different types of signals capture fault-related features from their respective perspectives, forming mutually complementary and mutually verifying data sources. Discharge fault characteristic signals directly reflect the core physical and electrical changes of the fault. Transient current signals can capture the current mutation at the moment of fault occurrence, and their waveform characteristics are closely related to the fault type and severity. For example, the transient current of arc grounding fault usually presents obvious sharp pulse characteristics, while the transient current of cable partial discharge presents a continuous sequence of small pulses. Ultrasonic signals are formed by the mechanical vibration generated by the medium breakdown during discharge, and their propagation path and intensity can reflect the position and energy of the discharge, which is of great significance for partial discharge faults that are difficult to detect directly through electrical signals. Transient voltage signals record the amplitude and phase changes of the line voltage before and after the fault, and different fault types will cause the voltage to present different characteristics of drop, oscillation or distortion. Auxiliary state signals do not directly reflect the fault itself, but can provide background information for the occurrence of the fault. Environmental temperature and humidity signals can reflect the influence of the external environment on the line insulation performance. High humidity environment can reduce the surface insulation resistance of the insulator, easily causing flashover fault, while high temperature environment can accelerate the aging of cable insulation and increase the risk of partial discharge. Line load current signals can distinguish between fault and normal load fluctuation, avoiding misjudgment of current mutation during large load start-up as a fault. Switch mechanical state signals can monitor the running state of the primary and secondary fusion switch itself, ensuring the reliability of subsequent control actions. If the switch mechanical structure has problems such as jamming, even if the fault is recognized, the control action cannot be effectively executed, so the collection of this signal provides a prerequisite guarantee for the effectiveness of control execution. The synchronous collection of these signals enables the fusion characteristic signal to comprehensively cover the electrical characteristics, physical characteristics, environmental factors and equipment state information of the fault, laying a foundation for subsequent accurate identification and control. Whether the fault is caused by electrical insulation damage, environmental factors or equipment aging, it can be effectively identified through the coordinated analysis of multi-dimensional signals. The time and space alignment and standardization processing link eliminates various disturbances and inconsistencies in the signal through fine operation, providing high-quality input for model processing. In the time and space alignment process, the unification of the time reference ensures the relevance of different signals in the time dimension. For example, the transient current mutation, the generation of ultrasonic signals and voltage drop at the moment of fault occurrence can be accurately corresponded. This temporal synchronization enables the model to accurately analyze the causal relationship between different features and avoid feature mismatching due to time deviation. The application of sliding time window converts continuous time series signals into fixed-length data segments that are convenient for model processing. Each data segment contains complete fault feature information within a certain time period, ensuring the integrity of the features and reducing the computational complexity of the model.The detrending operation in the standardization process effectively eliminates systematic errors caused by factors such as sensor drift and slow environmental changes. For example, sensor output shifts caused by a slow increase in temperature may be misjudged as fault features by the model if not eliminated. The standardization transformation solves the problems caused by differences in the dimensions and amplitudes of different signals, allowing different types of signals such as transient currents, ultrasound, and temperature / humidity signals to be analyzed at the same scale. The model no longer needs to adapt to the amplitude ranges of different signals and can focus more on pattern recognition of the features themselves. Through these processing steps, the original multi-source signals are transformed into fused feature signals with uniform scale, prominent features, and spatiotemporal consistency, greatly improving the accuracy and robustness of the model's recognition. Even in complex scenarios with noise interference and inconsistent sampling frequencies, it can ensure the effective extraction and identification of fault features.

[0025] In an embodiment of the present application, the fault identification and evaluation link of the model realizes accurate judgment of complex faults through the powerful feature extraction and modeling capability of the deep neural network. The extraction capability of the model for local features enables it to capture weak fault signals that are difficult to detect by traditional methods, such as the tiny current pulses and ultrasonic signals generated by early local discharge of the cable. These signals are often masked by background noise, but the model can effectively filter noise and strengthen fault features through convolution operations and other operations. The capture of time sequence dependency enables the model to understand the dynamic development process of the fault, such as the process from initial weak discharge to complete breakdown of the insulator flashover fault. The model can learn the evolution law through a recurrent network structure, thereby realizing the judgment of the fault development stage. The introduction of the attention mechanism further improves the recognition accuracy of the model. By dynamically adjusting the feature weight, the model can focus on the core features that play a key role in fault recognition, such as the sudden drop feature of the transient voltage and the peak feature of the transient current in the arc grounding fault. The model will automatically increase the weight of these features and weaken the influence of auxiliary signals such as temperature and humidity. For the insulator flashover fault caused by environmental humidity, the weight of the temperature and humidity signal will be correspondingly increased. The output of the comprehensive judgment signal quantizes and presents the type, location and danger level of the fault, providing clear basis for subsequent control strategy matching. This avoids the limitation of traditional methods that can only judge whether a fault occurs or not, but cannot provide more detailed information, greatly improving the pertinence and accuracy of control actions. The entire process from multi-source signal collection to feature processing to intelligent recognition forms a complete fault perception and judgment system, with each link closely connected and mutually supported. Through the cooperation of multi-dimensional signals, fine signal processing and intelligent model analysis, the overall perception and accurate judgment of the discharge fault of the distribution line are realized, laying a solid foundation for subsequent control execution and dynamic optimization, and completely solving the problems of low accuracy, incomplete information and poor adaptability in the fault identification link of traditional methods.

[0026] Further, the fault identification and evaluation model, as the core intelligent unit of the entire control method, adopts a hybrid neural network architecture based on deep learning, and realizes the deep mining and accurate mapping of complex discharge fault features through the collaborative work of multiple modules. This model integrates three core functional modules, which correspond to different dimensional requirements of feature extraction, forming a complete feature processing link from local to global and from static to dynamic. First is the convolutional neural network module, whose core function is to extract spatial features from the local segments of the fused feature signals. These features usually represent local key information such as mutation peaks, specific frequency components, and waveform inflection points in the signal, and are the basis for distinguishing different fault types. The convolutional neural network module is stacked with multiple convolutional layers and pooling layers. The convolutional layer performs sliding operations on the signal in the time dimension through learnable convolution kernels, automatically capturing local features of different scales. For example, the first layer of convolution kernels can capture high-frequency spikes in the transient current signal, and deep convolution kernels can extract more complex combined features. The pooling layer reduces the dimension of the convolution features through max-pooling or average-pooling operations, retaining key features while reducing computational complexity and improving model operation efficiency. After processing by the convolutional neural network module, the original one-dimensional time series signal is converted into a high-dimensional feature vector containing rich local features, laying the foundation for subsequent time series analysis. Next, the recurrent neural network module takes the output of the convolutional neural network and focuses on capturing the long-term dependencies of the feature vector in the time dimension. The occurrence and development of discharge faults is a dynamic process, and fault features will exhibit specific evolution patterns over time. For example, the amplitude of partial discharge signals will gradually increase with the degree of insulation deterioration, and the current signal of arc grounding faults will exhibit a decaying oscillation time series characteristic. These dynamic information cannot be obtained by simply extracting local features. The recurrent neural network module can selectively remember long-term effective feature information and forget irrelevant short-term interference through the introduction of a gating mechanism, thus accurately modeling the time evolution trend of fault features. In practical applications, the recurrent neural network module usually adopts a long short-term memory network or a gated recurrent unit structure, which can effectively solve the gradient vanishing problem of traditional recurrent neural networks and ensure the modeling ability of long sequence signals. Through the processing of the recurrent neural network module, the feature vector is assigned with context information in the time dimension, and the model can understand the complete process from the occurrence, development to stability of the fault, providing key support for fault location and risk level assessment.

[0027] Specifically, the attention mechanism module integrates and dynamically weights the local features extracted by the convolutional neural network and the time sequence features captured by the recurrent neural network. Under different fault scenarios, the contribution of various features to fault identification varies significantly. For example, in arc grounding fault, the sudden drop of transient voltage and the sharp peak of transient current play a dominant role, while in insulator flashover fault, the influence of environmental temperature and humidity features and ultrasonic signal features is more prominent. The attention mechanism module automatically identifies the key features in the current fault scenario by calculating the attention weight of each feature dimension, and gives higher weight coefficients, while suppressing the influence of secondary features and interference features. The weight calculation is based on multiple indicators such as the activation strength of the feature and the similarity to historical fault features, and the weight normalization is realized through a fully connected layer and a softmax function. After optimization by the attention mechanism module, the fused feature vector can focus on the most valuable information for current fault judgment, greatly improving the accuracy of the model output. Finally, the fully connected layer maps the weighted and fused deep features to the target output space to generate a comprehensive judgment signal containing fault type classification, fault location coordinates, and danger level quantization value, achieving synchronous and accurate identification of multi-dimensional information of discharge faults. The pre-training process of the model needs to be completed in a controlled test environment. By building a fault simulation platform close to actual working conditions, a large number of high-quality training samples are obtained to ensure the model has good generalization ability. The test platform needs to completely restore the topology, line parameters and operating environment of the distribution line, including the mixed wiring mode of overhead lines and cables, the deployment of different types of insulators, and the key elements such as load variation range. The fault simulation unit can generate discharge faults of different types, locations and severity by adjusting the parameters of the fault trigger device, such as simulating arc grounding faults of different severity by adjusting the discharge energy of the arc generator, simulating insulator flashover faults by changing the pollution level and humidity conditions on the surface of the insulator, and simulating partial discharge faults by setting defects at different positions of the cable. During the fault simulation process, high-precision reference detection equipment is deployed simultaneously, including millimeter-level precision fault locators, picocoulomb-level sensitive partial discharge detectors, and high-speed oscilloscopes, which can provide real labels of fault type, location, and danger level as the target reference for model training.

[0028] The construction of the training sample set needs to cover all common discharge fault types of the power distribution line, and each fault type needs to include sample data under different load conditions, environmental conditions and equipment states to ensure the diversity and representativeness of the samples. For each group of samples, the sensor components of the primary and secondary fusion switch synchronously collect the discharge fault characteristic signals and auxiliary state signals, generate the fusion characteristic signals after time and space alignment and standardization processing, and form complete training sample pairs with the real labels provided by the reference detection device. In order to improve the robustness of the model, noise interference samples also need to be added to the sample set to simulate the electromagnetic interference, sensor noise and other situations that may occur in actual operation. The training sample set is divided into training set, validation set and test set in proportion, the training set is used for model parameter learning, the validation set is used for adjusting model hyperparameters and preventing overfitting, and the test set is used for evaluating the final performance of the model. The model training adopts a supervised learning method, takes the fusion characteristic signals as the input and the real labels as the target, and adjusts the model parameters through iterative optimization. The design of the loss function needs to consider the needs of multi-task learning, and the cross-entropy loss function is used for the fault type classification task to measure the difference between the predicted class and the real class; the mean square error loss function is used for the fault location positioning and danger level evaluation task to quantify the deviation between the predicted value and the real value. In the training process, an adaptive optimizer is used to dynamically adjust the learning rate according to the gradient information of the model parameters to speed up the model convergence; at the same time, an early stopping mechanism and a regularization technique are introduced to stop training when the validation set loss does not decrease for consecutive multiple iteration periods to avoid model overfitting and ensure that the model still maintains good performance on unseen test data. Through large-scale sample training, the model can gradually learn the complex nonlinear mapping relationship between the fusion features and the multi-dimensional fault information, and finally has the ability to accurately extract fault features from disturbed multi-source signals and output reliable comprehensive judgment signals.

[0029] As shown in Figure 2 The present application also provides a discharge fault control system for a power distribution line based on a primary and secondary fusion switch, which comprises a multi-source signal acquisition module, a signal preprocessing and fusion module, a fault identification and control decision module, and a control execution and feedback module. The multi-source signal acquisition module is integrated in the primary and secondary fusion switch and comprises a fault characteristic sensing unit and an auxiliary state sensing unit, which are responsible for real-time acquisition of discharge fault characteristic signals and auxiliary state signals. The signal preprocessing and fusion module is connected to the multi-source signal acquisition module and performs time and space alignment and standardization processing to generate fusion characteristic signals. The fault identification and control decision module is built-in with a pre-trained fault identification and evaluation model and a pre-set control strategy library, receives fusion characteristic signals to generate comprehensive judgment signals and accurate control execution signals. The control execution and feedback module drives the primary and secondary fusion switch to execute control actions and collects line state feedback signals to support online adaptive updating of the model and the control strategy library.

[0030] Specifically, the control strategy matching link based on the comprehensive judgment signal is the core of the hierarchical design and dynamic adaptation mechanism of the preset control strategy library, which ensures that the control action can accurately match the needs of different fault scenarios. The control strategy library is a structured strategy set constructed based on the distribution line fault handling procedures, operation experience and simulation analysis results. Its core design concept is "classified strategy, hierarchical control", which constructs a multi-dimensional strategy matching index according to the three core dimensions (fault type, fault location, danger level) in the comprehensive judgment signal. Among them, the danger level is the core basis for strategy stratification, which divides the control strategy into three levels: emergency control, regular control and early warning control. Each level corresponds to a specific fault risk range and control objective. The emergency control strategy is for high-risk level faults. Such faults usually develop rapidly and are highly destructive, such as severe arc grounding faults and cable phase-to-phase discharge faults, which can cause equipment explosions, line fires or large-scale power outages if not handled in time. The core objective of this level of strategy is "quick isolation and prevention of spread", and the control logic uses a "quick tripping + fault isolation + load transfer" linkage mechanism. After receiving the control signal, the primary and secondary fusion switch needs to start the quick tripping mode, which shortens the tripping time to milliseconds by optimizing the driving voltage of the tripping coil and the action timing of the mechanical structure, ensuring that the fault line and the system are disconnected before the fault energy expands; at the same time, the system quickly determines the fault isolation range based on the line topology information, sends tripping instructions to adjacent switch devices, and forms a closed fault isolation area to prevent the fault from spreading to non-fault sections; for important loads on the fault line, the linkage standby line switch completes load transfer at the same time as fault isolation, minimizing power loss. The regular control strategy is suitable for medium-risk level faults, which develop relatively slowly and have limited damage, such as minor insulator flashover and early-stage cable partial discharge faults. The core objective is "precise disposal and power restoration". The control logic uses a "precise tripping + coordinated reclosing" mode, in which the switch precisely cuts off the fault section based on fault location information to avoid unnecessary large-scale power outages; after tripping, the reclosing operation is performed after a preset time interval, and if reclosing is successful, it indicates that the fault is transient, and the line is restored to normal power supply; if reclosing fails, it is determined to be a permanent fault, and the fault isolation program is immediately started, and the fault alarm information is sent to the operation and maintenance platform to prompt the staff to handle the scene. The key to this level of strategy is the dynamic adjustment of the reclosing interval time, which automatically matches the optimal interval time according to the fault type and line load state, avoiding premature reclosing that causes fault expansion or unnecessary power outage duration.

[0031] The power distribution line discharge fault control method and system based on a primary and secondary fusion switch of the present application synchronously collect discharge fault characteristic signals and line auxiliary state signals through a multi-source sensing assembly of the primary and secondary fusion switch, generate fusion characteristic signals through space-time alignment and standardization processing, input a pre-trained AI model to extract deep features, output comprehensive judgment signals of fault type, position and danger level, match a layered control strategy library to generate accurate control signals, drive the switch to execute actions, and simultaneously realize online adaptive updating of the model and the strategy library through feedback signals, thereby efficiently solving the fault management and control problem.

[0032] Therefore, the power distribution line discharge fault control method and system based on a primary and secondary fusion switch of the present application can solve the problems of slow discharge fault identification and poor control adaptability of the power distribution line.

[0033] The above embodiments only exemplarily illustrate the principles and effects of the present application, and are not used to limit the present application. Any person skilled in the art can modify or change the above embodiments without departing from the spirit and category of the present application. Therefore, all equivalent modifications or changes completed by those skilled in the art without departing from the spirit and technical thought disclosed by the present application should be covered by the claims of the present application.

Claims

1. A method for controlling discharge faults in distribution lines based on a primary and secondary integrated switch, characterized in that, include: S1: Collect discharge fault characteristic signals of the power distribution line through the sensing components of the primary and secondary integrated switch, and simultaneously collect at least one auxiliary status signal reflecting the line operation status to form multi-source time sequence data. S2: Perform spatiotemporal alignment and standardization processing on the discharge fault characteristic signal and the auxiliary state signal to eliminate signal dimension differences and time deviations, and generate a fused characteristic signal; S3: Input the fused feature signal into the pre-trained fault identification and evaluation model. The model extracts deep features related to the discharge fault in the signal and generates a comprehensive judgment signal containing the fault type, fault location and hazard level. S4: Based on the comprehensive judgment signal, call the preset control strategy library, match the control logic adapted to the current fault scenario, and generate a precise control execution signal for the primary and secondary fusion switch; S5: Drive the primary and secondary fusion switches to perform corresponding control actions according to the control execution signal, and at the same time collect the line status feedback signal after control. Use the signal to perform online adaptive updates of the fault identification and evaluation model and control strategy library to continuously optimize the control effect.

2. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch as described in claim 1, characterized in that, The discharge fault characteristic signals in step S1 include the fault transient current signal collected by the transient current sensor, the ultrasonic signal collected by the partial discharge ultrasonic sensor, and the transient voltage signal collected by the transient voltage sensor. The auxiliary status signals include the ambient temperature and humidity signal, the line load current signal, and the switch mechanical status signal. These signals capture fault information from multiple dimensions, including the electrical characteristics, physical characteristics, and external influencing factors of the fault occurrence. This allows the fused characteristic signals to comprehensively reflect the occurrence mechanism and evolution trend of the discharge fault, providing diversified data support for subsequent accurate fault identification.

3. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch as described in claim 1, characterized in that, The spatiotemporal alignment and standardization process in step S2 specifically includes unifying all time-series signals to the same time reference, achieving strict time synchronization of signals with different sampling frequencies through signal interpolation and timestamp calibration, segmenting the synchronized signals using a sliding time window to ensure spatiotemporal consistency of signals within each data segment, and performing detrending and standardization transformations on various types of signals within each data segment to eliminate the influence of dimensional differences between different sensing components and signal amplitude fluctuations, generating fused feature signals with uniform scale and prominent characteristics, providing stable and high-quality input data for fault identification and assessment models, and improving the accuracy and robustness of model identification.

4. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch as described in claim 1, characterized in that, The fault identification and assessment model is a hybrid neural network model based on deep learning. It includes a convolutional neural network module for extracting local fault features, a recurrent neural network module for capturing the temporal dependencies of signals, and an attention mechanism module for feature fusion and decision-making. The fused feature signal is first input into the convolutional neural network module to extract local detailed features of the fault, and the output result is sent to the recurrent neural network module to mine the temporal evolution law of the features. The attention mechanism module assigns dynamic weights to features of different dimensions to strengthen the influence of key fault features. Finally, a comprehensive judgment signal containing accurate fault type classification, fault location and hazard level quantitative assessment is output through a fully connected layer, realizing the synchronous identification of multi-dimensional information of discharge faults.

5. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch as described in claim 1, characterized in that, The training method of the pre-trained fault identification and evaluation model includes building a power distribution line discharge fault simulation platform in a controlled test environment to simulate discharge fault scenarios of different types, locations and severity. Various fault characteristic signals and auxiliary status signals are collected synchronously through the sensing components of the primary and secondary integrated switch. At the same time, high-precision fault detection equipment is used to obtain real labels of fault type, location and hazard level to construct a large-scale labeled training sample set. Using real labels as the training target, the hybrid neural network model is trained in a supervised manner by taking the fused feature signals as input. The model parameters are adjusted through iterative optimization so that the model learns the complex mapping relationship between the fused features and the multi-dimensional information of the fault until the model recognition error meets the preset accuracy requirements.

6. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch according to claim 1, characterized in that, The preset control strategy library in step S4 contains hierarchical control strategies for different fault types, locations and hazard levels. Based on the hazard level in the comprehensive judgment signal, it is divided into three levels: emergency control, normal control and early warning control. The emergency control strategy corresponds to high-hazard level faults and adopts the linkage control logic of fast tripping, fault isolation and load transfer. The conventional control strategy is for medium-risk faults, and executes precise tripping and reclosing control. The early warning control strategy is for low-risk faults, and activates the fault monitoring enhancement and operation and maintenance prompt linkage mechanism. The control strategy library also contains adaptive control parameters for different fault types and locations, ensuring that the control execution signal can accurately match the fault scenario and achieve differentiated and efficient management and control of discharge faults.

7. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch according to claim 1, characterized in that, The online adaptive update in step S5 specifically includes continuously monitoring the line status feedback signal after control. When the line is detected to be in a stable operating state or a completely cleared fault state, the current comprehensive judgment signal, control execution signal, and status feedback signal are correlated and analyzed to calculate the fault identification deviation and control effect evaluation index. If the evaluation index does not reach the preset optimal threshold, the fault identification and evaluation model is incrementally trained using the correlation analysis data, and the model parameters are dynamically adjusted to adapt to changes in line operating status and equipment characteristic drift. At the same time, the adaptation parameters in the control strategy library are fine-tuned based on the control effect feedback to optimize the control logic under different fault scenarios, so that the system can continuously adapt to the dynamic changes of the power distribution line and maintain efficient fault control performance.

8. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch according to claim 1, characterized in that, The method also includes fault prediction and early warning steps. In step S3, in addition to generating a comprehensive judgment signal, the fault identification and assessment model also generates a fault development prediction signal based on the trend changes of the fused feature signal. This signal includes the time window for possible fault escalation and the trend of risk level changes. The system starts the control strategy preloading mechanism in advance according to the fault development prediction signal. When the fault risk level reaches the preset threshold, the corresponding control action is automatically triggered. At the same time, an early warning information is sent to the operation and maintenance management platform to realize the forward-looking management and control of discharge faults and reduce the losses caused by fault expansion.

9. The method for controlling discharge faults in distribution lines based on a primary and secondary fusion switch according to claim 1, characterized in that, The method also includes a control process safety verification step. After the control execution signal is generated in step S4, the system verifies the execution conditions, action sequence, and safety boundaries of the control signal. The verification content includes the feasibility of line load transfer, the reliability of switch mechanical action, and the rationality of fault isolation range. If the verification passes, the control execution signal is issued. If the verification fails, the control strategy parameters are adjusted and the control signal is regenerated, or a backup control scheme is triggered. At the same time, the verification anomaly information is recorded for subsequent system optimization to ensure the safety and reliability of control actions and avoid secondary faults caused by improper control.

10. A power distribution line discharge fault control system based on a primary and secondary integrated switch, characterized in that, The power distribution line discharge fault control method based on primary and secondary fusion switch according to any one of claims 1-9 includes a multi-source signal acquisition module, a signal preprocessing and fusion module, a fault identification and control decision module, and a control execution and feedback module. The multi-source signal acquisition module is integrated into the primary and secondary fusion switch, including a fault feature sensing unit and an auxiliary status sensing unit, which is responsible for real-time acquisition of discharge fault feature signals and auxiliary status signals. The signal preprocessing and fusion module is connected to the multi-source signal acquisition module, and performs spatiotemporal alignment and normalization processing to generate fused feature signals; The fault identification and control decision module has a built-in pre-trained fault identification and evaluation model and a preset control strategy library. It receives fused feature signals to generate comprehensive judgment signals and precise control execution signals. The control execution and feedback module drives the primary and secondary fused switches to perform control actions, collects line status feedback signals, and supports online adaptive updates of the model and control strategy library.

Images