A cable state evaluation method and system based on a fault arc model

Abstract

This invention discloses a cable condition assessment method based on a fault arc model, belonging to the field of power equipment condition monitoring and fault diagnosis technology. The method includes the following steps: S1: Real-time monitoring of the electrical signals of the target cable, capturing and recording the original waveform data within a preset time window before and after the occurrence of a fault arc; S2: Preprocessing the original waveform data and extracting a multi-dimensional feature parameter set to characterize the arc characteristics and cable condition; S3: Inputting the multi-dimensional feature parameter set into a pre-trained cable condition assessment model; S4: Generating a current condition assessment result and maintenance recommendations for the target cable based on the output of the cable condition assessment model. This invention, through multi-dimensional intelligent analysis of fault arc signals, transforms arc events from a single fault indication into a dynamic indicator for assessing the health of cable insulation, achieving accurate diagnosis and trend prediction of cable condition.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring and fault diagnosis technology, specifically relating to a cable condition assessment method and system based on a fault arc model. Background Technology

[0002] Currently, cables, as a critical carrier of electrical energy transmission, are significantly affected by insulation degradation, a major cause of power system failures. Arc faults are a typical manifestation of cable insulation defects; persistent arc faults accelerate insulation aging, ultimately leading to serious accidents such as short circuits and fires. Existing cable condition monitoring technologies primarily rely on methods such as partial discharge detection, infrared thermography, and distributed fiber optic sensing. For arc faults, existing technologies (such as Arc Fault Circuit Interrupters (AFCIs) mainly focus on the immediate detection and circuit interruption of arcs to achieve electrical fire protection.

[0003] Publication No. CN120370083B discloses a method, apparatus, and non-volatile storage medium for cable condition assessment. The method includes: constructing a fault arc model and a circuit model corresponding to the target cable; generating simulation waveforms based on the fault arc model and the circuit model; determining target fault parameters based on the measured waveforms and simulation waveforms of the fault points in the target cable, wherein the target fault parameters characterize index data describing fault characteristics and conditions; determining the waveform characteristics of the fault points according to the target fault parameters; and determining the condition assessment result of the target cable based on the waveform characteristics.

[0004] Publication No. CN112487714B discloses a method for generating a decision tree model for identifying the fire status of a cable shaft, comprising: Step S1, collecting the residual current and residual voltage values ​​of the cable in the cable shaft, obtaining the resistive residual current component in the residual current, and obtaining the fault arc characteristics of the residual voltage; Step S2, collecting the ultrasonic signals generated by partial discharge in the cable shaft, and determining the size and location of the partial discharge based on the ultrasonic signals, as partial discharge information; Step S3, collecting the ambient temperature, detection point temperature, and smoke concentration in the cable shaft; Step S4, evaluating the feature items according to preset rules and generating evaluation results; and inputting the evaluation results and the corresponding input features as training data into a preset model for training to generate a decision tree model.

[0005] The technical solutions disclosed in the aforementioned patents all rely on the time-domain characteristics of the current waveform for threshold judgment. These solutions are susceptible to interference from nonlinear loads such as motor starting and thyristor devices, leading to false alarms or missed alarms. Furthermore, they cannot distinguish between arc types and severity, and cannot extract effective information from arc signals for assessing the overall aging state and insulation degradation trend of cables. They are passive protection methods lacking predictive maintenance capabilities. Therefore, a cable condition assessment method and system based on a fault arc model is needed. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a cable condition assessment method and system based on a fault arc model. By performing multi-dimensional and in-depth feature extraction and intelligent analysis on fault arc signals, the arc event is transformed from a single fault indication into a dynamic indicator for assessing the health status of cable insulation, thereby achieving accurate diagnosis, trend prediction, and proactive maintenance of cable condition and solving the technical problems mentioned in the background.

[0007] The objective of this invention is achieved as follows: a cable condition assessment method based on a fault arc model, comprising the following steps: S1: real-time monitoring of the electrical signals of the target cable, capturing and recording the original waveform data within a preset time window before and after the occurrence of a fault arc; S2: preprocessing the original waveform data and extracting a multi-dimensional feature parameter set from it to characterize the arc characteristics and cable condition; S3: inputting the multi-dimensional feature parameter set into a pre-trained cable condition assessment model; S4: generating a current condition assessment result and maintenance recommendations for the target cable based on the output of the cable condition assessment model.

[0008] The working principle is as follows: The fault arc signal is transformed from a traditional fault alarm signal into an information carrier that can be used for in-depth analysis of the cable insulation health status. By capturing waveform data of the entire arc event process and extracting multi-dimensional features that comprehensively reflect the physical characteristics of the arc and the cable insulation status, a comprehensive evaluation is performed using an intelligent model, thus achieving a shift from passive protection to proactive early warning. In implementation, the first step (S1) involves continuously monitoring the cable current / voltage using sensors. Once a possible arc disturbance is detected, the complete waveform before and after the event is recorded with high precision. The second step (S2) involves cleaning and aligning the recorded raw waveforms, and extracting a set of quantitative feature values ​​using various signal analysis algorithms (such as time-frequency analysis). The third step (S3) involves inputting these feature values ​​into a pre-trained intelligent evaluation model (such as a machine learning model). The fourth step (S4) involves the model outputting quantitative evaluation results (such as health score and risk level) and specific maintenance recommendations. This achieves intelligent and quantitative evaluation of cable status, breaking through the limitations of traditional technologies that can only determine the presence or absence of an arc, and providing direct data decision support for predictive maintenance.

[0009] In step S1, the electrical signals include current signals and voltage signals. Both current and voltage sensors need to be installed during deployment. During data processing, a digital filter is first used to remove high-frequency noise and power frequency interference. Then, the signal is segmented according to the power frequency cycle, and an algorithm is used to accurately locate the start time of each arc pulse, providing clean and time-aligned standard data segments for subsequent feature analysis.

[0010] In step S2, the preprocessing includes filtering and denoising, power frequency cycle segmentation, and precise calibration of the arc pulse start point. The multi-dimensional feature parameter set includes time-domain features, frequency-domain features, and time-frequency-domain features. The time-domain features include the rise rate, fluctuation entropy, zero-crossing distortion rate, and correlation coefficient with the standard arc waveform library of the current waveform. The frequency-domain features include the energy ratio of specific harmonic components and the ratio of high-frequency noise energy to fundamental wave energy. The time-frequency-domain features include the energy distribution of each component and singular spectral entropy obtained through wavelet transform or empirical mode decomposition. Since single-type features (such as only considering current amplitude) are prone to misjudgment and lack sufficient information, a more comprehensive and stable feature fingerprint is constructed by fusing features from the three dimensions of time domain (waveform shape), frequency domain (harmonic components), and time-frequency-domain (time-varying characteristics), thereby more accurately characterizing the complex physical processes of the arc and the insulation degradation mode of the cable. In practice, calculation tools are used to calculate from the waveform: the rate of change of current in the time domain (rise rate), the degree of waveform disorder (wave entropy), the zero-crossing anomaly (zero-crossing distortion rate), and the similarity to typical fault waveforms (correlation coefficient); the proportion of harmonic content and the noise energy of specific high-frequency bands in the frequency domain; and the distribution characteristics of signal energy at different times and frequencies are analyzed in the time and frequency domains using tools such as wavelet transform.

[0011] In step S3, the method for constructing and training the cable condition assessment model includes: S31: Obtaining a historical cable sample set, which contains the multi-dimensional feature parameter set corresponding to multiple cables of different aging states and specifications under various load conditions when fault arcs are induced, as well as the real state labels of each cable sample determined through offline detection; S32: Using the multi-dimensional feature parameter set as input features and the corresponding real state labels as supervision targets, training and optimizing a selected machine learning model or deep learning model to obtain a preliminary assessment model; S33: Using cable operating environment parameters and historical arc event sequence features as auxiliary inputs, fusing them with the multi-dimensional feature parameter set to enhance the preliminary assessment model, obtaining the cable condition assessment model capable of performing time-series correlation analysis. The working principle is as follows: The model is trained using massive amounts of historical fault data covering various operating conditions, enabling it to learn the complex mapping relationship from features to the actual cable state. Then, environmental parameters and historical event sequences are introduced, allowing the model to understand that the cable state is a process of dynamic evolution over time in a specific operating environment, thereby achieving a more realistic trend judgment. In implementation, firstly, a large number of samples are collected in the laboratory or on-site to establish a correspondence between feature sets and real-state labels for each cable sample. Then, this data is used to train a basic machine learning model (such as a random forest or neural network). Finally, information such as the cable's operating load, temperature and humidity, and historical arc events are integrated into the basic model to enhance its training, enabling it to perform time-series analysis by combining historical data with current conditions. Through supervised learning, the model possesses reliable evaluation capabilities; by integrating environmental and historical information, the model exhibits adaptability and predictive ability, enabling it to assess the current state's position within historical trends, thereby more accurately predicting future degradation rates and making the evaluation results more closely aligned with actual operating conditions.

[0012] In step S4, the status assessment results include at least one of the following: cable insulation health level score, estimated remaining reliable operating time, and location of the most likely deterioration point. Following step S4, the system further includes the following steps: S5: Establishing a full lifecycle status file for the target cable, storing the feature parameter set and status assessment results obtained from each assessment in a time series within the status file; S6: Based on the data in the status file, analyzing the cable status degradation trend, and generating graded early warning information when the status degradation rate exceeds a preset threshold or the assessment result is lower than a safety threshold. In implementation, the system establishes an independent electronic file for each cable, storing the features and results of each assessment in chronological order. The system continuously analyzes this file, for example, calculating the decline slope of the insulation health score. Once accelerated degradation (slope exceeding the threshold) or a score below the safety line is detected, different levels of early warning (such as attention, abnormal, severe) are automatically triggered. After outputting the assessment results, this data is further utilized for full lifecycle management and trend early warning. The principle is to place the results of a single assessment into a time series, and by analyzing the rate of change of status parameters over time, to achieve quantitative tracking and early warning of cable aging trends. Maintenance personnel can identify which cables are aging rapidly in advance, thus enabling them to scientifically plan maintenance schedules and allocate resources, greatly improving the initiative and economy of power grid operation and maintenance.

[0013] A cable condition assessment system based on a fault arc model, used to implement the aforementioned method, includes: a data acquisition module for real-time monitoring and high-precision sampling of current and voltage signals of the target cable; an arc detection and data caching module for real-time detection of fault arc events from the sampled signals and capturing raw waveform data for a set time period before and after the event; a feature extraction module for preprocessing the cached raw waveform data and calculating and generating the multi-dimensional feature parameter set; a condition assessment engine module, embedded with the pre-trained cable condition assessment model, for receiving the feature parameter set, performing calculations and analysis, and outputting condition assessment results; and a result output and early warning module for visually displaying the assessment results, generating maintenance reports, and issuing early warnings in case of anomalies. In implementation, the data acquisition module (sensor) is responsible for acquiring signals; the arc detection and data caching module (edge ​​computing unit) is responsible for real-time judgment and capturing valid data segments; the feature extraction module (computing server) is responsible for running complex feature calculation algorithms; the condition assessment engine module (core server) loads the model and performs inference; and the result output and early warning module (human-machine interface) is responsible for display and alarms.

[0014] It also includes: a model update and optimization module, used to receive feedback on the actual cable status after field verification, and to perform incremental learning or periodic retraining on the cable status assessment model; and a cloud-based collaborative platform, used to receive assessment data from multiple systems, perform big data aggregation and analysis to optimize and distribute a globally applicable assessment model, and provide data support for regional performance cable network status analysis. During implementation, the model update and optimization module will use the predicted and verified data pairs as new samples after each field inspection confirms the actual cable status, and fine-tune the model (incremental learning). The cloud-based collaborative platform gathers data from hundreds or thousands of field systems, trains a more powerful global model, and then distributes it to each field system. It also provides a comprehensive cable health report for the regional power grid, solving the problems of limited data from a single system and insufficient model generalization ability.

[0015] The data acquisition module includes: a distributed wideband Rogowski coil with a frequency response range of 0.1Hz-10MHz; a high-precision differential voltage probe, an embedded sensor integrated at the cable connector; and a synchronization clock unit to ensure time alignment accuracy of data acquired from multiple endpoints. Fault arc signals contain rich high-frequency transient components that ordinary power frequency sensors cannot capture. Therefore, wideband, high-precision dedicated sensors must be used, and strict time synchronization of multi-point measurements must be ensured to guarantee the accuracy of waveform analysis. In implementation, Rogowski coils and embedded voltage probes are installed at key monitoring points on the cable, their high-frequency response ensuring the capture of MHz-level arc high-frequency signals. The synchronization clock unit, such as one based on GPS or a precision clock source, ensures that all sensor-acquired data has a unified timestamp, facilitating advanced analysis such as traveling wave localization.

[0016] The cloud-based collaborative platform includes a group health status prediction unit and a cross-line knowledge transfer unit. The cross-line knowledge transfer unit is used to perform adversarial neural network transfer learning on the degradation mode characteristics of different cable lines. Through transfer learning, especially the application of adversarial neural networks, it aims to solve the problem of model universality caused by the different data distributions of different lines and cable models. It allows knowledge learned on one line to be safely and effectively transferred to another new line with scarce data. The group health status prediction unit performs clustering and trend analysis on the overall cable status. The cross-line knowledge transfer unit uses adversarial neural networks to extract common features from different cable aging modes and remove line-specific features, thereby generating a more generalized evaluation model that can quickly adapt to new lines, greatly reducing the model deployment cost and data accumulation time for new lines and cable models.

[0017] The result output and early warning module includes: an augmented reality interactive interface that overlays the location of electric arc risks onto the actual cable laying scenario using mixed reality devices; and an adaptive maintenance strategy generator that automatically matches spare parts inventory, personnel scheduling, and power outage time windows based on the assessment result report. When maintenance personnel wear mixed reality devices for inspection, the devices automatically identify cables and overlay risk information such as the most likely deterioration locations assessed by the system onto real cable joints and terminals in the form of highlights and icons. At the same time, the adaptive maintenance strategy generator automatically generates a preliminary maintenance plan including a spare parts list, work order, and recommended power outage windows based on the assessment results, spare parts inventory, shift schedule, and power grid operation mode. Transforming the assessment results from paper reports into intuitive, three-dimensional visualizations greatly facilitates on-site location and verification.

[0018] The beneficial effects of this invention are as follows: It moves from passive protection to active prediction. By intelligently analyzing arc characteristics and converting them into quantitative indicators of cable insulation status, it effectively distinguishes arc types and severity. Deep mining of arc signals reveals valuable information for assessing the overall aging state and insulation degradation trends of cables, enabling dynamic evaluation of the cable aging process and early fault prediction. Threshold judgments do not rely on the time-domain characteristics of current waveforms, thus avoiding interference from nonlinear loads such as motor starting and thyristor devices, directly preventing false alarms or missed alarms. The integration of multi-dimensional, cross-domain feature parameters effectively overcomes load interference, enabling more accurate differentiation of fault arc types (such as series and parallel arcs) and their severity, significantly reducing the false alarm rate. By establishing cable status archives and performing trend analysis, it provides data support for maintenance decisions, such as maintenance timing and replacement priorities, facilitating condition-based maintenance and lean asset management. It can operate independently on a single cable or be networked to achieve centralized intelligent monitoring of a regional cable network. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. It should be noted that this is only for the purpose of more clearly illustrating and explaining the present invention.

[0021] like Figure 1As shown, this embodiment discloses a cable condition assessment method based on a fault arc model, including the following steps: S1: Real-time monitoring of the electrical signals of the target cable, capturing and recording the original waveform data within a preset time window before and after the occurrence of the fault arc; S2: Preprocessing the original waveform data and extracting a multi-dimensional feature parameter set to characterize the arc characteristics and cable condition; S3: Inputting the multi-dimensional feature parameter set into a pre-trained cable condition assessment model; S4: Generating a current condition assessment result and maintenance suggestions for the target cable based on the output of the cable condition assessment model.

[0022] The working principle is as follows: The fault arc signal is transformed from a traditional fault alarm signal into an information carrier that can be used for in-depth analysis of the cable insulation health status. By capturing waveform data of the entire arc event process and extracting multi-dimensional features that comprehensively reflect the physical characteristics of the arc and the cable insulation status, a comprehensive evaluation is performed using an intelligent model, thus achieving a shift from passive protection to proactive early warning. In implementation, the first step (S1) involves continuously monitoring the cable current / voltage using sensors. Once a possible arc disturbance is detected, the complete waveform before and after the event is recorded with high precision. The second step (S2) involves cleaning and aligning the recorded raw waveforms, and extracting a set of quantitative feature values ​​using various signal analysis algorithms (such as time-frequency analysis). The third step (S3) involves inputting these feature values ​​into a pre-trained intelligent evaluation model (such as a machine learning model). The fourth step (S4) involves the model outputting quantitative evaluation results (such as health score and risk level) and specific maintenance recommendations. This achieves intelligent and quantitative evaluation of cable status, breaking through the limitations of traditional technologies that can only determine the presence or absence of an arc, and providing direct data decision support for predictive maintenance.

[0023] In step S1, the electrical signals include current signals and voltage signals. Both current and voltage sensors need to be installed during deployment. During data processing, a digital filter is first used to remove high-frequency noise and power frequency interference. Then, the signal is segmented according to the power frequency cycle, and an algorithm is used to accurately locate the start time of each arc pulse, providing clean and time-aligned standard data segments for subsequent feature analysis.

[0024] In step S2, the preprocessing includes filtering and denoising, power frequency cycle segmentation, and precise calibration of the arc pulse start point. The multi-dimensional feature parameter set includes time-domain features, frequency-domain features, and time-frequency-domain features. The time-domain features include the rise rate, fluctuation entropy, zero-crossing distortion rate, and correlation coefficient with the standard arc waveform library of the current waveform. The frequency-domain features include the energy ratio of specific harmonic components and the ratio of high-frequency noise energy to fundamental wave energy. The time-frequency-domain features include the energy distribution of each component and singular spectral entropy obtained through wavelet transform or empirical mode decomposition. Since single-type features (such as only considering current amplitude) are prone to misjudgment and lack sufficient information, a more comprehensive and stable feature fingerprint is constructed by fusing features from the three dimensions of time domain (waveform shape), frequency domain (harmonic components), and time-frequency-domain (time-varying characteristics), thereby more accurately characterizing the complex physical processes of the arc and the insulation degradation mode of the cable. In practice, calculation tools are used to calculate from the waveform: the rate of change of current in the time domain (rise rate), the degree of waveform disorder (wave entropy), the zero-crossing anomaly (zero-crossing distortion rate), and the similarity to typical fault waveforms (correlation coefficient); the proportion of harmonic content and the noise energy of specific high-frequency bands in the frequency domain; and the distribution characteristics of signal energy at different times and frequencies are analyzed in the time and frequency domains using tools such as wavelet transform.

[0025] In step S3, the method for constructing and training the cable condition assessment model includes: S31: Obtaining a historical cable sample set, which contains the multi-dimensional feature parameter set corresponding to multiple cables of different aging states and specifications under various load conditions when fault arcs are induced, as well as the real state labels of each cable sample determined through offline detection; S32: Using the multi-dimensional feature parameter set as input features and the corresponding real state labels as supervision targets, training and optimizing a selected machine learning model or deep learning model to obtain a preliminary assessment model; S33: Using cable operating environment parameters and historical arc event sequence features as auxiliary inputs, fusing them with the multi-dimensional feature parameter set to enhance the preliminary assessment model, obtaining the cable condition assessment model capable of performing time-series correlation analysis. The working principle is as follows: The model is trained using massive amounts of historical fault data covering various operating conditions, enabling it to learn the complex mapping relationship from features to the actual cable state. Then, environmental parameters and historical event sequences are introduced, allowing the model to understand that the cable state is a process of dynamic evolution over time in a specific operating environment, thereby achieving a more realistic trend judgment. In implementation, firstly, a large number of samples are collected in the laboratory or on-site to establish a correspondence between feature sets and real-state labels for each cable sample. Then, this data is used to train a basic machine learning model (such as a random forest or neural network). Finally, information such as the cable's operating load, temperature and humidity, and historical arc events are integrated into the basic model to enhance its training, enabling it to perform time-series analysis by combining historical data with current conditions. Through supervised learning, the model possesses reliable evaluation capabilities; by integrating environmental and historical information, the model exhibits adaptability and predictive ability, enabling it to assess the current state's position within historical trends, thereby more accurately predicting future degradation rates and making the evaluation results more closely aligned with actual operating conditions.

[0026] In step S4, the status assessment results include at least one of the following: cable insulation health level score, estimated remaining reliable operating time, and location of the most likely deterioration point. Following step S4, the system further includes the following steps: S5: Establishing a full lifecycle status file for the target cable, storing the feature parameter set and status assessment results obtained from each assessment in a time series within the status file; S6: Based on the data in the status file, analyzing the cable status degradation trend, and generating graded early warning information when the status degradation rate exceeds a preset threshold or the assessment result is lower than a safety threshold. In implementation, the system establishes an independent electronic file for each cable, storing the features and results of each assessment in chronological order. The system continuously analyzes this file, for example, calculating the decline slope of the insulation health score. Once accelerated degradation (slope exceeding the threshold) or a score below the safety line is detected, different levels of early warning (such as attention, abnormal, severe) are automatically triggered. After outputting the assessment results, this data is further utilized for full lifecycle management and trend early warning. The principle is to place the results of a single assessment into a time series, and by analyzing the rate of change of status parameters over time, to achieve quantitative tracking and early warning of cable aging trends. Maintenance personnel can identify which cables are aging rapidly in advance, thus enabling them to scientifically plan maintenance schedules and allocate resources, greatly improving the initiative and economy of power grid operation and maintenance.

[0027] A cable condition assessment system based on a fault arc model, used to implement the aforementioned method, includes: a data acquisition module for real-time monitoring and high-precision sampling of current and voltage signals of the target cable; an arc detection and data caching module for real-time detection of fault arc events from the sampled signals and capturing raw waveform data for a set time period before and after the event; a feature extraction module for preprocessing the cached raw waveform data and calculating and generating the multi-dimensional feature parameter set; a condition assessment engine module, embedded with the pre-trained cable condition assessment model, for receiving the feature parameter set, performing calculations and analysis, and outputting condition assessment results; and a result output and early warning module for visually displaying the assessment results, generating maintenance reports, and issuing early warnings in case of anomalies. In implementation, the data acquisition module (sensor) is responsible for acquiring signals; the arc detection and data caching module (edge ​​computing unit) is responsible for real-time judgment and capturing valid data segments; the feature extraction module (computing server) is responsible for running complex feature calculation algorithms; the condition assessment engine module (core server) loads the model and performs inference; and the result output and early warning module (human-machine interface) is responsible for display and alarms.

[0028] It also includes: a model update and optimization module, used to receive feedback on the actual cable status after field verification, and to perform incremental learning or periodic retraining on the cable status assessment model; and a cloud-based collaborative platform, used to receive assessment data from multiple systems, perform big data aggregation and analysis to optimize and distribute a globally applicable assessment model, and provide data support for regional performance cable network status analysis. During implementation, the model update and optimization module will use the predicted and verified data pairs as new samples after each field inspection confirms the actual cable status, and fine-tune the model (incremental learning). The cloud-based collaborative platform gathers data from hundreds or thousands of field systems, trains a more powerful global model, and then distributes it to each field system. It also provides a comprehensive cable health report for the regional power grid, solving the problems of limited data from a single system and insufficient model generalization ability.

[0029] The data acquisition module includes: a distributed wideband Rogowski coil with a frequency response range of 0.1Hz-10MHz; a high-precision differential voltage probe, an embedded sensor integrated at the cable connector; and a synchronization clock unit to ensure time alignment accuracy of data acquired from multiple endpoints. Fault arc signals contain rich high-frequency transient components that ordinary power frequency sensors cannot capture. Therefore, wideband, high-precision dedicated sensors must be used, and strict time synchronization of multi-point measurements must be ensured to guarantee the accuracy of waveform analysis. In implementation, Rogowski coils and embedded voltage probes are installed at key monitoring points on the cable, their high-frequency response ensuring the capture of MHz-level arc high-frequency signals. The synchronization clock unit, such as one based on GPS or a precision clock source, ensures that all sensor-acquired data has a unified timestamp, facilitating advanced analysis such as traveling wave localization.

[0030] The cloud-based collaborative platform includes a group health status prediction unit and a cross-line knowledge transfer unit. The cross-line knowledge transfer unit is used to perform adversarial neural network transfer learning on the degradation mode characteristics of different cable lines. Through transfer learning, especially the application of adversarial neural networks, it aims to solve the problem of model universality caused by the different data distributions of different lines and cable models. It allows knowledge learned on one line to be safely and effectively transferred to another new line with scarce data. The group health status prediction unit performs clustering and trend analysis on the overall cable status. The cross-line knowledge transfer unit uses adversarial neural networks to extract common features from different cable aging modes and remove line-specific features, thereby generating a more generalized evaluation model that can quickly adapt to new lines, greatly reducing the model deployment cost and data accumulation time for new lines and cable models.

[0031] The result output and early warning module includes: an augmented reality interactive interface that overlays the location of electric arc risks onto the actual cable laying scenario using mixed reality devices; and an adaptive maintenance strategy generator that automatically matches spare parts inventory, personnel scheduling, and power outage time windows based on the assessment result report. When maintenance personnel wear mixed reality devices for inspection, the devices automatically identify cables and overlay risk information such as the most likely deterioration locations assessed by the system onto real cable joints and terminals in the form of highlights and icons. At the same time, the adaptive maintenance strategy generator automatically generates a preliminary maintenance plan including a spare parts list, work order, and recommended power outage windows based on the assessment results, spare parts inventory, shift schedule, and power grid operation mode. Transforming the assessment results from paper reports into intuitive, three-dimensional visualizations greatly facilitates on-site location and verification.

[0032] During implementation, a 10kV XLPE insulated power cable was used as the evaluation object.

[0033] Data acquisition: A Rogowski coil current sensor with a bandwidth of 0.1Hz-10MHz and a capacitive voltage divider are installed at the beginning of the cable, with the sampling rate set to 10MHz.

[0034] Arc event capture: The edge computing unit monitors the current signal in real time. When it detects that the current waveform has high-frequency oscillation and meets the initial arc characteristics, it triggers recording and saves complete waveform data for 5 power frequency cycles before and after the event.

[0035] Feature extraction: Time domain: Calculate the correlation coefficient between the current waveform and the standard carbonization path arc waveform library (a drop below 0.85 is considered abnormal), and calculate the current rise rate di / dt.

[0036] Frequency domain: Perform FFT transformation to calculate the ratio of the total content of the 3rd, 5th, and 7th harmonics to the fundamental frequency content, and calculate the energy proportion of the 150kHz-500kHz frequency band.

[0037] Time-frequency domain: Perform 5-level decomposition of db4 wavelet packets and calculate the energy entropy of each node.

[0038] This results in a set of parameter vectors containing 12 core features.

[0039] Model Evaluation: The feature vectors were input into the trained XGBoost ensemble learning model. This model was trained in the laboratory using cable arc test data under different aging levels and loads. The model output a "Insulation Health Index" of 72 (out of 100) for this arc event, assessing it as a "Attention State," and predicting a 15% probability that the insulation resistance value will drop to the threshold within the next 3 months.

[0040] Warning and Feedback: The system marked the cable in yellow on the monitoring screen and generated an inspection work order, recommending that maintenance personnel focus on the cable termination connection during the next monthly inspection. All data from this inspection was archived. One month later, infrared thermography revealed an abnormal temperature rise at one cable termination. The processed results were used as feedback data input into the model for fine-tuning of model parameters.

[0041] This invention transforms arc characteristics into quantitative indicators of cable insulation status through intelligent analysis, enabling dynamic assessment of the cable aging process and early prediction of faults.

[0042] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A cable condition assessment method based on a fault arc model, characterized in that, Includes the following steps: S1: Real-time monitoring of the electrical signals of the target cable, capturing and recording the original waveform data within a preset time window before and after the occurrence of the fault arc; S2: Preprocess the original waveform data and extract a set of multi-dimensional feature parameters to characterize the arc characteristics and cable condition; S3: Input the multi-dimensional feature parameter set into the pre-trained cable condition assessment model; S4: Based on the output of the cable condition assessment model, generate the current condition assessment results and maintenance recommendations for the target cable.

2. The cable condition assessment method based on the fault arc model according to claim 1, characterized in that, In step S2, the multi-dimensional feature parameter set includes time-domain features, frequency-domain features, and time-frequency-domain features; wherein, the time-domain features include the rise rate of the current waveform, fluctuation entropy, zero-crossing distortion rate, and correlation coefficient with the standard arc waveform library; the frequency-domain features include the energy ratio of specific harmonic components and the ratio of high-frequency noise energy to fundamental wave energy; the time-frequency-domain features include the energy distribution of each component and singular spectral entropy obtained by wavelet transform or empirical mode decomposition.

3. The cable condition assessment method based on the fault arc model according to claim 2, characterized in that, In step S3, the method for constructing and training the cable condition assessment model includes: S31: Obtain a historical cable sample set, which includes the multi-dimensional feature parameter set corresponding to multiple cables of different aging states and specifications when fault arcs are induced under various load conditions, as well as the real state label of each cable sample determined by offline detection. S32: Using the multi-dimensional feature parameter set as input features and the corresponding real state label as supervision target, train and optimize the selected machine learning model or deep learning model to obtain a preliminary evaluation model. S33: Using cable operating environment parameters and historical arc event sequence features as auxiliary inputs, and fusing them with the multi-dimensional feature parameter set, the preliminary evaluation model is reinforced and trained to obtain the cable condition evaluation model capable of performing time-series correlation analysis.

4. The cable condition assessment method based on the fault arc model according to claim 3, characterized in that, In step S4, the condition assessment results include at least one of the following: cable insulation health level score, estimated remaining reliable operating time, and location of the most likely deterioration point; and after step S4, the following steps are also included: S5: Establish a full life cycle status file for the target cable, and store the characteristic parameter set and status evaluation results obtained from each evaluation into the status file in a time series. S6: Based on the data in the status file, analyze the cable status degradation trend, and generate graded early warning information when the status degradation rate exceeds a preset threshold or the evaluation result is lower than a safety threshold.

5. The cable condition assessment method based on a fault arc model according to any one of claims 1 to 4, characterized in that, In step S1, the electrical signal includes a current signal and a voltage signal; in step S2, the preprocessing includes filtering and noise reduction, power frequency cycle segmentation, and precise calibration of the arc pulse start point.

6. A cable condition assessment system based on a fault arc model, used to implement the method as described in claim 5, characterized in that, include: The data acquisition module is used to monitor and sample the current and voltage signals of the target cable in real time with high precision; The arc detection and data caching module is used to detect fault arc events in real time from the sampled signals and capture the raw waveform data before and after the event for a set time period. The feature extraction module is used to preprocess the cached raw waveform data and calculate and generate the multi-dimensional feature parameter set. The state assessment engine module, which is embedded with the pre-trained cable state assessment model, is used to receive the feature parameter set, perform calculation and analysis, and output the state assessment result. The results output and early warning module is used to visualize the evaluation results, generate maintenance reports, and issue early warnings when anomalies occur.

7. The cable condition assessment system based on the fault arc model according to claim 6, characterized in that, Also includes: The model update and optimization module is used to receive feedback on the actual cable status after field verification and to perform incremental learning or periodic retraining on the cable status assessment model. The cloud-based collaborative platform is used to receive evaluation data from multiple systems, perform big data aggregation and analysis to optimize and distribute globally applicable evaluation models, and provide data support for regional performance cable network status analysis.

8. The cable condition assessment system based on the fault arc model according to claim 6, characterized in that, The data acquisition module includes: a distributed wideband Rogowski coil with a frequency response range of 0.1Hz-10MHz; a high-precision differential voltage probe, an embedded sensor integrated into the cable connector; and a synchronous clock unit to ensure the time alignment accuracy of data acquired from multiple endpoints.

9. The cable condition assessment system based on the fault arc model according to claim 7, characterized in that, The cloud-based collaborative platform includes a group health status prediction unit and a cross-line knowledge transfer unit. The cross-line knowledge transfer unit is used to perform adversarial neural network transfer learning on the degradation mode characteristics of different cable lines.

10. The cable condition assessment system based on the fault arc model according to claim 6, characterized in that, The result output and early warning module includes: an augmented reality interactive interface that overlays the location of electric arc risk onto the actual cable laying scenario using mixed reality devices; and an adaptive maintenance strategy generator that automatically matches spare parts inventory, personnel scheduling, and power outage time windows based on the evaluation result report.

Images