Multi-region power cable aging state joint evaluation system fusing federated learning and narrowband internet of things

Abstract

The application discloses a multi-region power cable aging state joint evaluation system fusing federal learning and narrowband Internet of Things, and relates to the technical field of power system state monitoring. The system comprises a data acquisition layer, an edge computing layer, a federal learning layer and an evaluation decision layer. Multidimensional cable operation data are collected through narrowband Internet of Things, data preprocessing and local feature extraction are completed by an edge node, the federal learning layer realizes multi-region model collaborative training based on an improved federal average algorithm, and the evaluation decision layer completes aging state grade evaluation through a multi-scale feature fusion model. On the premise of protecting data privacy, the system breaks the multi-region data island, realizes the collaborative linkage of distributed training and centralized decision-making, significantly improves the comprehensiveness, real-time performance and reliability of power cable aging state evaluation, provides a scientific basis for power system operation and maintenance, and has wide engineering application value.

Description

Technical Field

[0001] This invention relates to the field of power system condition monitoring and assessment technology, specifically a multi-regional power cable aging condition joint assessment system that integrates federated learning and narrowband Internet of Things. Background Technology

[0002] As the core carrier of electrical energy transmission in a power system, the operating status of power cables directly affects the safety, stability, and reliability of the power system. As power cables age, they are subjected to long-term effects from various factors such as electricity, heat, machinery, and the environment. The cable insulation, conductors, and sheaths will gradually age and deteriorate. If this is not detected in time and targeted maintenance measures are not taken, it may lead to serious faults such as insulation breakdown and short circuits, causing large-scale power outages and resulting in huge economic losses and social impacts. Therefore, achieving accurate and efficient assessment of the aging status of power cables is a key technical requirement in the operation and maintenance management of power systems.

[0003] Currently, power cable aging status assessment technologies are mainly divided into two categories: traditional offline testing and online monitoring. Traditional offline testing requires power outage sampling of cables and analysis of cable aging through laboratory testing. This method suffers from drawbacks such as long testing cycles, high costs, and inability to achieve real-time monitoring. Furthermore, power outage testing disrupts normal power supply, making it difficult to meet the uninterrupted operation requirements of modern power systems. Online monitoring technology, by deploying sensors on cable lines to collect operational data, achieves real-time monitoring without power outages and has become the mainstream development direction. However, existing online monitoring technologies still have many shortcomings: First, the data collection dimensions are limited, mostly monitoring only single physical quantities such as partial discharge and temperature, failing to comprehensively cover the key influencing factors of cable aging, leading to one-sided assessment results. Second, the data transmission and processing modes are limited. Traditional online monitoring systems mostly adopt a centralized architecture, transmitting data collected from various areas to a central server for unified processing, which not only faces the challenge of massive data transmission... The transmission of power cables presents several challenges. First, the bandwidth pressure and data privacy risks are significant, especially for cross-regional and cross-departmental power cable networks, where the conflict between data sharing and privacy protection is more pronounced. Second, multi-regional collaborative assessment capabilities are insufficient. Due to differences in the operating environment, load characteristics, and maintenance levels of power cables in different regions, data silos are formed. Existing technologies struggle to effectively integrate and collaboratively assess data from multiple regions, resulting in poor generalization ability of assessment models and difficulty in adapting to complex and ever-changing real-world operating scenarios. Third, the integration of monitoring technology and assessment algorithms is not deep enough. Narrowband Internet of Things (NB-IoT), as a low-power, wide-coverage, and high-connectivity communication technology, has been initially applied in the field of power monitoring. However, current applications are mostly limited to data transmission and have not been deeply integrated with advanced machine learning algorithms, failing to fully explore the potential value of the data. At the same time, existing assessment algorithms are mostly trained based on single-region data, lacking adaptability to heterogeneous data from multiple regions, and the accuracy and reliability of assessments need to be improved.

[0004] Furthermore, the application of existing federated learning technology in power systems is still in its early stages, mainly focusing on scenarios such as load forecasting and fault diagnosis. There are few applications for joint assessment of the aging status of power cables, and existing technologies do not fully consider the heterogeneity and temporal characteristics of power cable data, as well as the issues of coordination and stability in multi-region model training. Therefore, how to integrate the efficient data acquisition capabilities of NB-IoT with the privacy protection features of federated learning to build a system that can break down data silos, achieve multi-regional collaboration, and accurately assess the aging status of power cables has become a pressing technical challenge in the field of power system condition monitoring. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-regional power cable aging status joint assessment system that integrates federated learning and narrowband Internet of Things, thus solving the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: a multi-regional power cable aging status joint assessment system integrating federated learning and narrowband Internet of Things, comprising a data acquisition layer, an edge computing layer, a federated learning layer and an assessment decision layer connected in sequence;

[0007] The data acquisition layer establishes communication with power cable monitoring points through narrowband IoT to collect multi-dimensional heterogeneous data during cable operation; the edge computing layer preprocesses the collected data and extracts local features to generate standardized feature vectors; the federated learning layer generates a global evaluation model based on the local model parameters of multi-region edge nodes through a collaborative training algorithm; the evaluation decision layer calls the global evaluation model, combines cable operating environment parameters and historical data, and outputs cable aging status evaluation results and maintenance suggestions.

[0008] Each level uses an encrypted communication protocol to ensure secure transmission of data and model parameters, forming a closed-loop evaluation architecture.

[0009] Optionally, the data acquisition layer includes several NB-IoT acquisition terminals, each of which is configured with multiple types of sensor modules, including partial discharge sensors, temperature sensors, humidity sensors, insulation resistance sensors, and cable operating load sensors.

[0010] The data acquisition terminal is based on key influencing factors of cable aging, presets the acquisition frequency and data accuracy thresholds, optimizes communication energy consumption through NB-IoT's eDRX power-saving mode and PSM sleep mechanism, and after the acquired data is initially filtered locally on the terminal, it is transmitted to the edge computing node in the corresponding area through the NB-IoT base station in a preset data format.

[0011] Optionally, the edge computing layer includes several regional edge nodes, and each edge node is configured with a data preprocessing unit and a local feature extraction unit;

[0012] The data preprocessing unit adopts a 3D-based approach. The outlier removal algorithm and the missing value imputation algorithm of the linear interpolation method are used to standardize the collected data and generate a normalized data matrix. The local feature extraction unit extracts the time-domain and frequency-domain features of the data through the local feature extraction module of the convolutional neural network, and specifically achieves feature mapping through the following formula:

[0013]

[0014] in, To standardize the data matrix, The convolution kernel weight matrix is... The kernel size is [size]. Step size, For bias terms, This represents the convolution operation. It is the ReLU activation function. These are local feature vectors.

[0015] Optionally, the federated learning layer includes a federated coordination node and several regional local training nodes, and the federated coordination node and each regional local training node establish a connection through an encrypted communication link.

[0016] The local training nodes in the region are trained using an improved federated averaging algorithm based on the local feature vectors output by the edge computing layer. The local model parameter update formula is as follows:

[0017]

[0018] in, For the first The training nodes in the region are at the... Wheel model parameters, For learning rate, The gradient of the loss function. For the first The local feature datasets of each region; the federated coordination node receives model parameters uploaded by the local training nodes of each region, and adaptively adjusts the aggregation weights based on the regional data quality weights and model training effects. The global model parameter update formula is:

[0019]

[0020] in, For the first Global model parameters of the wheel, For the number of regions, For the first The aggregate weight of each region satisfies ,and It is calculated jointly by regional data integrity, data sample size and local model loss value.

[0021] Optionally, the evaluation decision layer includes a feature fusion unit, an aging state evaluation unit, and a decision output unit; the feature fusion unit employs a multi-scale feature fusion algorithm combining attention mechanism and weighted fusion to fuse the global features output by the federated learning layer with the local features of each region, and the fusion formula is:

[0022]

[0023] in, To fuse feature vectors, For global feature vectors, For the first Local feature vectors of each region For global feature weights, For the first Local feature weights for each region, and , and Dynamically adjusted through attention mechanisms.

[0024] Optionally, the aging status assessment unit constructs a cable aging status mapping model based on fused feature vectors, and outputs the aging status level using a fuzzy comprehensive evaluation method combined with a BP neural network. This is specifically achieved through the following steps:

[0025] (1) Map the fused feature vectors to an aging influence factor matrix;

[0026] (2) Determine the weight coefficients of each influencing factor using the analytic hierarchy process;

[0027] (3) Construct a fuzzy evaluation matrix based on the fuzzy membership function;

[0028] (4) The fuzzy evaluation results are nonlinearly fitted by the BP neural network to output the aging status level, including four levels: mild aging, moderate aging, severe aging and critical aging. The activation function of the output layer of the BP neural network is the Softmax function, and the output vector is the probability distribution of each aging level.

[0029] Optionally, a data security transmission unit is set up between the edge computing layer and the federated learning layer. The local feature vector and model parameters are encrypted and transmitted using a symmetric encryption method based on the national cryptographic algorithm SM4. At the same time, data tampering and identity forgery are prevented through timestamp verification and digital signature mechanisms.

[0030] The federated coordination node is equipped with a model parameter verification unit to perform consistency verification on the local model parameters uploaded from each region, remove abnormal parameters, and then perform aggregation calculations to ensure the stability of the global model.

[0031] Optionally, the evaluation decision layer also includes a model adaptive update unit, which monitors the distribution changes of cable operation data in each region in real time. When the data distribution difference exceeds a preset threshold, it triggers the federated learning layer to retrain collaboratively and update the global evaluation model parameters. At the same time, the system is configured with a compatible interface module that can interface with the existing power system operation and maintenance management platform to realize the real-time push of evaluation results and the two-way interaction of operation and maintenance instructions.

[0032] Optionally, the NB-IoT acquisition terminal is configured with an adaptive sampling unit to dynamically adjust the acquisition frequency based on changes in cable operating load: when the load is within the rated range, a low-frequency acquisition mode is used; when the load exceeds 80% of the rated range or a sudden change occurs, it automatically switches to a high-frequency acquisition mode. The acquisition frequency adjustment formula is as follows:

[0033]

[0034] in, This is the actual sampling frequency. The lowest sampling frequency, This is the highest sampling frequency. To run the load in real time, For rated load, This represents the maximum allowable load.

[0035] Optionally, the system also includes a remote monitoring and maintenance unit, which provides real-time monitoring interfaces to operation and maintenance personnel via web and mobile terminals, displaying the cable aging status assessment results, model training progress, and equipment operating status in each area;

[0036] It also supports remote parameter configuration and system upgrades, and reduces the amount of data transmitted during upgrades through differential upgrade technology, thereby reducing the impact on the normal operation of the power system.

[0037] This multi-regional joint assessment system for the aging status of power cables, which integrates federated learning and narrowband Internet of Things, solves the problems of partial data collection, insufficient privacy protection, weak collaborative assessment capabilities, and limited assessment accuracy in existing technologies through innovative architecture design, algorithm optimization, and module linkage. It has the following significant benefits:

[0038] Firstly, the system deploys multiple types of sensor modules based on NB-IoT technology to collect multi-dimensional data closely related to cable aging, such as partial discharge, temperature, humidity, insulation resistance, and operating load, covering various influencing factors including electricity, heat, environment, and machinery. At the same time, the acquisition terminal is equipped with an adaptive sampling unit that dynamically adjusts the acquisition frequency according to the cable's operating load, reducing system energy consumption and data transmission pressure while ensuring data integrity, thus achieving accurate and efficient data acquisition.

[0039] Secondly, the system adopts a distributed training mode, where each region only uploads model parameters instead of raw data, avoiding the risk of privacy leakage during cross-regional data transmission and meeting the strict requirements of power system data security management. At the same time, the improved federated averaging algorithm introduces an adaptive aggregation weight mechanism, which dynamically adjusts the weight allocation based on regional data quality and model training effect, thereby improving the generalization ability and stability of the global model.

[0040] Furthermore, the feature fusion unit constructed by the system combines attention mechanism and weighted fusion method to fully explore the complementary value of global features and local features of each region, thereby enhancing the representational ability of feature vectors. The aging status assessment unit adopts an algorithm that combines fuzzy comprehensive evaluation method and BP neural network. It can not only reasonably allocate the weight of each influencing factor through hierarchical analysis method, but also fit complex nonlinear mapping relationship through neural network. It can effectively deal with the uncertainty brought about by heterogeneous data in multiple regions, significantly improve the accuracy and robustness of aging status level assessment, and provide a scientific and reliable basis for operation and maintenance decision-making.

[0041] Meanwhile, the system is equipped with a compatible interface module, which can seamlessly connect with the existing power system operation and maintenance management platform without the need for large-scale modification of existing equipment, thus reducing the system deployment cost and implementation difficulty. In addition, the model adaptive update unit can respond to changes in data distribution in real time and dynamically adjust the evaluation model parameters to ensure that the system maintains good evaluation performance throughout the entire life cycle of the cable.

[0042] Finally, by accurately assessing the aging status of power cables, we can guide maintenance personnel to develop targeted maintenance strategies, effectively reducing maintenance costs and power outage losses. At the same time, by providing early warnings of potential fault risks, we can avoid serious power accidents, ensure the safe and stable operation of the power system, improve power supply reliability, and provide strong support for social and economic development. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the system flow of the invention. Detailed Implementation

[0044] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0045] Please see Figure 1 The present invention provides a technical solution: a multi-regional power cable aging status joint assessment system integrating federated learning and narrowband Internet of Things, comprising a data acquisition layer, an edge computing layer, a federated learning layer and an assessment decision layer connected in sequence;

[0046] The data acquisition layer establishes communication with power cable monitoring points through narrowband IoT to collect multi-dimensional heterogeneous data during cable operation. The data acquisition layer includes several NB-IoT acquisition terminals, each of which is equipped with multiple types of sensor modules, including partial discharge sensors, temperature sensors, humidity sensors, insulation resistance sensors, and cable operation load sensors.

[0047] Based on the key influencing factors of cable aging, the data acquisition terminal presets the acquisition frequency and data accuracy thresholds. It optimizes communication energy consumption through NB-IoT's eDRX power-saving mode and PSM sleep mechanism. After the acquired data is initially filtered locally on the terminal, it is transmitted to the edge computing node in the corresponding area through the NB-IoT base station in a preset data format.

[0048] The NB-IoT data acquisition terminal is equipped with an adaptive sampling unit that dynamically adjusts the acquisition frequency based on changes in cable operating load: when the load is within the rated range, a low-frequency acquisition mode is used; when the load exceeds 80% of the rated range or a sudden change occurs, it automatically switches to a high-frequency acquisition mode. The acquisition frequency adjustment formula is as follows:

[0049]

[0050] in, This is the actual sampling frequency. The lowest sampling frequency, This is the highest sampling frequency. To run the load in real time, For rated load, Maximum allowable load;

[0051] The edge computing layer preprocesses the collected data and extracts local features to generate standardized feature vectors. The edge computing layer includes several regional edge nodes, and each edge node is configured with a data preprocessing unit and a local feature extraction unit.

[0052] The data preprocessing unit adopts a 3D-based approach. The outlier removal algorithm based on the criteria and the missing value imputation algorithm based on linear interpolation are used to standardize the collected data and generate a normalized data matrix. The local feature extraction unit extracts the time-domain and frequency-domain features from the data through the local feature extraction module of the convolutional neural network, specifically through the feature mapping achieved by the following formula:

[0053]

[0054] in, To standardize the data matrix, The convolution kernel weight matrix is... The kernel size is [size]. Step size, For bias terms, This represents the convolution operation. It is the ReLU activation function. These are local feature vectors;

[0055] The federated learning layer generates a global evaluation model based on the local model parameters of multiple regional edge nodes through a collaborative training algorithm. The federated learning layer includes a federated coordination node and several regional local training nodes. The federated coordination node and each regional local training node are connected through an encrypted communication link.

[0056] The local training nodes in the region are based on the local feature vectors output by the edge computing layer. An improved federated averaging (FedAvg) algorithm is used for local model training. The local model parameter update formula is as follows:

[0057]

[0058] in, For the first The training nodes in the region are at the... Wheel model parameters, For learning rate, The gradient of the loss function. For the first Local feature datasets for each region; the federated coordination node receives model parameters uploaded by local training nodes in each region, adaptively adjusts the aggregation weights based on regional data quality weights and model training performance, and the global model parameter update formula is:

[0059]

[0060] in, For the first Global model parameters of the wheel, For the number of regions, For the first The aggregate weight of each region satisfies ,and It is calculated jointly by regional data integrity, data sample size, and local model loss value;

[0061] A secure data transmission unit is set up between the edge computing layer and the federated learning layer. The local feature vectors and model parameters are encrypted and transmitted using a symmetric encryption method based on the national cryptographic algorithm SM4. At the same time, data tampering and identity forgery are prevented through timestamp verification and digital signature mechanisms.

[0062] The federal coordination node sets up a model parameter verification unit to perform consistency verification on the local model parameters uploaded by each region, remove abnormal parameters, and then perform aggregation calculations to ensure the stability of the global model.

[0063] The evaluation decision layer invokes the global evaluation model, combines cable operating environment parameters and historical data, and outputs cable aging status evaluation results and maintenance recommendations. The evaluation decision layer includes a feature fusion unit, an aging status evaluation unit, and a decision output unit. The feature fusion unit employs a multi-scale feature fusion algorithm combining attention mechanisms and weighted fusion to fuse the global features output by the federated learning layer with the local features of each region. The fusion formula is as follows:

[0064]

[0065] in, To fuse feature vectors, For global feature vectors, For the first Local feature vectors of each region For global feature weights, For the first Local feature weights for each region, and , and Dynamically adjusted through attention mechanisms;

[0066] The aging status assessment unit constructs a cable aging status mapping model based on fused feature vectors, and outputs the aging status level using a fuzzy comprehensive evaluation method combined with a BP neural network. This is achieved through the following steps:

[0067] (1) Map the fused feature vectors to an aging influence factor matrix;

[0068] (2) Determine the weight coefficients of each influencing factor using the analytic hierarchy process;

[0069] (3) Construct a fuzzy evaluation matrix based on the fuzzy membership function;

[0070] (4) The fuzzy evaluation results are nonlinearly fitted by the BP neural network to output the aging status level, including four levels: mild aging, moderate aging, severe aging and critical aging. The activation function of the output layer of the BP neural network is the Softmax function, and the output vector is the probability distribution of each aging level.

[0071] The evaluation decision layer also includes a model adaptive update unit, which monitors the distribution changes of cable operation data in various regions in real time. When the data distribution difference exceeds a preset threshold, it triggers the federated learning layer to retrain collaboratively and update the global evaluation model parameters. At the same time, the system is configured with a compatible interface module that can be connected to the existing power system operation and maintenance management platform to realize the real-time push of evaluation results and the two-way interaction of operation and maintenance instructions.

[0072] Each level uses an encrypted communication protocol to achieve secure transmission of data and model parameters, forming a closed-loop evaluation architecture;

[0073] The system also includes a remote monitoring and maintenance unit, which provides real-time monitoring interfaces to maintenance personnel via web and mobile terminals, displaying the cable aging status assessment results, model training progress and equipment operating status in each area;

[0074] It also supports remote parameter configuration and system upgrades, and reduces the amount of data transmitted during upgrades through differential upgrade technology, thereby reducing the impact on the normal operation of the power system.

[0075] The above description is only a preferred embodiment 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 inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-regional power cable aging status joint assessment system integrating federated learning and narrowband Internet of Things, characterized in that, It includes a data acquisition layer, an edge computing layer, a federated learning layer, and an evaluation and decision-making layer that are connected in sequence. The data acquisition layer establishes communication with power cable monitoring points through narrowband IoT to collect multi-dimensional heterogeneous data during cable operation; the edge computing layer preprocesses the collected data and extracts local features to generate standardized feature vectors; the federated learning layer generates a global evaluation model based on the local model parameters of multi-region edge nodes through a collaborative training algorithm; the evaluation decision layer calls the global evaluation model, combines cable operating environment parameters and historical data, and outputs cable aging status evaluation results and maintenance suggestions. Each level uses an encrypted communication protocol to ensure secure transmission of data and model parameters, forming a closed-loop evaluation architecture.

2. The system according to claim 1, characterized in that, The data acquisition layer includes several NB-IoT acquisition terminals, each of which is configured with multiple types of sensor modules, including partial discharge sensors, temperature sensors, humidity sensors, insulation resistance sensors, and cable operating load sensors. The data acquisition terminal is based on key influencing factors of cable aging, presets the acquisition frequency and data accuracy thresholds, optimizes communication energy consumption through NB-IoT's eDRX power-saving mode and PSM sleep mechanism, and after the acquired data is initially filtered locally on the terminal, it is transmitted to the edge computing node in the corresponding area through the NB-IoT base station in a preset data format.

3. The system according to claim 1, characterized in that, The edge computing layer includes several regional edge nodes, and each edge node is configured with a data preprocessing unit and a local feature extraction unit. The data preprocessing unit adopts a 3D-based approach. The outlier removal algorithm and the missing value imputation algorithm of the linear interpolation method are used to standardize the collected data and generate a normalized data matrix. The local feature extraction unit extracts the time-domain and frequency-domain features of the data through the local feature extraction module of the convolutional neural network, and specifically achieves feature mapping through the following formula: in, To standardize the data matrix, The convolution kernel weight matrix is... The kernel size is [size]. Step size, For bias terms, This represents the convolution operation. It is the ReLU activation function. These are local feature vectors.

4. The system according to claim 1, characterized in that, The federated learning layer includes a federated coordination node and several regional local training nodes. The federated coordination node and each regional local training node are connected through an encrypted communication link. The local training nodes in the region are trained using the local feature vectors output by the edge computing layer, and an improved FedAvg algorithm is employed for local model training. The local model parameter update formula is as follows: in, For the first The training nodes in the region are at the... Wheel model parameters, For learning rate, The gradient of the loss function. For the first The local feature datasets of each region; the federated coordination node receives model parameters uploaded by the local training nodes of each region, and adaptively adjusts the aggregation weights based on the regional data quality weights and model training effects. The global model parameter update formula is: in, For the first Global model parameters of the wheel, For the number of regions, For the first The aggregate weight of each region satisfies ,and It is calculated jointly by regional data integrity, data sample size and local model loss value.

5. The system according to claim 1, characterized in that, The evaluation decision layer includes a feature fusion unit, an aging state evaluation unit, and a decision output unit. The feature fusion unit employs a multi-scale feature fusion algorithm combining attention mechanism and weighted fusion to fuse the global features output by the federated learning layer with the local features of each region. The fusion formula is as follows: in, To fuse feature vectors, For global feature vectors, For the first Local feature vectors of each region For global feature weights, For the first Local feature weights for each region, and , and Dynamically adjusted through attention mechanisms.

6. The system according to claim 5, characterized in that, The aging status assessment unit constructs a cable aging status mapping model based on fused feature vectors, and outputs the aging status level using a fuzzy comprehensive evaluation method combined with a BP neural network. This is achieved through the following steps: (1) Map the fused feature vectors to an aging influence factor matrix; (2) Determine the weight coefficients of each influencing factor using the analytic hierarchy process; (3) Construct a fuzzy evaluation matrix based on the fuzzy membership function; (4) The fuzzy evaluation results are nonlinearly fitted by the BP neural network to output the aging status level, including four levels: mild aging, moderate aging, severe aging and critical aging. The activation function of the output layer of the BP neural network is the Softmax function, and the output vector is the probability distribution of each aging level.

7. The system according to claim 1, characterized in that, A secure data transmission unit is set up between the edge computing layer and the federated learning layer. The local feature vectors and model parameters are encrypted and transmitted using a symmetric encryption method based on the national cryptographic algorithm SM4. At the same time, data tampering and identity forgery are prevented through timestamp verification and digital signature mechanisms. The federated coordination node is equipped with a model parameter verification unit to perform consistency verification on the local model parameters uploaded from each region, remove abnormal parameters, and then perform aggregation calculations to ensure the stability of the global model.

8. The system according to claim 1, characterized in that, The evaluation decision layer also includes a model adaptive update unit, which monitors the distribution changes of cable operation data in various regions in real time. When the data distribution difference exceeds a preset threshold, it triggers the federated learning layer to retrain collaboratively and update the global evaluation model parameters. At the same time, the system is configured with a compatible interface module, which can interface with the existing power system operation and maintenance management platform to realize the real-time push of evaluation results and the two-way interaction of operation and maintenance instructions.

9. The system according to claim 2, characterized in that, The NB-IoT acquisition terminal is equipped with an adaptive sampling unit that dynamically adjusts the acquisition frequency based on changes in cable operating load: when the load is within the rated range, a low-frequency acquisition mode is used; when the load exceeds 80% of the rated range or a sudden change occurs, it automatically switches to a high-frequency acquisition mode. The acquisition frequency adjustment formula is as follows: in, This is the actual sampling frequency. The lowest sampling frequency, This is the highest sampling frequency. To run the load in real time, For rated load, This represents the maximum allowable load.

10. The system according to claim 1, characterized in that, The system also includes a remote monitoring and maintenance unit, which provides real-time monitoring interfaces to maintenance personnel via web and mobile terminals, displaying the cable aging status assessment results, model training progress and equipment operating status in each area; It also supports remote parameter configuration and system upgrades, and reduces the amount of data transmitted during upgrades through differential upgrade technology, thereby reducing the impact on the normal operation of the power system.

Images