New energy cable intelligent management method and system based on internet of things

Abstract

This invention relates to the field of new energy cable management technology, specifically to an intelligent management method and system for new energy cables based on the Internet of Things (IoT). The system includes: an array of sensors deployed on the new energy cable itself and at monitoring points along the cable line to continuously collect real-time operational status data such as cable sheath temperature, conductor current load, joint vibration amplitude, and ambient air humidity; the data is transmitted to an edge computing node for preprocessing and standardization to form a cable health data package; this data package is then input into an improved fault probability prediction algorithm based on a historical fault database and an environmental stress evolution model to calculate the instantaneous probability values ​​of different levels of faults in each monitoring section at a specific time in the future; adaptive inspections are triggered based on these probability values, generating intelligent work orders that are sent to a mobile inspection terminal; and the algorithm parameters are calibrated using the returned on-site verification results. This method can accurately predict fault risks, rationally allocate inspection resources, continuously optimize algorithm accuracy, and ensure the safe and stable operation of new energy cables.

Description

Technical Field

[0001] This invention relates to the field of new energy cable management technology, and in particular to an intelligent management method and system for new energy cables based on the Internet of Things. Background Technology

[0002] As the core carrier in new energy power generation and transmission systems, the operational stability of new energy cables is directly related to the safe and efficient operation of the entire new energy system. Currently, the industry mainly uses sensor devices to collect cable operation data and combines manual inspection or fixed-cycle inspection mode to achieve cable management. Some technologies will use simple data analysis to judge the cable operation status and assist in the troubleshooting work.

[0003] In existing technologies, the preprocessing of sensor data often involves simple filtering of single-dimensional data without forming standardized health data packages. Fault prediction often relies on single real-time data and lacks integrated consideration of historical fault patterns and changes in environmental stress, making it impossible to accurately predict the probability of faults occurring in different sections and at different levels. Inspection work often adopts a fixed cycle pattern without developing targeted inspection strategies based on differences in fault risk, resulting in problems such as wasted inspection resources and missed inspections of high-risk sections. Furthermore, fault prediction algorithms lack effective on-site verification and parameter calibration mechanisms, and their prediction accuracy gradually decreases after long-term use.

[0004] Current cable management technologies struggle to accurately predict fault risks, suffer from low inspection efficiency and unreasonable resource allocation, and the stability and accuracy of fault prediction algorithms cannot be consistently guaranteed. Therefore, a cable management solution is needed that can accurately predict fault probabilities, intelligently allocate inspection resources, and dynamically optimize algorithms. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent management method and system for new energy cables based on the Internet of Things.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent management method for new energy cables based on the Internet of Things, comprising: An array of sensors deployed on the new energy cable body and monitoring points along the line continuously collects real-time operating status data of the cable. The real-time operating status data includes cable sheath temperature, conductor current load, joint vibration amplitude, and ambient air humidity. The real-time operating status data collected by the sensor array is transmitted to the edge computing node. The edge computing node preprocesses and standardizes the cable sheath temperature, conductor current load, joint vibration amplitude and ambient air humidity to form a cable health data package. The cable health data packet is input into the improved fault probability prediction algorithm, which is based on a historical fault database and an environmental stress evolution model to calculate the instantaneous probability value of different levels of faults occurring in each monitoring section of the cable within a specific future time period. Based on the instantaneous probability value output by the improved fault probability prediction algorithm, an adaptive inspection decision process is triggered. The adaptive inspection decision process is used to generate an intelligent work order containing the coordinates of the inspection target, the inspection priority, and the list of inspection items. The intelligent work order is sent to the mobile inspection terminal, and the on-site verification results returned by the mobile inspection terminal after performing the inspection are received. The parameters of the improved fault probability prediction algorithm are calibrated using the on-site verification results.

[0007] As a further aspect of the present invention, the improved fault probability prediction algorithm is an improved gradient boosting decision tree algorithm, the working principle of which includes: Extract multidimensional feature sequences from the historical fault database. The multidimensional feature sequences include at least the cable sheath temperature time series, conductor current load time series, joint vibration amplitude time series, air humidity time series, historical fault occurrence time, and corresponding fault type codes. The multidimensional feature sequence is divided into time windows, and a label is added to indicate whether a fault occurs in the subsequent unit of time for each time window, thus constructing a sample set for model training. On the sample set, multiple basic decision tree models are initialized, and the multiple basic decision tree models are integrated by weighted voting to form a preliminary fault probability prediction model. The environmental stress evolution model is introduced to simulate the coupled effects of future changes in ambient temperature, solar radiation intensity, and rainfall on the thermodynamic state of the cable, and to calculate the predicted offset of the cable sheath temperature. The predicted offset of the cable sheath temperature is incorporated as a dynamic penalty term into the loss function calculation process of the fault probability prediction model, so that the fault probability prediction model can make penal corrections for future state deviations caused by environmental changes when learning historical patterns. Through iterative training, the loss function value after incorporating the dynamic penalty term converges, resulting in the improved gradient boosting decision tree algorithm. When the improved gradient boosting decision tree algorithm receives a new cable health data packet, it outputs the instantaneous probability value of different levels of faults occurring in each monitoring section of the cable within a specific future time period.

[0008] As a further aspect of the present invention, the real-time operating status data collected by the sensor array is transmitted to an edge computing node. The edge computing node preprocesses and standardizes the cable sheath temperature, conductor current load, joint vibration amplitude, and ambient air humidity to form a cable health data package, including: The edge computing node receives sensor data streams from different physical interfaces and performs timestamp alignment and unit unification processing on the sensor data streams of cable sheath temperature, conductor current load, joint vibration amplitude and air humidity. For the timestamp-aligned cable sheath temperature data, a sliding window filtering technique is used to remove spike noise caused by instantaneous changes in solar radiation; The real-time load rate percentage is calculated by combining the standardized conductor current load data with the cable model nameplate parameters. For the unit-standardized joint vibration amplitude data, frequency domain transformation is performed to extract the amplitude values ​​of characteristic frequencies; Calculate the gradient of change of air humidity data over a unit of time after unitization. The processed cable sheath temperature, real-time load rate percentage of conductor current load, characteristic frequency amplitude value of joint vibration amplitude, and air humidity change gradient are packaged together with the original acquisition time and sensor geographical location information to form the structured cable health data package.

[0009] As a further aspect of the present invention, based on the instantaneous probability value output by the improved fault probability prediction algorithm, an adaptive inspection decision process is triggered, including: Set high-level fault probability thresholds and medium-level fault probability thresholds; The instantaneous probability values ​​of each monitoring section of the cable output by the improved fault probability prediction algorithm are compared with the high-level fault probability threshold and the medium-level fault probability threshold, respectively. For monitoring sections where the instantaneous probability value exceeds the high-level fault probability threshold, their coordinates, current probability value, and corresponding fault level are marked as emergency inspection targets and assigned the highest priority. For monitoring sections whose instantaneous probability values ​​exceed the medium-level fault probability threshold but do not exceed the high-level fault probability threshold, calculate the growth rate of their probability values, assign high priority to monitoring sections with fast growth rate, and assign standard priority to monitoring sections with slow growth rate. Based on the potential fault types predicted by the improved fault probability prediction algorithm, the corresponding detection operations are matched from the preset detection item library to form a unique detection item list for each inspection target. Integrate the coordinates, priorities, and inspection item lists of all inspection targets to generate the aforementioned smart work order for the day or shift.

[0010] As a further aspect of the present invention, the method further includes a work order optimization step based on resource constraints in the adaptive inspection decision process: Obtain the number of currently available mobile inspection terminals, the skill and qualification information of maintenance personnel, and the range and route information of inspection vehicles; The geographical coordinates of all inspection targets in the intelligent work order are input into the path planning model. Combined with the driving range and path information of the inspection vehicle, the theoretical shortest path covering all inspection targets and the required working time are calculated. The required time for the theoretical shortest path is compared with the currently available manpower and time resources. If the required time exceeds a preset threshold, work order optimization is initiated. The work order optimization is based on the priority of the inspection targets, starting with the lowest priority inspection targets, and assessing the impact of removing them from the work orders of the day on the overall risk coverage. Under the premise of ensuring that the overall risk coverage is not lower than the preset safety level, a simplified intelligent work order consisting of the highest priority and high priority inspection targets is generated, and the actual inspection path is replanned for the simplified intelligent work order.

[0011] As a further aspect of the present invention, the improved fault probability prediction algorithm is calibrated using the field verification results, including: From the field verification results returned by the mobile inspection terminal, extract the description of the actual operating status of the cable, fault confirmation information or normal operation confirmation information; The on-site verification results are compared with the prediction results made by the improved fault probability prediction algorithm for the location of the monitoring section in the same time period to generate a prediction deviation record. Collect all prediction deviation records within a specific period, and calculate the average prediction accuracy and false alarm rate of the improved fault probability prediction algorithm for various types of faults; When the average prediction accuracy is lower than the set lower limit or the false alarm rate is higher than the set upper limit, the algorithm parameter calibration process is triggered. The algorithm parameter calibration process uses samples containing recent field verification results to perform small-scale iterative retraining of the weight parameters of the basic decision tree model in the improved fault probability prediction algorithm until the average prediction accuracy and false alarm rate on the new validation set are restored to an acceptable range.

[0012] As a further aspect of the present invention, the method also includes online identification of cable operating modes and early warning of anomalies: In the edge computing node, pattern features are extracted from the continuously input cable health data packets. The pattern features include at least the change period of the load rate percentage, the temperature difference stability between the cable sheath temperature and the ambient temperature, and the fluctuation variance of the amplitude value of the joint characteristic frequency. The extracted mode features are matched with the standard mode features in the cable operation mode feature library, which contains feature templates for various preset modes such as normal operation, light overload, continuous overload, external stress interference, and early stage of insulation aging. When the matched standard modal features are modalities other than those in normal operation, record the modality type, start time, and parameter values ​​of the modality features; If the same anomalous mode continues to occur within multiple consecutive time windows, an early anomaly warning signal is generated. The early anomaly warning signal includes the anomalous mode type, the location of occurrence, the duration, and key characteristic parameters. The early anomaly warning signal is fused and analyzed with the instantaneous probability value output by the improved fault probability prediction algorithm to correct the determination of the priority of inspection targets in the adaptive inspection decision process.

[0013] As a further aspect of the present invention, the step of matching the extracted mode features with the standard mode features in the cable operation mode feature library includes: Calculate the multidimensional Euclidean distance between the extracted mode feature vector and each standard mode feature vector in the cable operation mode feature library; From all the calculated multidimensional Euclidean distances, select the distance with the smallest value and record the operating mode type represented by its corresponding standard mode feature vector; A modal matching confidence threshold is set. If the minimum multidimensional Euclidean distance is less than or equal to the modal matching confidence threshold, then the current operating state of the cable is determined to match the operating mode type. If the minimum multidimensional Euclidean distance is greater than the modal matching confidence threshold, the current operating state of the cable is determined to be undefined mode, and the current mode feature vector is stored in the temporary storage area of ​​the mode to be analyzed. The mode feature vectors in the temporary storage area of ​​the modes to be analyzed are periodically clustered. If a new stable cluster is found, its feature mean is used as a new standard mode feature vector and added to the cable operation mode feature library.

[0014] As a further aspect of the present invention, the steps for constructing the environmental stress evolution model include: Obtain historical meteorological data for the area along the target cable route, including hourly recorded ambient temperature, solar radiation intensity, rainfall, and wind speed. A coupled influence equation for meteorological parameters is established, which is used to describe the coupling relationship between changes in ambient temperature, solar radiation intensity, rainfall, and wind speed on the temperature field of the cable body. Based on the historical meteorological data, the coupling coefficients in the meteorological parameter coupling influence equation are fitted by the multiple regression analysis method to obtain the quantified meteorological parameter coupling influence equation. The thermal parameters of the cable material and the heat dissipation parameters of the cable laying method are obtained. Combined with the quantified meteorological parameter coupling effect equation, an environmental stress evolution model is constructed with ambient temperature, solar radiation intensity, rainfall and wind speed as inputs and cable sheath temperature change as output. The environmental stress evolution model is continuously validated and its parameters are fine-tuned using real-time environmental monitoring data to ensure its accuracy in predicting the trend of cable sheath temperature changes.

[0015] As a further aspect of the present invention, the present invention also includes an intelligent management system for new energy cables based on the Internet of Things (IoT). The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the intelligent management method for new energy cables based on the IoT described above.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: An improved fault probability prediction algorithm, integrating historical fault databases and environmental stress evolution models, is employed to calculate the instantaneous probability of different levels of faults occurring in specific future time periods for each monitoring section of new energy cables. Compared to conventional fault prediction techniques that rely solely on single real-time data and lack consideration of historical patterns and environmental factors, this algorithm comprehensively combines historical cable fault characteristics and environmental stress change trends to accurately distinguish the fault risk levels of different monitoring sections, clearly define the probability of different levels of fault occurrence, avoid over-focusing on low-risk sections and under-focusing on high-risk sections, achieve accurate fault risk prediction, and reduce false positives and false negatives.

[0017] Based on the instantaneous probability values ​​output by the improved fault probability prediction algorithm, an adaptive inspection decision-making process is triggered, generating an intelligent work order containing the coordinates of the inspection target, inspection priority, and a list of inspection items. Simultaneously, the improved fault probability prediction algorithm's parameters are calibrated using on-site verification results transmitted from the mobile inspection terminal. Compared to the conventional fixed-cycle inspection mode, which lacks specificity and suffers from unreasonable resource allocation, adaptive inspection decision-making can rationally allocate inspection resources according to the fault probability, prioritizing high-risk sections, clarifying inspection priorities and items, reducing ineffective inspection operations, and improving inspection efficiency. Algorithm parameter calibration can promptly compensate for initial algorithm deviations and prediction errors caused by environmental changes, cable aging, and other factors during operation. As inspection data accumulates, algorithm accuracy is continuously optimized, forming a closed-loop mechanism of prediction, inspection, and calibration, ensuring the long-term stability and accuracy of fault prediction. Attached Figure Description

[0018] Figure 1 This is a flowchart of the intelligent management method for new energy cables based on the Internet of Things as described in this invention; Figure 2 A flowchart illustrating the preprocessing and standardization of cable health data packets for edge computing nodes; Figure 3 This is a flowchart for optimizing work orders based on resource constraints. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 This invention provides an intelligent management method for new energy cables based on the Internet of Things, the specific method including: An array of sensors, including temperature, current, vibration, and humidity sensors, is deployed on the new energy cable itself and at monitoring points along the route. This array continuously collects data on cable sheath temperature, conductor current load, joint vibration amplitude, and ambient air humidity. The sensor array sends this real-time operational data to the nearest edge computing node. The edge computing node performs timestamp alignment, unit unification, noise filtering, load rate calculation, frequency domain feature extraction, and humidity gradient calculation on this data to form a structured cable health data package. This cable health data package is then input into an improved fault probability prediction algorithm. Based on a historical fault database and an environmental stress evolution model, this algorithm outputs instantaneous probability values ​​for different levels of faults occurring in each monitoring section of the cable within a specific future time period. Based on the comparison between the instantaneous probability values ​​and a set threshold, an adaptive inspection decision process is triggered, generating an intelligent work order containing the coordinates of the inspection target, priority, and a list of inspection items. The intelligent work order is then sent to a mobile inspection terminal, which performs the inspection and transmits the on-site verification results back. These results are used to calibrate the parameters of the improved fault probability prediction algorithm.

[0022] In one embodiment of the present invention, the improved fault probability prediction algorithm employs an improved gradient boosting decision tree algorithm. A multidimensional feature sequence is extracted from a historical fault database. This sequence includes cable sheath temperature time series, conductor current load time series, joint vibration amplitude time series, air humidity time series, historical fault occurrence times, and corresponding fault type codes. The multidimensional feature sequence is divided into fixed time windows, with each window labeled with whether a fault occurred within a subsequent unit of time, forming a model training sample set. Multiple basic decision tree models are initialized on the sample set and integrated through weighted voting to form a preliminary fault probability prediction model. The construction process of the environmental stress evolution model is as follows: historical meteorological data along the cable route is obtained, including hourly ambient temperature, solar radiation intensity, rainfall, and wind speed; a meteorological parameter coupling influence equation describing the coupling effect of the above meteorological parameters on the cable temperature field is established, and the coupling coefficient in the equation is fitted using multivariate regression analysis; combined with the thermal parameters of the cable material and the heat dissipation parameters during laying, an environmental stress evolution model is constructed with meteorological parameters as input and cable sheath temperature changes as output, and is verified and fine-tuned through real-time environmental monitoring data. The predicted offset of cable sheath temperature output by the environmental stress evolution model is used as a dynamic penalty term and incorporated into the loss function of the fault probability prediction model. This allows the model to penalize deviations from the state caused by the future environment when learning historical patterns. Through iterative training, the loss function converges, resulting in an improved gradient boosting decision tree algorithm. This algorithm receives new cable health data packets and outputs the instantaneous probability values ​​of different levels of faults in each monitoring section within a specific future time period.

[0023] In this implementation, the improved fault probability prediction algorithm involved in this embodiment is a specific implementation of the improved gradient boosting decision tree algorithm, as well as the construction steps of the environmental stress evolution model. The multidimensional feature sequences extracted from the historical fault database include cable sheath temperature time series, conductor current load time series, joint vibration amplitude time series, air humidity time series, historical fault occurrence times, and corresponding fault type codes. These data originate from long-term monitoring records and fault event logs. In the specific implementation, the multidimensional feature sequences are divided into fixed-length time windows. The length of each time window can be set from several minutes to several hours depending on the actual application scenario. The label indicating whether a fault occurred within the subsequent unit time corresponding to each time window is marked based on historical fault records, thereby constructing a sample set for model training.

[0024] In specific implementation, multiple basic decision tree models are initialized on the sample set, each with different initial parameter settings. These basic decision tree models are integrated through weighted voting to form a preliminary fault probability prediction model, where the weight of each basic decision tree model is related to its performance on the validation set. In some embodiments, the construction process of the environmental stress evolution model is as follows: Historical meteorological data of the area along the target cable is obtained. The historical meteorological data includes hourly recorded ambient temperature, solar radiation intensity, rainfall, and wind speed. This data can be obtained from meteorological observation stations or dedicated meteorological sensors. A meteorological parameter coupling influence equation is established. This equation describes the coupling relationship between changes in ambient temperature, solar radiation intensity, rainfall, and wind speed on the temperature field of the cable body. The expression of this equation is as follows:

[0025] in: This indicates the predicted offset of the cable sheath temperature. Indicates the change in ambient temperature. This represents the change in solar radiation intensity. Indicates the change in rainfall. This indicates the change in wind speed. , , , These represent the coupling coefficients of various meteorological parameters. Based on historical meteorological data, the coupling coefficients in the meteorological parameter coupling influence equation are fitted using a multiple regression analysis method to obtain the quantified meteorological parameter coupling influence equation. The thermal parameters of the cable material and the heat dissipation parameters of the cable laying method are obtained. Combined with the quantified meteorological parameter coupling influence equation, an environmental stress evolution model is constructed with ambient temperature, solar radiation intensity, rainfall, and wind speed as inputs and changes in cable sheath temperature as output.

[0026] In practical implementation, the predicted offset of cable sheath temperature calculated by the environmental stress evolution model is used as a dynamic penalty term and incorporated into the loss function calculation of the fault probability prediction model. This dynamic penalty term is proportional to the absolute value of the predicted offset, enabling the fault probability prediction model to make penalized corrections for future state deviations caused by environmental changes when learning historical patterns. Through iterative training, the loss function value after incorporating the dynamic penalty term gradually decreases and eventually converges, resulting in an improved gradient boosting decision tree algorithm. Upon receiving new cable health data packets, the improved gradient boosting decision tree algorithm can calculate and output the instantaneous probability values ​​of different levels of faults occurring in each monitoring section of the cable within a specific future time period, based on the input feature data and the output of the environmental stress evolution model.

[0027] In one embodiment of the present invention, see [reference] Figure 2 Edge computing nodes receive sensor data streams from different physical interfaces and perform timestamp alignment and unit standardization on the data streams of cable sheath temperature, conductor current load, joint vibration amplitude, and air humidity. For the timestamp-aligned cable sheath temperature data, a sliding window filtering technique is used to remove spike noise caused by instantaneous changes in sunlight. For the unit-standardized conductor current load data, the real-time load percentage is calculated based on the cable model nameplate parameters. For the unit-standardized joint vibration amplitude data, frequency domain transformation is performed and the amplitude values ​​of characteristic frequencies are extracted. For the unit-standardized air humidity data, its gradient over a unit of time is calculated. The processed real-time load percentage of cable sheath temperature, conductor current load, characteristic frequency amplitude values ​​of joint vibration amplitude, and air humidity gradient are encapsulated together with the original acquisition time and sensor geographical location information to form a structured cable health data package.

[0028] In its specific implementation, this embodiment involves transmitting real-time operating status data collected by a sensor array to an edge computing node and performing preprocessing and standardization to form a cable health data packet. The edge computing node receives sensor data streams from different physical interfaces. These data streams originate from temperature sensors, current transformers, vibration sensors, and humidity sensors distributed in different locations. Due to differences in sampling clocks and communication delays among the sensors, the edge computing node first performs timestamp alignment and unit standardization on the sensor data streams for cable sheath temperature, conductor current load, joint vibration amplitude, and air humidity. Timestamp alignment involves interpolating and aligning data points from different sources based on a unified system time reference. Unit standardization involves converting temperature data to degrees Celsius, current data to amperes, vibration data to millimeters, and humidity data to relative humidity percentages to ensure that all data are under a consistent measurement system. In practical implementation, a sliding window filtering technique is used to remove spike noise caused by instantaneous changes in solar radiation from the timestamp-aligned cable sheath temperature data. The window width of the sliding window filtering technique can be set according to the typical time scale of solar radiation changes. Within the window, a moving average calculation is performed on the temperature data to smooth instantaneous fluctuations and retain trend information reflecting the true thermal state of the cable. For the unit-unified conductor current load data, the edge computing nodes combine the rated current carrying capacity parameters in the cable model nameplate parameters to calculate the real-time load rate percentage. The calculation formula is as follows:

[0029] in: This represents the real-time load percentage. This represents the measured value of conductor current load after unit standardization. This represents the rated current carrying capacity from the cable model nameplate parameters. For the standardized joint vibration amplitude data, a Fast Fourier Transform (FFT) is performed to achieve frequency domain transformation, extracting the amplitude values ​​of characteristic frequencies related to cable joint loosening or external mechanical interference. These characteristic frequencies are pre-set based on the joint structure and common fault modes. For the standardized air humidity data, its gradient over a unit of time is calculated, i.e., the ratio of the humidity difference between adjacent sampling times to the time interval, to characterize the dynamic changes in ambient humidity. In practical implementation, the edge computing node encapsulates the processed cable sheath temperature, real-time conductor current load percentage, characteristic frequency amplitude values ​​of joint vibration amplitude, and air humidity gradient, along with the original acquisition time and sensor geographical location information, into a structured cable health data packet. This data packet is organized in a predefined binary or JSON format, facilitating efficient transmission over the network and direct parsing and use by subsequent improved fault probability prediction algorithms.

[0030] In one embodiment of the present invention, a high-level fault probability threshold and a medium-level fault probability threshold are set. The instantaneous probability values ​​of each monitoring segment output by the improved fault probability prediction algorithm are compared with the two thresholds respectively. For monitoring segments whose instantaneous probability values ​​exceed the high-level fault probability threshold, their coordinates, current probability values, and corresponding fault levels are marked as emergency inspection targets and assigned the highest priority. For monitoring segments whose instantaneous probability values ​​exceed the medium-level fault probability threshold but not the high-level threshold, the growth rate of their probability values ​​is calculated. Those with a fast growth rate are assigned high priority, and those with a slow growth rate are assigned standard priority. Based on the predicted potential fault type, corresponding detection operations are matched from a preset detection item library to form a detection item list exclusive to each inspection target. The coordinates, priorities, and detection item lists of all inspection targets are integrated to generate a smart work order for the day or shift. In the adaptive inspection decision process, work order optimization based on resource constraints is further performed: see [link to relevant documentation]. Figure 3 The system obtains the number of currently available mobile inspection terminals, the skills and qualifications of maintenance personnel, and the range and route information of inspection vehicles. It inputs the geographical coordinates of all inspection targets in the smart work order into the route planning model, and calculates the theoretical shortest path and required working time to cover all targets by combining the range and route information. If the working time exceeds a preset threshold, it evaluates the impact of removing low-priority targets on the overall risk coverage rate in order of priority from low to high. Under the premise of ensuring that the risk coverage rate is not lower than the preset safety level, it generates a simplified smart work order consisting of high-priority targets and replans the actual inspection route for them.

[0031] In practical implementation, this embodiment involves triggering an adaptive inspection decision process based on the instantaneous probability value output by the improved fault probability prediction algorithm, and optimizing work orders based on resource constraints. The adaptive inspection decision process sets high-level fault probability thresholds and medium-level fault probability thresholds. These two thresholds are pre-configured values ​​based on historical maintenance records and risk assessment strategies. For example, the high-level fault probability threshold is set to 0.15, and the medium-level fault probability threshold is set to 0.08, used to distinguish potential fault risks of different urgency levels. In practical implementation, the instantaneous probability values ​​of each monitoring section of the cable output by the improved fault probability prediction algorithm are compared with the high-level and medium-level fault probability thresholds respectively. For monitoring sections whose instantaneous probability values ​​exceed the high-level fault probability threshold, the system automatically marks their coordinates, current probability value, and corresponding fault level as emergency inspection targets and assigns them the highest priority in the work order generation logic. For monitoring sections whose instantaneous probability values ​​exceed the medium-level fault probability threshold but do not exceed the high-level fault probability threshold, the growth rate of their probability values ​​is calculated. The formula for calculating the growth rate is:

[0032] in: Indicates the rate of growth trend. This represents the instantaneous probability value at the current moment. Indicates the previous time interval The instantaneous probability value at any given moment. Monitoring sections are prioritized based on their growth rate; sections with faster growth rates are assigned higher priority, while those with slower growth rates are assigned standard priority. In practice, based on the potential fault types predicted by the improved fault probability prediction algorithm, corresponding detection operations are matched from a pre-set detection item library. For example, if a joint overheating fault is predicted, infrared temperature measurement and contact resistance testing are matched, forming a unique detection item list for each inspection target. The coordinates, priorities, and detection item lists of all inspection targets are integrated to generate a smart work order for the day or shift. The structure of the smart work order is shown in Table 1.

[0033] Table 1: Smart Work Order Structure Table Inspection target coordinates Instantaneous probability value Fault Level Priority List of testing items (x1,y1) 0.21 high Highest Infrared temperature measurement and partial discharge detection (x2, y2) 0.12 middle high Visual inspection, vibration test In some embodiments, the adaptive inspection decision process further executes a resource-constrained work order optimization step. The number of currently available mobile inspection terminals, maintenance personnel skill and qualification information, and the range and route information of inspection vehicles are obtained and synchronized in real time from the operation and maintenance resource management system. The geographical coordinates of all inspection targets in the intelligent work order are input into the path planning model. Combined with the range and route information of the inspection vehicles, the shortest path algorithm is used to calculate the theoretical shortest path covering all inspection targets and its required working time. The working time required for the theoretical shortest path is compared with the currently available human and time resources. If the required working time exceeds a preset threshold, such as exceeding the maximum working time for a single shift, work order optimization is initiated. Work order optimization proceeds according to the priority of inspection targets, starting with the lowest priority targets, and assesses the impact of removing them from the daily work order on the overall risk coverage rate. The overall risk coverage rate is defined as the proportion of the total risk contribution of the remaining inspection targets. Under the premise of ensuring that the overall risk coverage is not lower than the preset safety level, a simplified intelligent work order consisting of the highest priority and high priority inspection targets is generated, and the actual inspection path is replanned for the simplified intelligent work order to adapt to the actual resource constraints.

[0034] In one embodiment of the present invention, the actual operating status description, fault confirmation information, or normal operation confirmation information of the cable are extracted from the field verification results returned by the mobile inspection terminal. The field verification results are compared with the prediction results of the improved fault probability prediction algorithm for the corresponding monitoring section in the same time period to generate a prediction deviation record. All prediction deviation records within a specific period are statistically analyzed, and the average prediction accuracy and false alarm rate of the algorithm for various types of faults are calculated. When the average prediction accuracy is lower than a set lower limit or the false alarm rate is higher than a set upper limit, the algorithm parameter calibration process is triggered. This calibration process uses samples containing recent field verification results to perform small-scale iterative retraining of the weight parameters of the basic decision tree model in the improved fault probability prediction algorithm until the average prediction accuracy and false alarm rate on the new validation set recover to an acceptable range.

[0035] In its specific implementation, this embodiment involves a procedure for calibrating the parameters of an improved fault probability prediction algorithm using on-site verification results transmitted from a mobile inspection terminal. The actual operating status description of the cable, fault confirmation information, or normal operation confirmation information are extracted from the on-site verification results transmitted from the mobile inspection terminal. This information includes the cable's appearance, measured joint temperature, partial discharge detection results, and conclusions regarding the presence or absence of a fault, as recorded by maintenance personnel on-site. In practice, the on-site verification results are compared with the prediction results made by the improved fault probability prediction algorithm for the same monitoring section location during the same time period. For example, the confirmed fault occurrence on-site is compared with the fault probability value output by the algorithm during the same time period, generating a prediction deviation record. The prediction deviation record includes the monitoring section identifier, prediction time range, prediction probability value, actual verification results, and the consistency status of the two.

[0036] The system collects all prediction deviation records within a specific period, the length of which can be set to one week or one month depending on operational needs. Based on the statistical results, it calculates the average prediction accuracy and false alarm rate of the improved fault probability prediction algorithm for various types of faults. The formula for calculating the average prediction accuracy is as follows:

[0037] in: This represents the average prediction accuracy. This indicates the number of records where the prediction results are consistent with the field verification results within a specific period. This represents the total number of records of field verification results within a specific period. The false alarm rate (FAR) refers to the proportion of records predicted as faults by the algorithm but verified as normal in the field, out of all records predicted as faults. When the average prediction accuracy falls below a set lower limit or the FAR exceeds a set upper limit, the system automatically triggers the algorithm parameter calibration process. In some embodiments, the algorithm parameter calibration process uses samples containing recent field verification results to construct a calibration training set, performs small-scale iterative retraining of the weight parameters of the basic decision tree model in the improved fault probability prediction algorithm, adjusts the weight allocation of the basic decision tree model in the ensemble, until the average prediction accuracy and FAR return to an acceptable range on the new validation set. Refer to Table 2 for a schematic structure of the prediction deviation record statistics.

[0038] Table 2: Statistical Table of Prediction Deviation Records Monitoring section Predicted time Predicted Fault Level Predicted probability Actual verification results Consistency judgment S001 2025-03-0110:00 high 0.22 Overheating fault of connector Consistent S002 2025-03-0214:30 middle 0.11 Normal operation Inconsistent In one embodiment of the present invention, pattern features are extracted from continuously input cable health data packets at an edge computing node. The extracted features include the periodicity of load percentage changes, the stability of the temperature difference between the cable sheath temperature and the ambient temperature, and the variance of the amplitude of the joint characteristic frequency. The extracted pattern features are matched with standard modal features in a cable operating modal feature library. This feature library contains feature templates for various preset modes, including normal operation, mild overload, continuous overload, external stress interference, and early insulation aging. The matching process is as follows: the multidimensional Euclidean distance between the extracted pattern feature vector and each standard modal feature vector in the library is calculated, and the standard modal type corresponding to the minimum distance is selected. If the minimum distance is less than or equal to the modal matching confidence threshold, the current operating state is determined to match the modal type. If the minimum distance is greater than the threshold, the current pattern feature vector is stored in the temporary storage area for the modes to be analyzed, and the feature vectors in this area are periodically clustered. If a new stable cluster category is found, its feature mean is added to the library as a new standard modal feature vector. When the matching result is an abnormal mode, the mode type, start time and feature parameters are recorded. If the same abnormal mode appears continuously in multiple consecutive time windows, an early abnormal warning signal containing the abnormal type, location, duration and key parameters is generated. This signal is fused with the instantaneous probability value output by the fault probability prediction algorithm to correct the determination of the priority of inspection targets in the adaptive inspection decision process.

[0039] In this embodiment, the online identification and early warning of cable operating modes are performed. At the edge computing node, pattern features are extracted from continuously input cable health data packets. These extracted features include at least the periodicity of load percentage changes, the stability of the temperature difference between the cable sheath and ambient temperature, and the variance of the amplitude of the joint characteristic frequency. These features reflect the periodicity of the cable's operating state, its thermal balance characteristics, and the stability of its mechanical vibration. In practice, the extracted pattern features are matched with standard modal features in a cable operating mode feature library. This library contains feature templates for various preset modes, including normal operation, light overload, continuous overload, external stress interference, and early insulation aging. Each feature template consists of typical value ranges or center values ​​for each mode feature under the corresponding mode.

[0040] In practice, the matching process is achieved by calculating the multidimensional Euclidean distance between the extracted mode feature vector and each standard mode feature vector in the cable operation mode feature library. The formula for calculating the multidimensional Euclidean distance is as follows:

[0041] in: Indicates the relationship with the first Multidimensional Euclidean distance of standard modal eigenvectors Indicates the extracted first Each pattern feature value, This represents the first characteristic in the cable operating mode feature library. The first standard modality eigenvector Each feature component This represents the number of mode features. The smallest distance is selected from all calculated multidimensional Euclidean distances, and the corresponding standard modal feature vector representing the operating mode type is recorded. A modal matching confidence threshold is set. If the smallest multidimensional Euclidean distance is less than or equal to the modal matching confidence threshold, the current operating state of the cable is determined to match the corresponding operating mode type; if the smallest multidimensional Euclidean distance is greater than the modal matching confidence threshold, the current operating state of the cable is determined to be an undefined mode, and the current mode feature vector is stored in the temporary storage area for modes to be analyzed. In some embodiments, the mode feature vectors in the temporary storage area for modes to be analyzed are periodically clustered. An unsupervised clustering algorithm is used to identify the clustering patterns in the feature space. If a new stable cluster category is found, its feature mean is used as a new standard modal feature vector and added to the cable operating mode feature library to expand the coverage of modal recognition.

[0042] When the matched standard modal characteristics are modes outside of normal operation, the system records the mode type, start time, and parameter values ​​of the mode characteristics. If the same abnormal mode appears continuously within multiple consecutive time windows, for example, if a slight overload mode is matched in three consecutive time windows, an early anomaly warning signal is generated. The early anomaly warning signal includes the abnormal mode type, location of occurrence, duration, and key characteristic parameters. In specific implementation, the early anomaly warning signal is fused and analyzed with the instantaneous probability value output by the improved fault probability prediction algorithm. For example, when the early anomaly warning signal indicates the early stage of insulation aging and the instantaneous probability value output by the fault probability prediction algorithm shows an upward trend, the system automatically increases the inspection priority of the corresponding monitoring section in the adaptive inspection decision process, thereby achieving a more accurate risk response.

[0043] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A smart management method for new energy cables based on the Internet of Things, characterized in that, The method includes: An array of sensors deployed on the new energy cable body and monitoring points along the line continuously collects real-time operating status data of the cable. The real-time operating status data includes cable sheath temperature, conductor current load, joint vibration amplitude, and ambient air humidity. The real-time operating status data collected by the sensor array is transmitted to the edge computing node. The edge computing node preprocesses and standardizes the cable sheath temperature, conductor current load, joint vibration amplitude and ambient air humidity to form a cable health data package. The cable health data packet is input into the improved fault probability prediction algorithm, which is based on a historical fault database and an environmental stress evolution model to calculate the instantaneous probability value of different levels of faults occurring in each monitoring section of the cable within a specific future time period. Based on the instantaneous probability value output by the improved fault probability prediction algorithm, an adaptive inspection decision process is triggered. The adaptive inspection decision process is used to generate an intelligent work order containing the coordinates of the inspection target, the inspection priority, and the list of inspection items. The intelligent work order is sent to the mobile inspection terminal, and the on-site verification results returned by the mobile inspection terminal after performing the inspection are received. The parameters of the improved fault probability prediction algorithm are calibrated using the on-site verification results, including: From the field verification results returned by the mobile inspection terminal, extract the description of the actual operating status of the cable, fault confirmation information or normal operation confirmation information; The on-site verification results are compared with the prediction results made by the improved fault probability prediction algorithm for the location of the monitoring section in the same time period to generate a prediction deviation record. Collect all prediction deviation records within a specific period, and calculate the average prediction accuracy and false alarm rate of the improved fault probability prediction algorithm for various types of faults; When the average prediction accuracy is lower than the set lower limit or the false alarm rate is higher than the set upper limit, the algorithm parameter calibration process is triggered. The algorithm parameter calibration process uses samples containing recent field verification results to perform small-scale iterative retraining of the weight parameters of the basic decision tree model in the improved fault probability prediction algorithm until the average prediction accuracy and false alarm rate on the new validation set are restored to an acceptable range.

2. The intelligent management method for new energy cables based on the Internet of Things according to claim 1, characterized in that, The improved fault probability prediction algorithm is an improved gradient boosting decision tree algorithm, and its working principle includes: Extract multidimensional feature sequences from the historical fault database. The multidimensional feature sequences include at least the cable sheath temperature time series, conductor current load time series, joint vibration amplitude time series, air humidity time series, historical fault occurrence time, and corresponding fault type codes. The multidimensional feature sequence is divided into time windows, and a label is added to indicate whether a fault occurs in the subsequent unit of time for each time window, thus constructing a sample set for model training. On the sample set, multiple basic decision tree models are initialized, and the multiple basic decision tree models are integrated by weighted voting to form a preliminary fault probability prediction model. The environmental stress evolution model is introduced to simulate the coupled effects of future changes in ambient temperature, solar radiation intensity, and rainfall on the thermodynamic state of the cable, and to calculate the predicted offset of the cable sheath temperature. The predicted offset of the cable sheath temperature is incorporated as a dynamic penalty term into the loss function calculation process of the fault probability prediction model, so that the fault probability prediction model can make penal corrections for future state deviations caused by environmental changes when learning historical patterns. Through iterative training, the loss function value after incorporating the dynamic penalty term converges, resulting in the improved gradient boosting decision tree algorithm. When the improved gradient boosting decision tree algorithm receives a new cable health data packet, it outputs the instantaneous probability value of different levels of faults occurring in each monitoring section of the cable within a specific future time period.

3. The intelligent management method for new energy cables based on the Internet of Things according to claim 1, characterized in that, The real-time operating status data collected by the sensor array is transmitted to an edge computing node. The edge computing node preprocesses and standardizes the cable sheath temperature, conductor current load, joint vibration amplitude, and ambient air humidity to form a cable health data package, including: The edge computing node receives sensor data streams from different physical interfaces and performs timestamp alignment and unit unification processing on the sensor data streams of cable sheath temperature, conductor current load, joint vibration amplitude and air humidity. For the timestamp-aligned cable sheath temperature data, a sliding window filtering technique is used to remove spike noise caused by instantaneous changes in solar radiation; The real-time load rate percentage is calculated by combining the standardized conductor current load data with the cable model nameplate parameters. For the unit-standardized joint vibration amplitude data, frequency domain transformation is performed to extract the amplitude values ​​of characteristic frequencies; Calculate the gradient of change of air humidity data over a unit of time after unitization. The processed cable sheath temperature, real-time load rate percentage of conductor current load, characteristic frequency amplitude value of joint vibration amplitude, and air humidity change gradient are packaged together with the original acquisition time and sensor geographical location information to form the structured cable health data package.

4. The intelligent management method for new energy cables based on the Internet of Things according to claim 1, characterized in that, Based on the instantaneous probability value output by the improved fault probability prediction algorithm, an adaptive inspection decision process is triggered, including: Set high-level fault probability thresholds and medium-level fault probability thresholds; The instantaneous probability values ​​of each monitoring section of the cable output by the improved fault probability prediction algorithm are compared with the high-level fault probability threshold and the medium-level fault probability threshold, respectively. For monitoring sections where the instantaneous probability value exceeds the high-level fault probability threshold, their coordinates, current probability value, and corresponding fault level are marked as emergency inspection targets and assigned the highest priority. For monitoring sections whose instantaneous probability values ​​exceed the medium-level fault probability threshold but do not exceed the high-level fault probability threshold, calculate the growth rate of their probability values, assign high priority to monitoring sections with fast growth rate, and assign standard priority to monitoring sections with slow growth rate. Based on the potential fault types predicted by the improved fault probability prediction algorithm, the corresponding detection operations are matched from the preset detection item library to form a unique detection item list for each inspection target. Integrate the coordinates, priorities, and inspection item lists of all inspection targets to generate the aforementioned smart work order for the day or shift.

5. The intelligent management method for new energy cables based on the Internet of Things according to claim 4, characterized in that, The method also includes a work order optimization step based on resource constraints in the adaptive inspection decision process: Obtain the number of currently available mobile inspection terminals, the skill and qualification information of maintenance personnel, and the range and route information of inspection vehicles; The geographical coordinates of all inspection targets in the intelligent work order are input into the path planning model. Combined with the driving range and path information of the inspection vehicle, the theoretical shortest path covering all inspection targets and the required working time are calculated. The required time for the theoretical shortest path is compared with the currently available manpower and time resources. If the required time exceeds a preset threshold, work order optimization is initiated. The work order optimization is based on the priority of the inspection targets, starting with the lowest priority inspection targets, and assessing the impact of removing them from the work orders of the day on the overall risk coverage. Under the premise of ensuring that the overall risk coverage is not lower than the preset safety level, a simplified intelligent work order consisting of the highest priority and high priority inspection targets is generated, and the actual inspection path is replanned for the simplified intelligent work order.

6. The intelligent management method for new energy cables based on the Internet of Things according to claim 1, characterized in that, The method also includes online identification of cable operating modes and early warning of anomalies: In the edge computing node, pattern features are extracted from the continuously input cable health data packets. The pattern features include at least the change period of the load rate percentage, the temperature difference stability between the cable sheath temperature and the ambient temperature, and the fluctuation variance of the amplitude value of the joint characteristic frequency. The extracted mode features are matched with the standard mode features in the cable operation mode feature library, which contains feature templates for various preset modes such as normal operation, light overload, continuous overload, external stress interference, and early stage of insulation aging. When the matched standard modal features are modalities other than those in normal operation, record the modality type, start time, and parameter values ​​of the modality features; If the same anomalous mode continues to occur within multiple consecutive time windows, an early anomaly warning signal is generated. The early anomaly warning signal includes the anomalous mode type, the location of occurrence, the duration, and key characteristic parameters. The early anomaly warning signal is fused and analyzed with the instantaneous probability value output by the improved fault probability prediction algorithm to correct the determination of the priority of inspection targets in the adaptive inspection decision process.

7. The intelligent management method for new energy cables based on the Internet of Things according to claim 6, characterized in that, The step of matching the extracted pattern features with the standard modal features in the cable operation modal feature library includes: Calculate the multidimensional Euclidean distance between the extracted mode feature vector and each standard mode feature vector in the cable operation mode feature library; From all the calculated multidimensional Euclidean distances, select the distance with the smallest value and record the operating mode type represented by its corresponding standard mode feature vector; A modal matching confidence threshold is set. If the minimum multidimensional Euclidean distance is less than or equal to the modal matching confidence threshold, then the current operating state of the cable is determined to match the operating mode type. If the minimum multidimensional Euclidean distance is greater than the modal matching confidence threshold, the current operating state of the cable is determined to be undefined mode, and the current mode feature vector is stored in the temporary storage area of ​​the mode to be analyzed. The mode feature vectors in the temporary storage area of ​​the modes to be analyzed are periodically clustered. If a new stable cluster is found, its feature mean is used as a new standard mode feature vector and added to the cable operation mode feature library.

8. The intelligent management method for new energy cables based on the Internet of Things according to claim 2, characterized in that, The steps for constructing the environmental stress evolution model include: Obtain historical meteorological data for the area along the target cable route, including hourly recorded ambient temperature, solar radiation intensity, rainfall, and wind speed. A coupled influence equation for meteorological parameters is established, which is used to describe the coupling relationship between changes in ambient temperature, solar radiation intensity, rainfall, and wind speed on the temperature field of the cable body. Based on the historical meteorological data, the coupling coefficients in the meteorological parameter coupling influence equation are fitted by the multiple regression analysis method to obtain the quantified meteorological parameter coupling influence equation. The thermal parameters of the cable material and the heat dissipation parameters of the cable laying method are obtained. Combined with the quantified meteorological parameter coupling effect equation, an environmental stress evolution model is constructed with ambient temperature, solar radiation intensity, rainfall and wind speed as inputs and cable sheath temperature change as output. The environmental stress evolution model is continuously validated and its parameters are fine-tuned using real-time environmental monitoring data to ensure its accuracy in predicting the trend of cable sheath temperature changes.

9. An intelligent management system for new energy cables based on the Internet of Things, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent management method for new energy cables based on the Internet of Things as described in any one of claims 1 to 8.

Images