A method for monitoring temperature rise of a cable branch box joint

Abstract

The application relates to the technical field of power equipment temperature monitoring, and discloses a cable branch box joint temperature rise monitoring method, which is used for solving the problems that the surface temperature is difficult to reflect the internal real temperature rise in time, early warning lag or missed report is caused, and the surface temperature of adjacent joints is abnormally high, which causes false report in traditional methods; the method first collects and checks the surface temperature of the joint, the box internal environment temperature and the load current, separates the environment temperature influence and the interphase thermal interference, compensates the surface temperature rise in combination with the heat transfer lagging relationship, then determines the internal temperature rise according to the structure heat transfer relationship and discriminates the abnormality of the change state, and accurate identification and timely early warning of the internal abnormal heating of the cable branch box joint are realized.

Description

Technical Field

[0001] This invention relates to the field of temperature monitoring technology for power equipment, specifically a method for monitoring the temperature rise of cable branch box joints. Background Technology

[0002] During long-term operation, cable joints in cable distribution boxes are prone to abnormal heating at the crimping points due to increased contact resistance. If this is not detected and addressed in time, it can easily lead to a decline in insulation performance and potentially cause the fault to escalate. For example, the published invention patent application CN104776938B discloses a method and system for inverting the cable core temperature of cable joints based on cable surface temperature. It mainly calculates the temperature of the cable core contact point by collecting the surface temperature of the joint and combining it with a thermal circuit model and simulation fitting. Another example is the published invention patent application CN113686461B, which discloses a T-type cable joint self-powered temperature sensor, a cable joint fault detection system and method. It mainly estimates the temperature of the internal metal conductor of the joint by collecting the hot end temperature and the cold end temperature and combining it with a non-invasive algorithm. The above technical solutions improve the temperature monitoring capability of cable joints to a certain extent, but their technical ideas are still mainly based on surface temperature collection and internal temperature estimation. However, existing technologies still have certain limitations under alternating load conditions in cable branch boxes. First, the external insulation layer of cable joints typically has high thermal resistance and thermal capacity, resulting in a significant lag in the heat transfer process from the inside of the joint to the surface. This makes it difficult for the surface temperature to reflect the true internal temperature rise in a timely manner, thus affecting the accuracy of internal temperature rise estimation based on surface temperature and easily causing delayed or missed warnings. Second, the internal space of cable branch boxes is relatively enclosed, and the installation distance between three-phase joints is usually close. When one phase joint heats up, it can easily affect adjacent joints through thermal radiation and air convection, forming interphase thermal interference. This can lead to abnormally high surface temperatures of adjacent joints, easily causing false alarms. Therefore, how to compensate for the heat transfer lag caused by the insulation layer without damaging the original high-voltage shielding structure of the joint, and how to effectively separate interphase thermal interference in the enclosed space, has become a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method for monitoring the temperature rise of cable branch box joints. This method solves the problems of traditional methods, such as the difficulty in timely reflecting the actual internal temperature rise due to surface temperature, which can lead to delayed or missed warnings, and the potential for false alarms due to abnormally high surface temperatures of adjacent joints.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for monitoring the temperature rise of cable branch box joints includes: S1 collects the surface temperature of each phase joint in the cable branch box, the ambient temperature inside the box, and the load current of each phase, and performs time alignment and abnormal data verification on the collected data. S2, based on the spatial arrangement relationship of each phase joint, the temperature distribution relationship inside the box and the interphase influence relationship, separates the influence of ambient temperature and interphase thermal interference in the surface temperature of each phase joint; S3, based on the load change process and the heat transfer hysteresis relationship of the joint insulation layer, performs hysteresis compensation on the surface temperature of each phase joint after thermal interference separation. S4. Based on the surface temperature of each phase joint and the heat transfer relationship of the joint structure after hysteresis compensation, the internal temperature rise of each phase joint is determined, and the internal temperature rise is correlated with the load change process. S5 performs anomaly detection based on the temperature rise changes inside each phase joint and outputs the detection results.

[0005] Furthermore, S1 includes: Assign a unified timestamp to the surface temperature of the fixed measuring point on the outer insulation layer of the three-phase connector, the ambient temperature of the upper, middle and lower parts of the box, and the corresponding load current, and align them according to a unified analysis window; Periods of load disturbance and environmental disturbance are marked, and anomaly verification, status identification, local repair, channel retry and continuity constraints are performed respectively to form an effective monitoring sequence and status record.

[0006] Furthermore, S2 includes: Establish spatial relationship information between the three-phase connector and the environmental measurement points inside the enclosure; The local ambient background temperature is determined based on the spatial correspondence between the target connector and the upper, middle and lower environmental measurement points, and the surface temperature is then corrected for environmental conditions. Identify the thermal impact state based on the three-phase load relationship and the environmentally corrected temperature rise relationship; Subtract the synchronous micro-amplitude crosstalk component under low-impact conditions; Construct or update the interphase influence matrix and estimate the thermal influence components of adjacent phases under enhanced thermal disturbance conditions; Through non-negative constraints, continuity constraints, and abnormal freezing and recovery processing, a net surface temperature rise sequence and corresponding state information are formed.

[0007] Furthermore, the thermally affected state is identified based on the three-phase load relationship and the environmentally corrected temperature rise relationship, including: The current thermal impact state is determined based on the interphase load deviation, the interphase environmental correction temperature rise deviation, and the degree of deviation of a single phase relative to the average values ​​of the other two phases. The current thermal impact state is then divided into low impact state, intermediate transition state, and enhanced thermal interference state. The intermediate transition state is first processed using the basic separation method, and after maintaining this state for multiple consecutive analysis windows, it is switched to the enhanced separation method.

[0008] Furthermore, S3 includes: Compensation determination is made based on the segmented state of load current and the relationship between net surface temperature rise and load change direction. The heat transfer hysteresis parameter template that matches the joint specification is called, the compensation amount is determined according to different operating states, and combined with parameter updates, version rollback, short-term missing conservative compensation and low-confidence freeze processing, a corrected surface temperature rise sequence and corresponding state information are formed.

[0009] Furthermore, compensation is determined based on the segmented state of the load current and the relationship between the net surface temperature rise and the direction of load change, including: Window analysis is performed on the load current, and steady-state, load increase, load decrease and fluctuation states are identified based on the current change rate, cumulative increase and decrease, and direction reversal. When the identification result is an increased load, decreased load, or fluctuating state, or when the direction of change of net surface temperature rise is continuously mismatched with the direction of change of corresponding load, hysteresis compensation determination is triggered.

[0010] Furthermore, S4 includes: Call upon the structural heat transfer template that matches the target joint; Based on the corrected surface temperature rise, the temperature rise of the internal core area is determined layer by layer according to the layered temperature difference mapping relationship; The current main value of internal temperature rise is determined by verifying the consistency between short and long windows; The status is determined by combining the template replacement status, confidence level, and the correlation between internal temperature rise and load change; Record the internal temperature rise sequence, rate of change sequence, and corresponding state information.

[0011] Furthermore, based on the corrected surface temperature rise and the layered temperature difference mapping relationship, the temperature rise of the internal core region is determined layer by layer, including: Based on the interlayer temperature difference mapping coefficient in the structural heat transfer template; The order is from the outer surface to the middle of the insulation layer, and from the middle of the insulation layer to the crimping area. Combine the current load range with the corrected surface temperature rise range; The temperature difference increment of each layer is determined by looking up a table or interpolation, and the temperature difference increment of each layer is sequentially added to the temperature rise of the correction surface. When the structural heat transfer template has more layers, it continues to accumulate according to the layer order given by the template.

[0012] Furthermore, S5 includes: When the internal temperature rise discrimination condition is met, a stratified threshold is constructed based on the historical baseline of internal temperature rise corresponding to the current load range and the thermal background range. The anomaly level is determined by combining the absolute value of internal temperature rise, the rate of rise, the duration of high temperature, the deviation from load changes, and the three-phase group discrimination relationship. It performs verification, recalculation, or rollback when there are critical states, changes in key intermediate quantities, or data link anomalies, and records the corresponding status information.

[0013] Furthermore, the anomaly level is determined by combining the absolute value of the internal temperature rise, the rate of rise, the duration of high temperature, the deviation from load changes, and the three-phase group discrimination relationship, including: The joint abnormality level is determined by combining the following factors: the duration of internal temperature rise exceeding the limit, the deviation of the internal temperature rise rate from the load change, the duration of temperature rise after the load stabilizes or falls, the synchronous relationship of the three-phase internal temperature rise, the relationship of local environmental background changes, and the degree of deviation of the single phase.

[0014] Compared with the prior art, the present invention provides a method for monitoring the temperature rise of cable branch box joints, which has the following advantages: 1. This invention collects, aligns, and verifies the surface temperature of each phase joint, the ambient temperature inside the box, and the load current in a unified manner. It also separates the influence of ambient temperature and interphase thermal interference by combining the spatial correspondence between the joint and the environmental measurement points. Furthermore, it compensates for the surface temperature rise based on the load change process and the heat transfer hysteresis relationship of the joint insulation layer. At the same time, it uses the heat transfer template of the joint structure to back-calculate the internal temperature rise layer by layer and correlates the internal temperature rise with the load change process. This can reduce the response lag, false alarms, and missed alarms caused by directly judging based on the surface temperature alone, and achieve accurate identification and timely early warning of abnormal heating state inside the cable branch box joint.

[0015] 2. This invention acquires joint operation information by employing a non-invasive surface temperature measurement method, and combines it with local environment mapping, interphase thermal influence constraints, hysteresis compensation, internal temperature rise back calculation, and verification, recalculation, and rollback processing. Without changing the original high-voltage shielding structure of the joint, it improves the applicability of monitoring data under complex operating conditions, reduces the interference of single fluctuations or channel anomalies on the judgment conclusion, and improves the continuity, stability, and reliability of anomaly judgment of cable branch box joints. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the process for monitoring the temperature rise of a cable branch box joint according to the present invention; Figure 2 This is a schematic diagram of the monitoring objects and measuring point arrangement of the cable branch box of the present invention; Figure 3 This is a logic diagram for separating environmental impact and interphase thermal interference in this invention. Figure 4 This is a curve showing the load change and surface temperature rise hysteresis compensation of the present invention; Figure 5 This is a schematic diagram illustrating the layered heat transfer and internal temperature rise of the joint in this invention. Figure 6 This is a flowchart of the anomaly level judgment and result output of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: Figures 1-6 A method for monitoring the temperature rise of cable branch box joints is presented, including: S1 collects the surface temperature of each phase joint in the cable branch box, the ambient temperature inside the box, and the load current of each phase, and performs time alignment and abnormal data verification on the collected data. S2, based on the spatial arrangement relationship of each phase joint, the temperature distribution relationship inside the box and the interphase influence relationship, separates the influence of ambient temperature and interphase thermal interference in the surface temperature of each phase joint; S3, based on the load change process and the heat transfer hysteresis relationship of the joint insulation layer, performs hysteresis compensation on the surface temperature of each phase joint after thermal interference separation. S4. Based on the surface temperature of each phase joint and the heat transfer relationship of the joint structure after hysteresis compensation, the internal temperature rise of each phase joint is determined, and the internal temperature rise is correlated with the load change process. S5 performs anomaly detection based on the temperature rise changes inside each phase joint and outputs the detection results.

[0019] Specifically, such as Figure 2 As shown: First, a unified process for collecting, aligning, and verifying the basic validity of the surface temperature of each phase joint, the ambient temperature inside the cable branch box, and the load current of each phase is established. This process is executed continuously after the branch box is put into operation, and the edge processing terminal uniformly receives, caches, sorts, and records the status of all collected data. The cable branch box is equipped with three-phase cable connectors corresponding to phases A, B, and C, respectively. These connectors are distinguished based on the branch box's primary wiring, phase sequence markings, and on-site installation location. Phases A, B, and C are connector bodies that electrically connect to their corresponding phase conductors. The surface temperatures of the A, B, and C connectors are collected using non-invasive temperature sensors fixed to specific positions on the outer insulation layer of each connector. The measuring points are located on the insulation surface outside the crimping area, maintaining consistency in axial distance and circumferential position. This is because the temperature distribution at different locations on the connector surface is not entirely uniform; ensuring the three measuring points are in the same structural position is crucial for obtaining accurate temperature readings. Only temperature data can be horizontally comparable; the ambient temperature inside the box is obtained through three ambient temperature acquisition points arranged in the upper, middle and lower parts of the branch box. The upper acquisition point is used to reflect the temperature of the hot air accumulation area, the middle acquisition point is used to reflect the temperature of the space near the joint, and the lower acquisition point is used to reflect the temperature of the cold air compensation area. The basis for this is that there is natural convection and temperature stratification inside the fully enclosed metal branch box, and it is difficult to accurately reflect the actual thermal background near the joint using only a single environmental measurement point; the load current of each phase is obtained through current acquisition devices installed at the corresponding incoming and outgoing line positions. The current acquisition device can be a current transformer or a Rogowski coil, as long as it can stably output the current timing data corresponding to each phase joint. To enable comparison of surface temperature, ambient temperature, and load current under the same time base, the edge processing terminal, after receiving data from each channel, first appends a unified timestamp to all data, and then resamples and aligns the windows according to the unified time base. The sampling period is determined by comprehensively considering the thermal inertia of the connector's outer insulation layer, the rate of change of ambient temperature, the rate of change of load current, sensor response time, and terminal processing capability, and is set to 0.1 seconds to 1 second. Since the connector surface temperature and the ambient temperature inside the enclosure are slow variables, while the load current is a fast variable, the sampling period for temperature data can be selected as 0.5 seconds, and the sampling period for load current can be selected as 0 seconds. After 1 second, the current data is aggregated and statistically analyzed in 0.5-second windows to ensure that temperature and current data fall into the same analysis window. The above values ​​can balance the continuity of temperature tracking, the resolution of load changes, and the data processing burden, and have practical engineering basis. When aligning time, if the data packet arrives early or slightly late, it is assigned to the nearest 0.5-second alignment window according to the timestamp. If the delay time exceeds 0.5 seconds, the data packet is recorded as delayed data and written to the delayed data buffer. It is not included in the current window calculation and will be used for historical correction after subsequent data entry is completed, so as to avoid misaligned data directly entering the subsequent analysis link. After time alignment, the edge processing terminal performs disturbance marking and anomaly verification on the aligned data. Disturbance marking is used to identify periods of rapid load change or changing environmental background so that relevant data can be prioritized for subsequent processing. When the average rate of change of the load current of the same phase exceeds 5% / s of the rated current in 5 consecutive current sampling points, the current period is recorded as a load disturbance period. When the cumulative change of the ambient temperature in the upper, middle and lower parts exceeds 0.5℃ in 10 consecutive temperature sampling points, the current period is recorded as an environmental disturbance period. The basis for the above thresholds is that a continuous change of rated current of more than 5% / s usually exceeds the range of daily measurement noise and ordinary small fluctuations, and can reflect the real load disturbance that is sufficient to affect the thermal state of the joint. A cumulative change of ambient temperature of 0.5℃ within 5 seconds also exceeds the common small fluctuations under stable natural convection inside the branch box, and can reflect a perceptible change in the thermal background inside the box. Data marked as disturbance periods are not directly deleted, but are marked with disturbance tags for subsequent thermal interference separation and hysteresis compensation. Anomaly checks are performed on surface temperature, ambient temperature, and load current. For surface temperature data, the current temperature value is first checked to see if it falls within the validity check range, which is -40°C to 150°C. This range is determined based on the common range of industrial-grade temperature sensors, the operating environment of the branch box, and the anomaly rejection boundary. It is used to determine the validity of the sampled value and is not considered as the normal operating temperature limit for the connector. Next, the temperature change amplitude between two adjacent 0.5-second sampling points is checked to see if it exceeds the short-term physical allowable change range. The short-term physical allowable change range is determined by considering the thickness of the connector's outer insulation layer, the material's thermal resistance, and thermal capacity. For common distribution connectors with insulation layer thicknesses of 30 mm to 60 mm, the range is 0. With a 5-second sampling period, the range can be selected from 0.2℃ to 1.0℃, with 0.5℃ being a commonly used value. This is because the surface temperature of this type of connector is significantly constrained by thermal inertia, and under normal circumstances, it will not experience a significant jump within 0.5 seconds. Then, check whether the surface temperature change matches the load change and environmental change within the same window. When, within three consecutive temperature sampling points, the current change amplitude is less than 2% of the rated current, the cumulative change in ambient temperature is less than 0.2℃, and the surface temperature change amplitude exceeds 0.5℃, the sampling point is judged as suspicious data. The basis for this judgment rule is that, under conditions where the load and environment are basically stable, the surface temperature should not independently deviate from thermal inertia and show a significant sudden increase or decrease. For ambient temperature data, the focus is on checking whether the temperature gradient between the three spatial measurement points conforms to the natural convection pattern within the enclosed enclosure. When the branch box load current is above 60% of the rated current, and the upper temperature remains more than 0.5℃ lower than the lower temperature within 10 consecutive temperature sampling points, the environmental acquisition status is considered abnormal. The basis for these thresholds is that during medium-to-high load operation, a stable stratification of upper-heat and lower-cooling typically forms within the enclosure. If a significant reverse gradient persists, it is often caused by sensor malfunction, door opening, or external airflow intrusion. For load current data, the following checks are performed sequentially: sudden jumps, continuous zeroing, short-term saturation, and sample loss. If two adjacent 0.1-second samples... If the current change amplitude between points exceeds 30% of the rated current and does not correspond to a switch action record or load switching record, it is judged as a sudden jump anomaly; if the switch remains closed and the current returns to zero for 20 consecutive sampling points, it is judged as a continuous zeroing anomaly; if 5 consecutive sampling points reach more than 98% of the upper limit of the acquisition range, it is judged as short-term saturation; if 3 consecutive sampling points are lost, it is judged as a sample loss anomaly. The basis for the above thresholds is that, under the background of no switch action, a change of more than 30% of the rated current within 0.1 seconds is usually not a stable operating characteristic, and a continuous return to zero for 2 seconds is also difficult to match with the closed state, while continuously reaching the upper limit of the range indicates that the acquisition channel is close to the distortion area. The edge processing terminal generates a status identifier for each type of data based on the verification results. The status identifier includes at least fields such as valid, suspicious, and invalid. Suspicious data is not directly removed but instead enters a local repair process. Local repair is performed in a fixed order: first, neighboring window interpolation is used to repair isolated outliers and missing data; when missing or outliers occur consecutively and for a short duration, the same historical short window fitting is used for repair; when the ambient temperature channel fails locally, adjacent environmental points are used for collaborative correction to restore local environmental data; for invalid data, the current window is left empty, and the corresponding acquisition channel is retried; the number of retries can be selected from 1 to 3, with 2 being a commonly used value. This is because a single retrieval is insufficient to eliminate instantaneous communication jitter, while more than 3 consecutive retries significantly lengthen the current window processing time. If the problem persists after testing, the corresponding channel is marked as abnormal and processed according to the degraded path in subsequent processing. To prevent single anomalies or short-term noise from directly entering subsequent thermal analysis, a continuity constraint is also set. The number of effective sampling points in the continuity constraint is determined based on the requirements for anomaly point filtering and the response time of subsequent processing, and is set to 5 to 20 consecutive sampling points. When the temperature sampling period is 0.5 seconds, the corresponding continuous observation duration is 2.5 to 10 seconds, of which 10 consecutive sampling points can be used as a common value, corresponding to a continuous observation duration of 5 seconds. This value can filter out isolated disturbances and short-term jitter without significantly reducing the response speed of subsequent thermal interference separation and hysteresis compensation. Only when a certain number of effective sampling points are obtained consecutively within a certain time period will the edge processing terminal write the data for that time period into the effective monitoring sequence. After the above processing, effective sequences of surface temperatures of each phase joint, effective sequences of ambient temperatures inside the tank, and effective sequences of load currents of each phase are formed, and corresponding status records are generated. The above effective sequences serve as input data for subsequent separation of ambient temperature influence and interphase thermal interference, and the status records serve as auxiliary basis for subsequent judgment of data reliability and selection of anomaly handling paths.

[0020] Specifically, such as Figure 3 As shown: After the preliminary data acquisition, time alignment, and validity verification processes are completed, the influence of ambient temperature and interphase thermal interference in the surface temperature of each phase joint are separated. This process is initiated after obtaining the effective sequence of surface temperature of each phase joint, the effective sequence of ambient temperature inside the box, and the effective sequence of load current of each phase, which meet the requirements of the preceding continuity constraints. When the temperature sampling period is 0.5 seconds and the number of continuous effective sampling points is 10, the corresponding continuous observation duration is 5 seconds. This value is based on the fact that a 5-second observation duration can cover the formation process of short-term thermal disturbances inside the branch box without significantly reducing the response speed of subsequent temperature rise compensation processing. At the same time, the spatial relationship information inside the branch box is pre-saved in the edge processing terminal, and the spatial relationship information includes at least the A phase joint. The fields include the center coordinates of the A-phase, B-phase, and C-phase connectors; the relative distance and orientation between any two phase connectors; the distance between each phase connector and the inner wall of the box; and the spatial mapping relationship between each phase connector and the upper, middle, and lower ambient temperature acquisition points. The A-phase, B-phase, and C-phase connectors correspond to the three-phase connector bodies that are electrically connected to the A-phase, B-phase, and C-phase conductors in the branch box, respectively. The distinction is based on the primary wiring relationship, phase sequence identification, and on-site installation location. The thermal impact parameters related to the connector arrangement are not written into the spatial relationship information as static geometric information, but are stored separately as inter-phase influence coefficients for use according to the straight-line, triangular, or asymmetrical arrangement. First, the influence of ambient temperature is separated. Based on the effective ambient temperature sequences of the upper, middle, and lower regions formed in the previous acquisition and verification stage, the edge processing terminal establishes corresponding local ambient background temperature sequences for each phase joint. Due to natural convection and spatial stratification within the enclosed branch box, a single ambient temperature value cannot accurately reflect the actual thermal background of the area adjacent to the joint. Therefore, the positional weights are determined according to the three-dimensional spatial distances between the target joint and the upper, middle, and lower ambient temperature acquisition points. The distances from the target joint to the upper, middle, and lower ambient temperature acquisition points are denoted as the [n]th, ... The first, second, and third distances are weighted by taking the reciprocal of their respective distances and then normalizing them so that the sum of the three weights is 1. When any distance is less than the preset minimum distance, that distance is included in the calculation as the preset minimum distance to avoid abnormal weight amplification caused by local installations being too close. The preset minimum distance ranges from 20 mm to 50 mm, with 30 mm being commonly used. This value is based on the common sensor installation spacing in the branch box, the size of the outer insulation layer of the connector, and the space for local measurement point arrangement. Within this range, the ability to distinguish the positions of measurement points in different environments can be maintained, while avoiding excessive influence on the calculation results due to individual measurement points being too close. The local ambient background temperature corresponding to each phase joint is the result of weighting the upper, middle, and lower ambient temperatures according to normalized weights. The basis for this processing method is that the closer the target joint is to the ambient temperature acquisition point, the stronger the representativeness of the ambient temperature acquisition point to the thermal background around the target joint. Using the inverse distance weighting can keep the environmental correction result consistent with the spatial location of the joint, thereby avoiding the stratification error caused by directly using the average temperature of the box. If an ambient temperature acquisition point has been determined to be invalid in the previous acquisition and verification stage, the acquisition point is removed from the weighted calculation, and the remaining valid ambient temperature acquisition points are re-weighted and normalized. If only one valid ambient temperature acquisition point remains, the fine environmental mapping is paused, and only the valid ambient temperature is used as a conservative background value. At the same time, the incomplete environmental mapping state is recorded and passed to the subsequent credibility judgment unit. After obtaining the local ambient background temperature sequence corresponding to each phase joint, the interphase thermal interference is further separated. The edge processing terminal first subtracts the corresponding local ambient background temperature from the surface temperature of each phase joint to obtain the environmentally corrected temperature rise of each phase. Then, combining the three-phase current relationship and the three-phase environmentally corrected temperature rise relationship within the same analysis window, the current thermal impact state is identified. When the current difference between any two phases does not exceed 5% of the rated current, and the environmentally corrected temperature rise difference between any two phases does not exceed 1.0℃, the current state is identified as a low-impact state and processed according to the basic separation method. The above threshold values ​​are based on the fact that when the interphase current difference is within 5% of the rated current, it can generally be regarded as the three-phase load base. When the current is balanced and the environmental correction temperature rise difference is within 1.0℃, it is usually still within the range that can be explained by environmental stratification, measurement deviation and individual differences of the joints. Correspondingly, when the current of one phase is more than 10% higher than the average of the other two phases, or the environmental correction temperature rise of one phase is more than 1.5℃ higher than the average of the other two phases, the current state is identified as a state of enhanced thermal interference and is handled in an enhanced separation manner. The above threshold values ​​are based on the fact that when the single-phase load is more than 10% higher than the rated current, the possibility of single-phase heat spreading to adjacent phases is significantly increased, and when the environmental correction temperature rise of one phase is more than 1.5℃ higher than the average of the other two phases, it usually exceeds the natural difference range under normal slight imbalance conditions. For cases where the aforementioned criteria for determining enhanced thermal interference are not met, but the condition for determining low-impact state is not stably found, the basic separation method is used first. If the condition remains unchanged for three consecutive analysis windows, the enhanced separation method is used. The values ​​for the three consecutive analysis windows are based on the fact that, under the effective data conditions formed by the aforementioned acquisition, alignment, and verification processes, the three consecutive analysis windows can indicate that the current thermal state deviation is not an instantaneous fluctuation, but has a continuous change characteristic, thereby avoiding the omission of continuously accumulated interphase thermal crosstalk. Through the above processing, while maintaining the clarity of the determination criteria, the stability of switching between the basic separation method and the enhanced separation method can be improved, so that the subsequent calculation of net surface temperature rise is based on a determination that is more in line with the actual thermal impact state. Under the basic separation path, the edge processing terminal prioritizes identifying low-intensity synchronous crosstalk. When the ambient temperature rise of a certain phase increases in the current analysis window relative to the previous analysis window, and the current change amplitude of two adjacent phases within the same time period is less than 2% of the rated current, and a temperature rise of 0.2℃ to 0.5℃ occurs simultaneously in the current analysis window or the next adjacent analysis window, this part of the synchronous micro-rise is identified as a suspected crosstalk component. The above values ​​are based on the fact that, under the condition of basically unchanged load, the temperature rise of adjacent surfaces usually will not synchronously increase in the same direction beyond the measurement noise level. With a sampling period of 0.5 seconds, a temperature rise of less than 0.2℃ is considered a potential crosstalk component. The temperature fluctuation is easily affected by sensor resolution and random jitter, and changes exceeding 0.5℃ are generally no longer suitable for treating as slight crosstalk. For identified suspected crosstalk components, the current synchronization micro-increase value is compared with the average synchronization offset during historical light load balancing periods, and the smaller value is subtracted to obtain the basic separation result. When the historical light load balancing period samples are insufficient, a conservative subtraction is performed based on the current synchronization micro-increase value. The basis for using the smaller value for subtraction is that the basic separation path corresponds to a low-impact window, at which point the interphase thermal interference is relatively weak. If the entire synchronization micro-increase is directly subtracted, it is easy to excessively weaken the true heat-generating components of the target phase. Under the enhanced separation path, the edge processing terminal constructs the interphase influence matrix within the current analysis window by combining the connector arrangement relationship and operating status. The interphase influence matrix consists of pre-calibration coefficients and online correction coefficients. The pre-calibration coefficients are pre-established during the installation and commissioning phase based on the connector arrangement, interphase distance, relative orientation, and enclosure structure. The online correction coefficients are updated based on typical operating conditions identified during operation. Typical operating conditions can be selected as the light load balancing period, the single-phase significant change period, and the steady-state maintenance period. The light load balancing period is defined as the time when the current of each phase is lower than 30% of the rated current, and the difference between any two phase currents does not exceed 5% of the rated current. The single-phase significant change period is defined as the time when the current of one phase changes by more than 10% of the rated current within 5 seconds, while the changes in the currents of the other two phases do not exceed 2% of the rated current. The steady-state maintenance period is defined as the time when the rate of change of the current of each phase is lower than 2% of the rated current within 10 consecutive analysis windows. The above division is based on the fact that the light load balancing period is conducive to extracting low background crosstalk characteristics. The single-phase significant change period is helpful in identifying the amplitude and delay relationship of the target phase load change to the adjacent phase temperature rise, while the steady-state maintenance period is helpful in correcting long-term bias. During online updates, the basic influence value is first determined according to the joint arrangement and phase distance, then the difference in spatial heat transfer direction is corrected according to the relative orientation, and finally the basic influence value is increased or decreased according to the phase temperature difference and load imbalance in the current analysis window, thus forming the phase influence matrix in the current analysis window. The delay window between the target phase load change and the adjacent environment corrected temperature rise change is taken as 1 to 4 temperature sampling windows, corresponding to 0.5 seconds to 2 seconds. The above values ​​are based on the fact that the interphase air heat transfer and radiation coupling inside the branch box are usually faster than the heat transfer inside the insulation layer, but will still show short-time delay characteristics. The amplitude ratio is determined according to the response ratio of the adjacent phase environment corrected temperature rise to the target phase load change under typical operating conditions, and is controlled between 0 and 1 to ensure that the interphase influence coefficient has clear physical meaning and interpretability. After obtaining the interphase influence matrix, the thermal influence components from the other two phases are estimated for each phase. The net surface temperature rise of the target phase is the result of subtracting the thermal influence components from the adjacent two phases from the target phase's environmentally corrected temperature rise. To prevent over-separation, non-negativity constraints and continuity constraints are set simultaneously within each analysis window. The non-negativity constraint is: when the net surface temperature rise is less than 0, it is corrected to 0. This is based on the fact that the net surface temperature rise is the temperature rise relative to the local environmental background temperature, and should not have a significantly negative value for a long period under normal operating conditions. The continuity constraint is: when the net surface temperature rise of the target phase is less than 0 in two adjacent analysis windows... If the temperature drop exceeds 0.5℃ and the corresponding phase current does not decrease, and the local ambient background temperature does not decrease synchronously, it is determined that there may be over-separation in the current analysis window. When the direction of net surface temperature rise change is continuously opposite to the direction of corresponding phase current change in three consecutive analysis windows, the enhanced separation results are paused and the results are reverted to the basic separation results. The above threshold values ​​are based on the fact that, under a 0.5-second sampling period, the net surface temperature rise usually does not show a significant sudden drop without load reduction support. However, if the direction of change is opposite in multiple consecutive analysis windows, it usually indicates that the interphase influence coefficient is too large or there is a deviation in the environmental mapping. To ensure consistency between the separation process and the preceding data acquisition and verification, an abnormal branching, freezing, and recovery mechanism are implemented. If the environmental acquisition status is determined to be abnormal during the preceding data acquisition and verification, or if a door opening signal, maintenance operation signal, or external wind disturbance signal is received, the interphase influence coefficient update is paused, and only the local ambient background temperature is deducted; enhanced separation is not performed. Enhanced separation is restarted after the environment stabilizes. Environmental stabilization is defined as a cumulative temperature change of less than 0.2℃ within 10 consecutive temperature sampling points, and this condition must persist for at least one analysis window after the door status returns to normal. This value is based on the premise that this condition can cover the short-term buffering process after the external airflow disturbance subsides. If the surface temperature channel of one phase in the three phases has been determined to be invalid during the preceding data acquisition and verification, the interphase influence coefficient update is paused. The impact matrix is ​​only calculated among the remaining valid channels, while the impact coefficient of the previous stable period is retained for the failed channel, and the net surface temperature rise of that phase is marked as a low confidence result. The low confidence result is only used for trend reference and anomaly interpretation in subsequent processing, and is not used for parameter updates and high-level anomaly determination. If the environmental mapping is incomplete, the typical operating conditions required for the update of the impact coefficient are not met in three consecutive analysis windows, or more than three windows in five consecutive analysis windows trigger continuity constraints, the current impact coefficient is frozen, and the most recent stable coefficient remains unchanged until the update conditions are met again before dynamic updates are resumed. The basis for the above freezing conditions is that frequent updates will amplify the instantaneous error under abnormal operating conditions, while keeping the most recent stable coefficient when conditions are insufficient is conducive to ensuring the continuity and interpretability of the separation results. After the above processing, at the end of each analysis window, the net surface temperature rise sequence corresponding to the A-phase joint, B-phase joint, and C-phase joint is generated, and information such as the local ambient background temperature component, interphase heat effect component, and interphase influence coefficient status are recorded simultaneously. The net surface temperature rise sequence serves as the direct input for subsequent heat transfer hysteresis compensation, while the local ambient background temperature component, interphase heat effect component, and interphase influence coefficient status serve as auxiliary basis for subsequent parameter updates, anomaly interpretation, and confidence judgment.

[0021] Specifically, such as Figure 4 As shown: After separating the effects of ambient temperature and interphase thermal interference to form the net surface temperature rise sequence of each phase joint, hysteresis compensation is performed on the net surface temperature rise by combining the load change process and the heat transfer hysteresis relationship of the joint insulation layer. Hysteresis compensation is used to correct the time lag between the external surface temperature rise of the joint and the actual internal heating under alternating load conditions, rather than simply amplifying the net surface temperature rise. After continuously obtaining the effective net surface temperature rise sequence and the corresponding effective load current sequence, the edge processing terminal performs window analysis on the load current of each phase and identifies the current operating status accordingly. The analysis window width is set to 0.5 seconds, which is based on the fact that this value is consistent with the time scale used in the previous thermal interference separation process, which can reflect the short-term fluctuations under alternating load without being affected by excessive window width. Short and amplified random noise; the edge processing terminal divides the operating state into a steady-state segment, a load-increasing segment, a load-reducing segment, and a fluctuation segment. The steady-state segment is defined as the period in which the current change rate is less than 2% of the rated current and the net surface temperature rise amplitude is less than 0.2℃ for 10 consecutive analysis windows. The load-increasing segment is defined as the period in which the current continuously increases for 3 consecutive analysis windows and the cumulative increase exceeds 5% of the rated current. The load-reducing segment is defined as the period in which the current continuously decreases for 3 consecutive analysis windows and the cumulative decrease exceeds 5% of the rated current. The fluctuation segment is defined as the period in which there are at least two reversals in the direction of current change, the peak-to-valley difference exceeds 5% of the rated current, and the deviation of the average current of the window from the average current of the previous steady-state segment does not exceed 10% of the rated current for 10 consecutive analysis windows. The above values ​​are based on the following: when the rate of change of current is less than 2% of the rated current, the operating state can generally be considered to be basically stable; when the cumulative increase or decrease exceeds 5% of the rated current, it is usually sufficient to cause a considerable change in the internal heating level of the joint; multiple reversals of direction under the condition that the load average has not shifted significantly can characterize the pulsating operating state under alternating load; when the current phase enters the load increase section, load decrease section or fluctuation section, hysteresis compensation is activated; in addition, when the direction of change of net surface temperature rise is inconsistent with the direction of change of load current of the corresponding phase for three consecutive analysis windows, hysteresis compensation is also activated; the basis for using three consecutive analysis windows as the judgment condition is that the directional deviation in a single analysis window may be caused by measurement disturbance or short-term fluctuation, while the directional mismatch in multiple consecutive analysis windows usually indicates that the external surface response has been significantly affected by the heat transfer hysteresis of the insulation layer, so further correction of net surface temperature rise is required; The set of joint heat transfer hysteresis parameters called for hysteresis compensation includes at least the following fields: joint model, voltage level, conductor cross-section, insulation layer thickness, insulation material type, first heat transfer time constant, second heat transfer time constant, third heat transfer time constant, hysteresis gain, attenuation coefficient, reference steady-state thermal resistance, and parameter version number. These parameters are primarily determined based on the step load temperature rise test results of joints of the same model. When corresponding step load temperature rise test results are lacking, an initial heat conduction calibration model is established based on the joint structural dimensions, insulation material parameters, and standard part specifications to obtain the initial values ​​of each parameter. Calibration data is used to make minor corrections to the initial values. Corresponding parameter templates are established for different joint models. When a dedicated parameter template is lacking, the most compatible general parameter template is called as the initial template according to the voltage level, conductor cross-section range, insulation layer thickness range, and insulation material type. The basis for adopting the above matching method is that the joint heat transfer hysteresis characteristics are simultaneously affected by conductor dimensions, insulation layer thickness, and material thermal properties. Matching only according to a single model is insufficient to cover the actual structural and material differences of similar joints in the field. The edge processing terminal determines the hysteresis compensation amount based on the data in the current analysis window. The hysteresis relationship model can be selected as a layered thermal inertia model, and a multi-time response model can be used as an alternative implementation. When using the layered thermal inertia model, the thermal response is divided into three layers along the radial heat transfer path of the joint: the conductor heating layer, the insulation transition layer, and the surface response layer. Each layer corresponds to a set of heat transfer time constants and weight parameters. The model input includes at least the current value, current change rate, net surface temperature rise value, net surface temperature rise change rate, and compensation status of the previous analysis window. During processing, the load driving quantity is first formed based on the current current value and current change rate. Then, the degree of thermal response hysteresis of the conductor heating layer, insulation transition layer, and surface response layer is estimated by combining the first heat transfer time constant, the second heat transfer time constant, and the third heat transfer time constant. Finally, the hysteresis compensation amount of the current analysis window is determined by combining the hysteresis gain and attenuation coefficient. For common electrical connectors with insulation layer thicknesses of 30 mm to 60 mm, the first heat transfer time constant is taken as 5 to 30 seconds, the second as 30 to 180 seconds, and the third as 120 to 600 seconds. These values ​​are based on the fact that the conductor heating layer, being closest to the heat source, has the fastest thermal response; the insulating transition layer undertakes the main heat transfer process, with the next fastest thermal response; the surface response layer directly corresponds to the temperature sensing results of the outer surface, exhibiting the most significant hysteresis, thus corresponding to a progressively increasing time constant. When the insulation layer thickness increases or the thermal conductivity of the insulating material decreases, the heat transfer in each layer... The time constant increases accordingly; when the insulation layer thickness decreases or the thermal conductivity of the insulation material increases, the heat transfer time constant of each layer decreases accordingly; the hysteresis gain is between 0.1 and 1.0, and its value is based on the fact that this parameter is used to control the response intensity of the compensation amount to load changes. When it is below 0.1, the compensation effect is not obvious, and when it is above 1.0, it is easy to amplify short-term disturbances; the attenuation coefficient is between 0.90 and 0.99, and its value is based on the fact that this parameter is used to control the attenuation rate of residual heat during the load reduction stage. When it is below 0.90, the compensation attenuation is too fast, and when it is above 0.99, the residual compensation stays for too long; Different hysteresis compensation methods are used for different operating states. In the steady-state phase, the internal heat generation and external surface temperature rise response of the joint are close to thermal equilibrium. The hysteresis compensation amount is mainly used to correct long-term bias and can be determined based on the deviation between the current net surface temperature rise rate and the reference steady-state thermal resistance, with the compensation range controlled within 5% of the current net surface temperature rise. This value is based on the fact that the heat transfer hysteresis effect is relatively weak during steady-state operation; if the compensation range is too large, it can easily introduce additional amplification to the already nearly balanced surface temperature rise. In the load-increasing phase, the internal heat generation of the joint usually precedes the external surface temperature rise. Therefore, the hysteresis compensation amount is determined according to the positive compensation method. The compensation range is determined by combining the current current change rate, the cumulative load increase, and the first and second heat transfer time constants. The faster the current rises and the larger the cumulative load increase, the larger the compensation amount. In the load-reducing phase, the release of residual heat on the external surface is usually slower than the decay of internal heat generation. Therefore, the compensation amount is determined by the positive compensation method. The compensation amount is based on the analysis window, multiplied by the attenuation coefficient, and then corrected by combining the current current decrease. During the fluctuation phase, to avoid overcompensation caused by a single pulsation, the short-window trend and long-window trend are weighted and combined to determine the compensation amount. The short window can be selected from the three most recent analysis windows, and the long window from the ten most recent analysis windows. The short-window weight can be 0.6 to 0.8, and the long-window weight can be 0.2 to 0.4, with commonly used values ​​of 0.7 and 0.3. These values ​​are based on the fact that short-term changes within the fluctuation phase are more sensitive to thermal response judgment, but relying solely on short-window information makes it susceptible to single pulsation disturbances. Combining the long-window trend helps improve the stability of the compensation result. The corrected surface temperature rise is formed by the net surface temperature rise and the hysteresis compensation amount determined according to the current operating state. For the load increase phase, positive compensation is used; for the load decrease phase, residual compensation after attenuation is used; and for the steady-state and fluctuation phases, the correction result is determined according to the corresponding compensation rules. To balance dynamic response capability and operational stability with hysteresis compensation, the edge processing terminal is equipped with a parameter update mechanism. Parameter updates are performed within the aforementioned steady-state range, prioritizing the light-load balance interval. The light-load balance interval can be selected as a period where the current of each phase is lower than 30% of the rated current, and the rate of change of current is lower than 2% of the rated current for 10 consecutive analysis windows. The basis for these values ​​is that the load drive is relatively stable within this interval, and the correspondence between surface temperature rise and current is more suitable for reverse verification of parameter drift. During the update process, the average absolute deviation between the model-predicted surface temperature rise and the current net surface temperature rise is used as the deviation index. When the deviation exceeds the average absolute deviation of three consecutive steady-state ranges... When the indicators increase progressively and the deviation of the current steady-state segment exceeds 10% compared to the deviation of the previous steady-state segment, a parameter update is triggered. During the parameter update, the first heat transfer time constant, the second heat transfer time constant, the third heat transfer time constant, the hysteresis gain, and the attenuation coefficient are corrected in small steps, with each adjustment ranging from 1% to 5%. The reason for this is that an adjustment of less than 1% does not significantly improve the model output, while an adjustment of more than 5% is likely to misjudge short-term anomalies as parameter drift and amplify their impact, thereby causing model oscillation or instability. After the parameter update is completed, the new parameter version number is recorded, and the previous parameter version is retained for subsequent rollback calls. To ensure the continuity and reliability of the hysteresis compensation process, the edge processing terminal also has an anomaly handling mechanism. When current data is briefly missing while the net surface temperature rise sequence remains continuous, the compensation amount is calculated again using the most recent valid current value, provided the missing duration does not exceed 5 current sampling points. Conservative compensation is then performed based on 20% to 50% of the compensation amount calculated by the model under the current operating state, with 30% being a commonly used value. The basis for this value is that a short-term current loss of less than 0.5 seconds usually does not completely disrupt the continuity of the thermal response, but in the case of incomplete input data, the compensation intensity should be appropriately reduced to avoid over-correction. When an abnormal jump of more than 0.5℃ occurs between two adjacent analysis windows in the net surface temperature rise, and cannot be explained by the status identifier formed by the previous thermal interference separation process, the compensation for the current analysis window is paused, and the previous parameter version and the previous window compensation status are called for recalculation. This 0.5℃ threshold is consistent with the short-term mutation judgment threshold used in the previous acquisition verification process and thermal interference separation process. The basis for this is to maintain the uniformity of the threshold system throughout the entire processing chain and avoid compensation deviations caused by inconsistent judgment standards. When a dedicated parameter template is missing, the aforementioned general parameter template is called back for initial compensation, and the error is gradually corrected during subsequent operation. When three consecutive analysis windows are in a low-confidence input state, parameter updates are frozen, and only the current compensation state output is retained until the input data returns to normal. After the above processing, a calibration surface temperature rise sequence corresponding to each phase joint is formed, and information such as compensation state identifier, parameter version information, and confidence level identifier corresponding to the calibration surface temperature rise sequence is generated simultaneously. The relevant information includes at least the phase identifier, analysis window number, compensation state category, parameter version number, and confidence level. The calibration surface temperature rise sequence serves as the input data for subsequent internal temperature rise determination, and the compensation state identifier, parameter version information, and confidence level identifier serve as auxiliary basis for subsequent internal temperature rise interpretation, parameter backtracking, and anomaly detection.

[0022] Specifically, such as Figure 5 As shown: After the pre-processing hysteresis compensation generates the temperature rise sequence of the corrected surface of each phase joint, the internal temperature rise of each phase joint is determined based on the heat transfer relationship of the joint structure, and the internal temperature rise is correlated with the load change process. The determination of the internal temperature rise is initiated after obtaining a complete judgment period of the corrected surface temperature rise sequence. The analysis window width is set to 0.5 seconds, and a complete judgment period can be selected as 10 consecutive analysis windows, corresponding to a continuous observation duration of 5 seconds. The number of reliable windows is set to no less than 8. The above values ​​are based on the fact that the 5-second duration can cover the short-term evolution process of the temperature rise of the corrected surface to the internal thermal state, while not significantly reducing the response speed of subsequent anomaly identification. Setting the number of reliable windows to no less than 8 is equivalent to the effective data ratio being no less than 80%, which allows for a small amount of single-window disturbance, while avoiding distortion of the internal temperature rise back-inference due to insufficient effective data. The edge processing terminal calls the structural heat transfer template matched with the target joint. When the dedicated template cannot be obtained, the alternative template closest to the target joint is called, and the template substitution status is recorded. The structural heat transfer template is primarily established based on the measured calibration results of the same type of joint. When measured calibration results are lacking, an initial template can be established based on the joint structural dimensions, thermal properties of the insulation material, and simulation calibration results. The measured data of the standard sample is used to correct the initial template. The alternative template is matched according to the voltage level, conductor cross-sectional area, insulation layer thickness range, and insulation material type. The relevant information in the structural heat transfer template includes at least the layered thermal resistance between the crimped area, conductor area, semiconductive layer, insulation layer, and outer surface, the interlayer temperature difference mapping coefficient, the equivalent thermal resistance from the conductor to the surface, the thickness of each layer, the material thermal conductivity range, and the steady-state response range under typical load. The above template parameters are set based on the fact that the internal temperature rise of the joint is not only affected by the surface temperature rise, but also directly related to the spatial thermal resistance distribution from the internal heat-generating part to the outer surface and the thermal conductivity of the material. Only by simultaneously retaining the layered thermal resistance, layer thickness, and interlayer mapping information can the layer-by-layer reverse calculation from the outer surface to the internal heat-generating area be achieved. The internal temperature rise is not directly output as an absolute internal temperature. Instead, it is first determined relative to the ambient background. The edge processing terminal uses the corrected surface temperature rise sequence as a basis, combined with the interlayer temperature difference mapping relationship in the structural heat transfer template, to reverse-engineer the internal thermal state within the current analysis window. The reverse-engineering adopts a layered progressive approach, first determining the temperature difference increment from the outer surface to the middle position of the insulation layer, then determining the temperature difference increment from the middle position of the insulation layer to the pressing area, and finally obtaining the internal core region temperature rise. When the structural heat transfer template contains more layers, the temperature difference increment is accumulated layer by layer according to the layering order given by the template. The internal core region temperature rise is obtained by summing the corrected surface temperature rise and the temperature difference increment of each layer. The temperature difference increment of each layer is determined by looking up a table or interpolating from the structural heat transfer template according to the current load range and the current correction surface temperature rise range. To balance the dynamic sensitivity and result stability of the internal temperature rise back calculation, the edge processing terminal calculates both short-window and long-window back calculation results simultaneously. The short window value can be selected from the most recent 3 to 5 analysis windows, and the long window value can be selected from the most recent 10 to 20 analysis windows. Commonly used values ​​are 3 analysis windows for the short window and 10 analysis windows for the long window, with corresponding durations of 1.5 seconds and 5 seconds, respectively. The above values ​​are based on the fact that the short window can quickly track the rapid changes in the internal thermal state of the joint, while the long window can suppress single disturbances and local compensation errors. When the short-window and long-window results change in the same direction, and the difference between them within the current analysis window does not exceed 0.5℃, or does not exceed 10% of the long-window result, the two are combined into an internal temperature rise result using a weighted average. The short-window weight can be selected from 0.6 to 0.8, and the long-window weight can be selected from 0.2 to 0.4, with commonly used values ​​being 0.7 and 0.3. The basis for these values ​​is that the short window should maintain a higher weight to preserve sensitivity to rapid changes, while the long window should have a lower weight to smooth out occasional disturbances. When the short-window result and the long-window result change in the same direction, the difference between the two within the current analysis window does not exceed 0.5℃, or does not exceed 10% of the long-window result. When the difference between window results exceeds the aforementioned threshold, the long window result is used as the current main internal temperature rise value, while the short window result is only used as a trend indicator, and the current analysis window is marked as a window to be reviewed. When the above deviation exists in three consecutive analysis windows, the current template update is paused, the corresponding result is not written into the high-confidence internal temperature rise valid sequence, only the internal temperature rise trend of the previous valid window is used, and the review status is recorded. The basis for the above processing is that the long window result is not sensitive to occasional disturbances and is more suitable as the current main value, while the short window result is more suitable as a trend judgment basis. After obtaining the internal temperature rise results, a correlation analysis is performed between the internal temperature rise and the load change process to distinguish between the increase in internal temperature rise caused by normal load increase and the abnormal internal temperature rise caused by abnormal contact conditions. During the correlation analysis, the internal temperature rise change curve and the load current change curve are compared simultaneously within the same analysis window. The internal temperature rise increment, internal temperature rise change rate, and duration are extracted, along with the corresponding load increment, load change rate, and duration, and a thermal response correspondence is established accordingly. The historical characteristic interval is established based on the statistical results of internal temperature rise of the same joint under healthy operating conditions, in similar load intervals, and in similar thermal background intervals. The load interval can be divided into levels based on 10% of the rated current, and the thermal background interval can be divided into levels based on 2°C. Within each load interval and thermal background interval, the historical characteristic interval can be selected. The range can be formed by adding or subtracting two standard deviations from the historical sample mean, or it can be formed by the 5% to 95th percentile range. The above values ​​are based on the fact that grading by 10% of the rated current can take into account both load resolution and sample size, and grading by 2℃ can distinguish the impact of environmental background differences on internal temperature rise. If the deviation between the increase in internal temperature rise and the corresponding load increment does not exceed 20% of the historical average of the same type, the deviation between the rate of change of internal temperature rise and the rate of change of corresponding load does not exceed 20% of the historical average of the same type, and the delay of internal temperature rise relative to load change is between 0 and 4 analysis windows, then it is judged as load-driven temperature rise. The above values ​​are based on the fact that although the internal temperature rise response of healthy connectors under similar operating conditions has a limited delay, its amplitude deviation and rate deviation usually do not exceed 20% of the historical average for a long period of time. If the change in the corresponding phase load current within three consecutive analysis windows is less than 2% of the rated current, or has fallen by more than 5% of the rated current relative to the most recent load peak, but the internal temperature rise continues to rise for three consecutive analysis windows with a cumulative increase of more than 0.5℃, then the tendency for abnormal heating is considered to be enhanced. If the internal temperature rise exceeds the upper limit of the historical characteristic range under similar load conditions by more than 10%, the tendency for abnormal heating is also considered to be enhanced. The basis for the above thresholds is that when the load remains basically unchanged or has fallen significantly, the internal temperature rise of a healthy connector should generally not continue to rise significantly. Exceeding the upper limit of the historical characteristic range by more than 10% usually indicates that the current thermal response has deviated from the healthy range. If the correlation analysis finds that the internal temperature rise and load change are continuously mismatched for three consecutive complete judgment cycles and cannot be explained by abnormal acquisition, template replacement status, or low confidence input status, then the status of that phase connector is upgraded to a key concern status and is given priority for subsequent anomaly judgment. The continuous mismatch threshold is set at three consecutive complete judgment cycles, which is based on the ability to exclude short-term deviations caused by environmental fluctuations, temporary load switching, or template replacement within a single judgment cycle. To ensure the continuous use of internal temperature rise results, the edge processing terminal is also equipped with an anomaly handling and rollback mechanism. When the number of reliable windows for the corrected surface temperature rise sequence within the current judgment period does not meet the aforementioned requirement, the effective internal temperature rise sequence is not directly updated. Instead, the internal temperature rise trend of the previous effective window is used, and the current analysis window is marked as an interpolation window. If the aforementioned template substitution state exists, the internal temperature rise results continue to be output, but the reliability level is lowered by one level. If the key parameters of the template are missing, the internal temperature rise trend continues to be output, but the reliability level is lowered by two levels, and the current result is not used for template updates. The reliability level can be selected as high reliability, medium reliability, and low reliability. High reliability is defined as when a dedicated template is called and the input is complete, medium reliability is defined as when the template is substituted, and low reliability is defined as when key parameters are missing or the effective window is insufficient. The basis for the above classification method is that the reliability of the results corresponding to different template sources and input completeness varies. Using classification processing is beneficial for distinguishing and using the internal temperature rise results in the future. If the difference between the short-window result and the long-window result for three consecutive analysis windows exceeds the aforementioned consistency threshold, and the correlation analysis continues to mismatch and cannot be explained by the previous data collection verification status, then the writing of the current judgment period result into the internal temperature rise effective sequence is paused, only the most recent stable result is retained, and the status pending verification is recorded. After the above processing, the internal temperature rise effective sequence, internal temperature rise change rate sequence, and correlation status information between internal temperature rise and load change are formed for each phase joint, and the template call status and confidence level are recorded simultaneously. The relevant information includes at least the following fields: phase identifier, analysis window number, internal temperature rise value, internal temperature rise change rate, correlation status category, template call status, and confidence level. The above internal temperature rise effective sequence serves as the main input for subsequent anomaly discrimination, the correlation status information is used to distinguish between normal thermal response and abnormal thermal response, and the template call status and confidence level are used for subsequent anomaly interpretation, result backtracking, and priority ranking. The value of the above three consecutive analysis windows is based on the fact that deviations within a single or a few analysis windows may be caused by instantaneous disturbances, while continuous deviations in three consecutive analysis windows usually reflect that the current back-calculation result has deviated from the normal stable state.

[0023] Specifically, such as Figure 6 As shown: After the preceding internal temperature rise determination process generates the effective sequence of internal temperature rise, the internal temperature rise rate sequence, and the correlation status information between internal temperature rise and load change for each phase joint, the edge processing terminal performs anomaly detection based on the change status of internal temperature rise in each phase joint and outputs the detection results. Anomaly detection is initiated when the following conditions are met in the current determination cycle: the proportion of effective internal temperature rise windows is not less than 80%, the number of reliable windows is not less than 8, and the number of continuous interpolation windows does not exceed 2; the width of the determination window is consistent with the preceding internal temperature rise determination process and can be selected as 0.5 seconds; one determination cycle can be selected as 10 consecutive determination windows, corresponding to a continuous observation duration of 5 seconds; the above values ​​are based on the fact that a duration of 5 seconds can cover the short-term evolution of internal temperature rise and its main response relationship with load change, while not significantly reducing the real-time performance of anomaly identification; setting the effective window proportion to not less than 80% and limiting the number of continuous interpolation windows to no more than 2 can reduce the probability of misjudgment due to insufficient data continuity while allowing a small amount of single-window disturbance. The input for discrimination includes at least the effective sequence of internal temperature rise of each phase joint, the sequence of internal temperature rise change rate, the correlation status information between internal temperature rise and load change, the status information of previous data acquisition and verification, the status information of thermal interference separation, the status information of hysteresis compensation, the status of template call, the credibility level, and the key intermediate quantity change identifier. Among them, the key intermediate quantities include at least the fields of net surface temperature rise, corrected surface temperature rise, internal temperature rise back-calculation result, interphase influence coefficient, and hysteresis parameter. Anomaly discrimination does not rely solely on a single absolute temperature value, but comprehensively considers the absolute value of internal temperature rise, the rate of internal temperature rise, the duration of high internal temperature rise, whether there is a mismatch between internal temperature rise and load change, and whether the state deteriorates continuously in adjacent time periods, thereby identifying a continuous abnormal temperature rise process that deviates from the normal thermal response law. The basis of the above discrimination method is that under high ambient temperature or short-term high load conditions in summer, the absolute value of internal temperature rise may increase, but it does not necessarily correspond to joint anomaly. Therefore, it is necessary to combine the temperature rise change trend, duration, and load correlation for comprehensive judgment. The edge processing terminal establishes a hierarchical threshold system based on connector operating history, equipment type, current load range, and current thermal background range. Load ranges can be divided into levels based on 10% of rated current, and thermal background ranges can be divided into levels based on 2°C. Within each load range and thermal background range, historical baselines for internal temperature rise and internal temperature rise rate are established based on healthy operating samples. The hierarchical threshold system includes at least a monitoring threshold, a warning threshold, and an alarm threshold. Thresholds are preferentially constructed using the historical sample mean plus the standard deviation. Specifically, the monitoring threshold can be the historical sample mean plus one standard deviation, the warning threshold can be the mean plus two standard deviations, and the alarm threshold can be the mean plus three standard deviations. When the target range sample distribution is significantly skewed, has many outliers, or the sample size is insufficient to stably calculate the standard deviation, the threshold is constructed using quantiles. The threshold can be taken as the 90th percentile of historical samples, the warning threshold can be taken as the 95th percentile, and the alarm threshold can be taken as the 99th percentile. The above-mentioned threshold construction method, which combines priority and substitution, is based on the fact that doubling the standard deviation of the mean is more suitable for scenarios with relatively stable sample distribution, while the quantile method is more suitable for scenarios with skewed sample distribution or more outliers, thereby improving the applicability of the threshold system under different data conditions. When the number of healthy samples in the target load interval and the thermal background interval is less than the preset lower limit, samples from adjacent load intervals or adjacent thermal background intervals can be merged to establish a temporary threshold. If the number of samples after merging is still insufficient, the equipment factory baseline threshold or the baseline threshold of the same type of healthy connector is used as the fallback threshold. The historical mean and historical health slope mean mentioned below refer to the statistical results of the healthy samples corresponding to the current load interval and the current thermal background interval. The internal temperature rise rate is obtained by dividing the difference in internal temperature rise between two adjacent judgment windows by the window duration. When the rise rate exceeds 20% of the historical average health slope under similar operating conditions for three consecutive judgment windows, it is judged as an excessive rise rate. The relative deviation here is calculated by dividing the difference between the current value and the historical average of the same type by the historical average of the same type. The duration of the internal temperature rise remaining high is determined according to the duration of the internal temperature rise continuously above the attention threshold, warning threshold, or alarm threshold. When the judgment window width is 0.5 seconds, the continuous high threshold can be constructed according to 10, 20, and 40 consecutive judgment windows corresponding to 5 seconds, 10 seconds, and 20 seconds, respectively. The above values ​​are based on the fact that the over-limit phenomenon in a single or a few judgment windows may be caused by instantaneous disturbances, while the high-level state lasting 5 seconds, 10 seconds, or 20 seconds can better reflect the continuity and stability of the abnormal internal thermal state of the joint, thus helping to distinguish between instantaneous over-limit and continuous abnormality. Unless otherwise specified, the following trend-based criteria use three consecutive judgment windows as the threshold for sustained judgment. Specifically, when the absolute value of the internal temperature rise exceeds the attention threshold for no more than three consecutive judgment windows, and the rate of increase in the internal temperature rise does not exceed 20% of the average historical healthy slope, and the relative deviation between the internal temperature rise and the load increase does not exceed 20% of the average historical value for the same type, it is judged as a state of concern. The above values ​​are based on the fact that short-term, minor exceedances that can be explained by load changes are more suitable for early attention than for direct escalation. When the internal temperature rise exceeds the warning threshold for three consecutive judgment windows, or although the absolute value of the internal temperature rise has not reached the warning threshold, its rate of increase exceeds 30% of the average historical healthy slope for three consecutive judgment windows, and the relative deviation between the internal temperature rise and the load change exceeds 20% of the average historical value for three consecutive judgment windows... When the internal temperature rise exceeds 20% of the historical average for similar conditions, it is considered a warning state. The above criteria reflect that the internal temperature rise has continuously exceeded the warning boundary, or the temperature rise trend has significantly deviated from the normal response pattern under healthy operating conditions. When the internal temperature rise reaches the alarm threshold, or the corresponding phase load has fallen by more than 5% of the rated current relative to the most recent peak, but the internal temperature rise still rises for three consecutive judgment windows, and the cumulative increase exceeds 0.5℃, it is considered an alarm state. The above values ​​are based on the fact that when the load has obviously fallen, the internal temperature rise of the healthy connector should generally not continue to rise significantly, and the cumulative increase of more than 0.5℃ for three consecutive judgment windows usually indicates that the abnormal heating has become persistent. If the internal temperature rise exceeds the upper limit of the historical characteristic range under similar load conditions by more than 10%, it can also be directly judged as an increased tendency for abnormal heating. For cases where multiple phase joints simultaneously experience temperature increases, a group-based assessment is necessary. When the internal temperature rise of the three phases is in the same direction, the temperature difference between any two phases does not exceed 1.0℃, and the local ambient background temperature rises synchronously by more than 0.5℃, it is preferentially determined to be an overall ambient temperature rise. When the internal temperature of one phase exceeds the average of the other two phases by more than 1.5℃, and this deviation exists for three consecutive assessment windows, it is preferentially determined to be a target phase joint anomaly. The basis for these values ​​is that an overall ambient temperature rise usually manifests as a synchronous increase in all three phases with small inter-phase differences, while a single-phase joint anomaly usually manifests as... If a single phase continuously deviates from the other two phases, and correlation analysis reveals a continuous mismatch between internal temperature rise and load change for three consecutive complete judgment cycles, and this mismatch cannot be explained by abnormal data acquisition, template replacement status, or low-confidence input status, then the phase connection status is elevated to a key concern status and prioritized for subsequent anomaly detection. The threshold for continuous mismatch is set at three consecutive complete judgment cycles, based on the fact that this threshold can exclude short-term deviations caused by environmental fluctuations, temporary load switching, or template replacement within a single judgment cycle, thereby improving the stability and reliability of anomaly detection results. To improve the robustness of the judgment results, the edge processing terminal is equipped with a verification mechanism and a recalculation mechanism. The verification mechanism is applicable to critical states, i.e., when the relative deviation between the current internal temperature rise value and the corresponding level threshold does not exceed 5% of the level threshold, and any of the following states exists simultaneously: template substitution state, low-confidence input state, interpolation state, or low-confidence state of previous processing, the judgment window is extended by 3 to 5 months for continued observation, and the previous net surface temperature rise sequence, corrected surface temperature rise sequence, and internal temperature rise back-calculation results can be called back for re-verification. The verification extension period is set to 3 to 5 judgment windows, based on the fact that small deviations near the threshold boundary are easily affected by single disturbances, and this length can better mitigate the impact of small deviations. It effectively distinguishes between instantaneous fluctuations and persistent anomalies; the recalculation mechanism applies to situations where previous key intermediate quantities have been corrected; when any key intermediate quantity in the current judgment period changes by more than 10% compared to the previous version due to data acquisition anomaly repair, phase-to-phase influence coefficient update, hysteresis parameter update, or structural heat transfer template update, the previous thermal interference separation processing, hysteresis compensation processing, and internal temperature rise determination processing are re-executed for the most recent 1 to 3 historical judgment periods, and the judgment is re-completed based on the recalculation results; the recalculation trigger threshold is set at 10%, based on the fact that when the change in key intermediate quantities reaches this level, it is usually sufficient to affect the final judgment conclusion, and the influence of the old results needs to be eliminated through local backcalculation; To ensure the controllability of anomaly detection even when the data link is unstable, the edge processing terminal also has a rollback mechanism. This rollback mechanism is applicable when the current data link is severely abnormal or the key template is unavailable. When a valid internal temperature rise sequence that meets the aforementioned integrity requirements cannot be formed within three consecutive judgment cycles, or when the key template is missing and a replacement template cannot be called, the output of a highly reliable anomaly conclusion is stopped, and only a data anomaly prompt and a status pending manual review are output. The rollback threshold is set at three consecutive judgment cycles because data anomalies within a single judgment cycle may be caused by temporary communication failures or short-term disturbances, while continuous anomalies within three consecutive judgment cycles usually indicate that the current automatic detection conditions are insufficient. By setting up review, recalculation, and rollback mechanisms, the terminal can adopt different processing methods—continued observation, partial backcalculation, and conclusion downgrading—in critical states, previous result correction states, and data link anomaly states, respectively, thereby ensuring the continuity, reliability, and traceability of the detection results. After the above processing, anomaly identification results and corresponding auxiliary status information for each phase joint are generated. The relevant information includes at least the joint identifier, phase identifier, identification time, anomaly level, internal temperature rise value, internal temperature rise rate of change, change status category, load association status, template call status, and data reliability. The change status category can be selected as heating, stable, cooling, rapid heating, and continuous abnormal heating, among which rapid heating and continuous abnormal heating can be used as auxiliary identification for early warning or alarm status. The identification results can be written to local records or uploaded to the upper-level monitoring platform. When in a state of concern, the concern result is output and continuous tracking is maintained. When in an early warning state, the early warning result is output and a review prompt message is generated. When in an alarm state, the alarm result is output and a priority handling identifier is generated. By outputting different anomaly levels in a tiered manner, corresponding basis can be provided for subsequent operation and maintenance decisions.

[0024] Example 2: Based on Example 1, the specific application process of a method for monitoring the temperature rise of cable branch box joints is further explained: Taking a 10kV cable branch box as the monitoring object, the branch box is equipped with three-phase joints that are respectively connected to the A-phase conductor, B-phase conductor, and C-phase conductor. Non-intrusive temperature sensors are arranged on the surface of the outer insulation layer of each phase joint. Ambient temperature acquisition points are arranged in the upper, middle and lower parts of the box. Current acquisition devices are arranged at the inlet and outlet positions. The edge processing terminal pre-stores the spatial arrangement relationship of the three-phase joints, the structural heat transfer template, the hysteresis compensation parameter template and the anomaly discrimination threshold system. During operation, the edge processing terminal continuously receives the surface temperature of each phase joint, the ambient temperature inside the box and the load current of each phase, and sequentially completes data acquisition verification, separation of environmental influence and interphase thermal interference, hysteresis compensation, internal temperature rise back calculation and anomaly discrimination processing to realize continuous monitoring of the abnormal heating state inside the joints. In this operational scenario, the branch box is put into operation during the high-load period in summer. The initial load currents of phases A, B, and C are stable at around 182A, 185A, and 179A, respectively. The ambient temperature inside the box is 36.8℃ in the upper part, 35.9℃ in the middle part, and 35.1℃ in the lower part. After initial sampling, the surface temperatures of each phase connector are 52.4℃ for phase A, 53.1℃ for phase B, and 52.0℃ for phase C. The edge processing terminal first adds a unified timestamp to the data of each channel and performs time alignment according to a 0.5-second analysis window. The temperature sampling period is 0.5 seconds, and the current sampling period is 0.1 seconds. The current data is aggregated according to the 0.5-second window and then entered into the same analysis window as the temperature data. After time alignment, basic validity checks are performed on various types of data. When a certain surface temperature is sampled... If a value exceeds the validity range of -40℃ to 150℃, or if a sudden jump of more than 0.5℃ occurs within a single window when both current and ambient temperature are relatively stable, the sampled value is marked as suspicious. If the ambient temperature at the upper sampling point is consistently more than 0.5℃ lower than that at the lower sampling point under medium-to-high load conditions, the environmental sampling status is marked as abnormal. If the change amplitude of two adjacent 0.1-second current sampling points exceeds 30% of the rated current and does not correspond to a switch action record or load switching record, it is determined to be an abnormal current jump. In this operating scenario, no abnormal surface temperature jump, environmental gradient anomaly, or current jump anomaly was detected in the initial stage of operation. Therefore, the data within 10 consecutive analysis windows are written into the valid monitoring sequence, forming the valid sequence of surface temperature of each phase joint, the valid sequence of ambient temperature, and the valid sequence of load current. Subsequently, the edge processing terminal performs separation processing of ambient temperature influence and interphase thermal interference; based on the internal spatial relationship information of the branch box, it reads the distances between the A-phase connector, B-phase connector, and C-phase connector and the upper, middle, and lower ambient temperature acquisition points, and calculates the local ambient background temperature of each phase connector using a distance-weighted method; taking the B-phase connector as an example, if the distances from the B-phase connector to the upper, middle, and lower ambient temperature acquisition points are 0.18m, 0.10m, and 0.22m respectively, then the corresponding weights are calculated according to 1 / 0.18, 1 / 0.10, and 1 / 0.22 and normalized, thus obtaining a local ambient background temperature of approximately 35.9℃ for the B-phase connector; A-phase and The local ambient background temperature of each phase connector can also be obtained. Then, the corresponding local ambient background temperature is subtracted from the surface temperature of each phase connector to obtain the ambient corrected temperature rise of each phase. When the current difference between any two phases does not exceed 5% of the rated current and the ambient corrected temperature rise difference between any two phases does not exceed 1.0℃, it is processed according to the basic separation path. When the current of one phase is higher than the average of the other two phases by more than 10% of the rated current, or the ambient corrected temperature rise of one phase is higher than the average of the other two phases by more than 1.5℃, it is processed according to the enhanced separation path. In this operating scenario, the three-phase load is basically balanced in the initial stage of operation, and the ambient corrected temperature rise difference is small, so it is processed according to the basic separation path first. When the system reaches a certain point, the load on phase B gradually increases due to increased power consumption in the downstream branch, rising from 185A to 206A within three consecutive analysis windows, while phases A and C remain around 183A and 181A respectively. At this time, the ambient temperature rise of phase B increases from 17.2℃ to 18.9℃, while phases A and C experience synchronous slight increases of approximately 0.3℃ and 0.2℃ respectively, and the current change amplitudes of the corresponding two phases are both less than 2% of the rated current. According to the basic separation path rule, the edge processing terminal identifies the 0.2℃ to 0.3℃ temperature rise in phases A and C that occurs synchronously with the increase in phase B load as a suspected crosstalk component, and subtracts it from the smaller value between the current synchronous slight increase amplitude and the average synchronous offset during the historical light load balancing period, thus obtaining a net surface temperature rise that is closer to the self-heating state of each phase. As the load on phase B continues to increase and its ambient temperature rise... If the temperature exceeds the average of the other two phases by more than 1.5℃, the system switches to the enhanced thermal interference state and enters the enhanced separation path. At this time, the edge processing terminal calls the pre-established interphase influence matrix, and estimates the thermal influence components of phase B on phases A and C by combining the joint arrangement, interphase distance, relative orientation, and current load imbalance. The corresponding components are then subtracted from the environmental correction temperature rise of phases A and C, ultimately forming the net surface temperature rise sequence of phases A, B, and C. Since non-negative constraints and continuity constraints are applied simultaneously in the enhanced separation path, there will be no physically meaningless net surface negative temperature rise, nor will there be an abnormal drop without load reduction support. After this processing, the net surface temperature rise of phase B remains high, while the artificially high surface temperature rise of phases A and C caused by crosstalk is reduced, thereby reducing the risk of false alarms caused by interphase thermal interference. After forming the net surface temperature rise sequence, the edge processing terminal continues to perform hysteresis compensation processing. Since the cumulative increase of the B-phase load current exceeds 5% of the rated current within three consecutive analysis windows, the system identifies that the B-phase has entered the load increase phase. At this time, the heat transfer hysteresis parameter set of the corresponding B-phase connector is called, which includes at least the first heat transfer time constant, the second heat transfer time constant, the third heat transfer time constant, hysteresis gain, attenuation coefficient, and reference steady-state thermal resistance. If the current insulation layer thickness of the B-phase connector is 40mm, the first heat transfer time constant can be selected as 10s to 20s, the second heat transfer time constant as 60s to 120s, and the third heat transfer time constant as 180s to 360s; the hysteresis gain can be selected as 0.4 to 0.7, and the attenuation coefficient as 0.94 to 0.98. The edge processing terminal calculates the current value, current change rate, net surface temperature rise value, net surface temperature rise change rate, and the previous window's data based on the current value, current change rate, net surface temperature rise value, net surface temperature rise change rate, and the previous window's data. In the compensation state, a positive hysteresis compensation amount is generated for the load increase phase. For example, during the process of phase B increasing from 185A to 206A, although the net surface temperature rise of phase B only increases from 17.2℃ to 18.9℃, since internal heat generation precedes surface temperature rise, the system can calculate a positive hysteresis compensation amount of approximately 1.6℃, thus forming a corrected surface temperature rise of approximately 20.5℃ for phase B. For phases A and C, since the load change is small, they are still in the steady state phase, so only a small correction is made. If the load of phase B subsequently drops, the system switches to the load decrease phase processing, and continues to correct according to the residual compensation value after multiplying the previous window compensation amount by the attenuation coefficient, in order to avoid misjudgment caused by the lag in the release of residual heat on the surface. If a short-term current loss occurs during operation, but the duration does not exceed 5 current sampling points, the edge processing terminal adopts the method of maintaining the most recent effective current value and performs conservative compensation at 30% of the normal compensation amount to maintain the continuity of the compensation chain. After obtaining the temperature rise sequence of each phase's corrected surface, the edge processing terminal uses the structural heat transfer template to infer the internal temperature rise of each phase joint. Taking the B-phase joint as an example, the structural heat transfer template includes at least the layered thermal resistance between the pressing area, conductor area, semiconductive layer, insulation layer, and outer surface, the interlayer temperature difference mapping coefficient, the equivalent thermal resistance from conductor to surface, the thickness of each layer, and the steady-state response range under typical load. The system first determines the temperature difference increment from the outer surface to the middle position of the insulation layer by looking up a table or interpolating the template based on the B-phase corrected surface temperature rise and load range. Then, it determines the temperature difference increment from the middle position of the insulation layer to the pressing area and accumulates it layer by layer to obtain the internal core area temperature rise. To enhance the stability of the inference results, the edge processing terminal simultaneously calculates two sets of results: short window and long window. If the short window value is the most recent 3 analysis windows, the long window value is the most recent 10 analysis windows. When the short-window and long-window results change in the same direction and the difference between them does not exceed 0.5℃ or 10% of the long-window result, the internal temperature rise result can be synthesized with weights of 0.7 and 0.3, respectively. For example, when the corrected surface temperature rise of phase B reaches 20.5℃, the internal temperature rise inferred from the short window can be 31.4℃, and the internal temperature rise inferred from the long window can be 30.9℃. The difference between the two does not exceed 0.5℃, so the synthesized result is an internal temperature rise of approximately 31.2℃. Since the load changes of phases A and C are small and the thermal interference has been removed, their internal temperature rise remains within the normal fluctuation range. If the deviation between the short-window and long-window results in a certain window exceeds the aforementioned consistency threshold for three consecutive analysis windows, the result of that period will not be written into the high-confidence internal temperature rise valid sequence. Instead, the internal temperature rise trend of the previous valid window will be used and marked as pending verification. After deducing the internal temperature rise, the edge processing terminal performs a correlation analysis between the internal temperature rise and the load change process. Based on historical samples of the same connector under healthy operating conditions, within similar load ranges, and within similar thermal background ranges, the system establishes a historical characteristic range for the internal temperature rise. If the historical average internal temperature rise of phase B in the current load range and thermal background range is 26.0℃, with a historical upper limit of approximately 28.5℃, and the current internal temperature rise of phase B has reached 31.2℃, it indicates that it has significantly exceeded the upper limit of the healthy range. Furthermore, if the load on phase B subsequently drops from 206A to 194A, with a drop exceeding 5% of the rated current, but the internal temperature rise of phase B... If the temperature continues to rise from 31.2℃ to 31.8℃, 32.1℃, and 32.4℃ within three consecutive analysis windows, with a cumulative increase exceeding 0.5℃, the system determines that phase B exhibits an increased tendency for abnormal heating. Simultaneously, if the internal temperature rises of phases A and C do not show the same magnitude or duration of deviation, and the internal temperature rise difference between the three phases exceeds 1.5℃, the system prioritizes identifying an abnormality at the phase B joint, rather than an overall environmental temperature rise. If, within the subsequent three complete judgment cycles, the internal temperature rise of phase B continues to be mismatched with load changes and cannot be explained by abnormal data acquisition, template substitution status, or low-confidence input status, then the phase B joint status is upgraded to a key concern status. After completing the correlation analysis, the edge processing terminal performs anomaly detection. The system comprehensively judges the situation based on the internal temperature rise value of phase B, the rate of change of internal temperature rise, the duration of high temperature, the deviation between internal temperature rise and load changes, and the continuous deterioration status, according to the attention threshold, warning threshold, and alarm threshold. If the current internal temperature rise of phase B only briefly exceeds the attention threshold and the load change is explainable, an attention status is output. If three consecutive judgment windows exceed the warning threshold, or if the rate of increase of internal temperature rise is excessively large for three consecutive windows and the deviation from load changes continues to exceed the limit, a warning status is output. If the internal temperature rise reaches the alarm threshold, or if the load has significantly decreased but the internal temperature rise continues to rise and the cumulative increase exceeds 0.5℃, an alarm status is output. In this operating scenario, because the internal temperature rise of phase B exceeds the upper limit of the similar historical range by 1... If the temperature rises above 0% and continues to rise after the load decreases, the system will eventually output an alarm judgment result for the B-phase connector. Simultaneously, it will generate corresponding connector identifier, phase identifier, judgment time, anomaly level, internal temperature rise value, internal temperature rise change rate, change state category, load association status, template call status, and data reliability information. This judgment result can be written to a local record or uploaded to a higher-level monitoring platform, and a priority handling identifier will be generated to prompt maintenance personnel to prioritize the review and handling of the B-phase connector. Through the above operational logic, it can be seen that this method does not merely rely on the connector surface temperature for a rough alarm, but rather achieves a more accurate identification of abnormal heating states of cable branch box connectors based on continuous processing including multi-source acquisition, thermal interference separation, hysteresis compensation, internal temperature rise back-calculation, and comprehensive anomaly judgment.

[0025] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0026] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0027] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0028] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the temperature rise of cable branch box joints, characterized in that, include: S1 collects the surface temperature of each phase joint in the cable branch box, the ambient temperature inside the box, and the load current of each phase, and performs time alignment and abnormal data verification on the collected data. S2, based on the spatial arrangement relationship of each phase joint, the temperature distribution relationship inside the box and the interphase influence relationship, separates the influence of ambient temperature and interphase thermal interference in the surface temperature of each phase joint; S3, based on the load change process and the heat transfer hysteresis relationship of the joint insulation layer, performs hysteresis compensation on the surface temperature of each phase joint after thermal interference separation. S4. Based on the surface temperature of each phase joint and the heat transfer relationship of the joint structure after hysteresis compensation, the internal temperature rise of each phase joint is determined, and the internal temperature rise is correlated with the load change process. S5 performs anomaly detection based on the temperature rise changes inside each phase joint and outputs the detection results.

2. The method for monitoring the temperature rise of a cable branch box joint according to claim 1, characterized in that, S1 includes: Assign a unified timestamp to the surface temperature of the fixed measuring point on the outer insulation layer of the three-phase connector, the ambient temperature of the upper, middle and lower parts of the box, and the corresponding load current, and align them according to a unified analysis window; Periods of load disturbance and environmental disturbance are marked, and anomaly verification, status identification, local repair, channel retry and continuity constraints are performed respectively to form an effective monitoring sequence and status record.

3. The method for monitoring the temperature rise of a cable branch box joint according to claim 1, characterized in that, S2 includes: Establish spatial relationship information between the three-phase connector and the environmental measurement points inside the enclosure; The local ambient background temperature is determined based on the spatial correspondence between the target connector and the upper, middle and lower environmental measurement points, and the surface temperature is then corrected for environmental conditions. Identify the thermal impact state based on the three-phase load relationship and the environmentally corrected temperature rise relationship; Subtract the synchronous micro-amplitude crosstalk component under low-impact conditions; Construct or update the interphase influence matrix and estimate the thermal influence components of adjacent phases under enhanced thermal disturbance conditions; Through non-negative constraints, continuity constraints, and abnormal freezing and recovery processing, a net surface temperature rise sequence and corresponding state information are formed.

4. The method for monitoring the temperature rise of a cable branch box joint according to claim 3, characterized in that, Identify the thermally affected state based on the three-phase load relationship and the environmentally corrected temperature rise relationship, including: The current thermal impact state is determined based on the interphase load deviation, the interphase environmental correction temperature rise deviation, and the degree of deviation of a single phase relative to the average values ​​of the other two phases. The current thermal impact state is then divided into low impact state, intermediate transition state, and enhanced thermal interference state. The intermediate transition state is first processed using the basic separation method, and after maintaining this state for multiple consecutive analysis windows, it is switched to the enhanced separation method.

5. The method for monitoring the temperature rise of a cable branch box joint according to claim 1, characterized in that, S3 includes: Compensation determination is made based on the segmented state of load current and the relationship between net surface temperature rise and load change direction. The heat transfer hysteresis parameter template that matches the joint specification is called, the compensation amount is determined according to different operating states, and combined with parameter updates, version rollback, short-term missing conservative compensation and low-confidence freeze processing, a corrected surface temperature rise sequence and corresponding state information are formed.

6. The method for monitoring the temperature rise of a cable branch box joint according to claim 5, characterized in that, Compensation determination is based on the segmented state of the load current and the relationship between the net surface temperature rise and the direction of load change, including: Window analysis is performed on the load current, and steady-state, load increase, load decrease and fluctuation states are identified based on the current change rate, cumulative increase and decrease, and direction reversal. When the identification result is an increased load, decreased load, or fluctuating state, or when the direction of change of net surface temperature rise is continuously mismatched with the direction of change of corresponding load, hysteresis compensation determination is triggered.

7. The method for monitoring the temperature rise of a cable branch box joint according to claim 1, characterized in that, S4 includes: Call upon the structural heat transfer template that matches the target joint; Based on the corrected surface temperature rise, the temperature rise of the internal core area is determined layer by layer according to the layered temperature difference mapping relationship; The current main value of internal temperature rise is determined by verifying the consistency between short and long windows; The status is determined by combining the template replacement status, confidence level, and the correlation between internal temperature rise and load change; Record the internal temperature rise sequence, rate of change sequence, and corresponding state information.

8. The method for monitoring the temperature rise of a cable branch box joint according to claim 7, characterized in that, Based on the corrected surface temperature rise, the temperature rise of the internal core region is determined layer by layer according to the layered temperature difference mapping relationship, including: Based on the interlayer temperature difference mapping coefficient in the structural heat transfer template; The order is from the outer surface to the middle of the insulation layer, and from the middle of the insulation layer to the crimping area. Combine the current load range with the corrected surface temperature rise range; The temperature difference increment of each layer is determined by looking up a table or interpolation, and the temperature difference increment of each layer is sequentially added to the temperature rise of the correction surface. When the structural heat transfer template has more layers, it continues to accumulate according to the layer order given by the template.

9. The method for monitoring the temperature rise of a cable branch box joint according to claim 1, characterized in that, S5 includes: When the internal temperature rise discrimination condition is met, a stratified threshold is constructed based on the historical baseline of internal temperature rise corresponding to the current load range and the thermal background range. The anomaly level is determined by combining the absolute value of internal temperature rise, the rate of rise, the duration of high temperature, the deviation from load changes, and the three-phase group discrimination relationship. It performs verification, recalculation, or rollback when there are critical states, changes in key intermediate quantities, or data link anomalies, and records the corresponding status information.

10. A method for monitoring the temperature rise of a cable branch box joint according to claim 9, characterized in that, The anomaly level is determined by combining the absolute value of internal temperature rise, rate of rise, duration of high temperature, deviation from load changes, and three-phase group discrimination relationship, including: The joint abnormality level is determined by combining the following factors: the duration of internal temperature rise exceeding the limit, the deviation of the internal temperature rise rate from the load change, the duration of temperature rise after the load stabilizes or falls, the synchronous relationship of the three-phase internal temperature rise, the relationship of local environmental background changes, and the degree of deviation of the single phase.

Images