A terminal contact degradation virtual temperature measurement and failure early warning method for low-voltage electric energy metering box

Abstract

The present application belongs to the technical field of low-voltage metering equipment thermal state perception and intelligent early warning, and discloses a terminal contact degradation virtual temperature measurement and failure early warning method for a low-voltage electric energy metering box. The method collects connection point related voltage, current, power, harmonic, environmental temperature and maintenance event data; establishes a reference electrothermal agent model according to the metering box connection topology to solve the reference hot spot temperature; extracts the voltage drop response coefficient, load square accumulation, harmonic thermal stress coefficient and phase imbalance thermal bias coefficient to construct a contact degradation characteristic vector; uses a lightweight temperature correction model with a storage occupation of no more than 256kB to output the virtual hot spot temperature; and determines the degradation level according to the virtual hot spot temperature, temperature rise slope and contact degradation index and outputs the failure early warning result and maintenance suggestion. The present application can continuously and online perceive the hidden hot spots of the metering box terminals and connection points without adding special temperature measurement hardware, effectively improves the failure early warning lead time and reduces the false negative rate.

Description

Technical Field

[0001] This invention belongs to the field of thermal status sensing and intelligent early warning technology for low-voltage metering equipment, and particularly relates to a method for virtual temperature measurement and failure early warning of terminal contact degradation in low-voltage power metering boxes. Background Technology

[0002] During long-term operation, low-voltage electricity metering boxes are prone to contact degradation at various points, including the incoming line terminals, the terminals before and after the meter, the busbar connection points, and the outgoing line terminals, due to factors such as loose screws, aging connectors, oxidation of contact surfaces, or conductor deformation caused by heat. Contact degradation can lead to increased local contact resistance, resulting in hidden hot spots, localized overheating, or even terminal failure.

[0003] Existing methods for troubleshooting terminal overheating mainly rely on manual temperature measurement, infrared inspection, or the installation of temperature sensors at specific locations. These methods suffer from limitations such as limited inspection frequency, inability to continuously monitor online conditions, incomplete coverage of measurement points, and high retrofit costs. In particular, in scenarios involving multi-position low-voltage power metering boxes, it is difficult to perform long-term, stable, real-time temperature measurement of all critical connection points.

[0004] Furthermore, judging terminal risk solely based on current magnitude or simple temperature rise thresholds makes it difficult to distinguish between normal temperature rise caused by load fluctuations and abnormal heating due to contact degradation. In contact degradation scenarios, in addition to current stress, voltage dips under sudden load changes, harmonic-induced thermal stress, and thermal bias caused by three-phase imbalance also affect hotspot evolution. If a degradation characteristic reflecting the connection point's contact state cannot be constructed and jointly modeled with a reference electrothermal relationship, it will be difficult to achieve virtual terminal temperature measurement and failure early warning without relying on dedicated temperature measurement hardware. Summary of the Invention

[0005] To overcome the problems of existing terminal heating identification methods that rely on manual temperature measurement or additional temperature sensors, have difficulty in continuously sensing hidden hot spots at low-voltage power metering box connection points, and lack the ability to identify contact degradation, this invention proposes a virtual temperature measurement and failure early warning method for terminal contact degradation in low-voltage power metering boxes. The method collects relevant voltage, current, power, harmonics, environmental, and maintenance event data for connection points. Based on the metering box connection topology, branch current, preset reference contact resistance, ambient temperature, thermal resistance coefficient, thermal coupling terms of adjacent nodes, and the identification results of shutdown or low-load cooling curves, a reference electrothermal proxy model for each connection point is established. Further, the voltage drop response coefficient, load square accumulation, harmonic thermal stress coefficient, and phase-to-phase unbalanced thermal bias coefficient are extracted. A lightweight temperature correction model with a storage footprint of no more than 256kB is used to obtain the virtual hot spot temperature of each connection point. Based on the virtual hot spot temperature, temperature rise slope, and contact degradation index, failure early warning results and maintenance suggestions are output, thereby achieving online virtual temperature measurement and degradation early warning for low-voltage power metering box terminals and connection points.

[0006] To achieve the above objectives, the specific technical solution of the present invention for a virtual temperature measurement and failure early warning method for terminal contact degradation in a low-voltage power metering box is as follows:

[0007] A virtual temperature measurement and failure early warning method for terminal contact degradation in a low-voltage energy metering box is applied to a low-voltage energy metering box or a low-voltage energy metering box monitoring terminal with functions for acquiring voltage, current, power, harmonics, ambient temperature, and recording maintenance events. The method includes the following steps:

[0008] S1. Collect multi-source operational data related to each connection point of the metering box within the most recent preset time window;

[0009] S2. Establish a reference electrothermal proxy model for each connection point based on the metering box connection topology, and solve for the reference hot spot temperature. When there is shutdown or low-load natural cooling data, obtain the thermal time constant based on the cooling curve of the connection point temperature decaying over time, so as to back-calculate or correct the equivalent thermal resistance coefficient.

[0010] S3. Extract the voltage drop response coefficient, load square accumulation, harmonic thermal stress coefficient and phase-to-phase unbalanced thermal bias coefficient of each connection point, and construct the contact degradation feature vector.

[0011] S4. Input the contact degradation feature vector and the reference hotspot temperature into the lightweight temperature correction model to obtain the virtual hotspot temperature of each connection point;

[0012] S5. Calculate the risk index based on the virtual hotspot temperature, temperature rise slope and contact degradation index, and determine the connection point degradation level and failure warning result based on the risk index;

[0013] S6. Output the location, risk level, and maintenance recommendations for high-risk connection points.

[0014] Furthermore, the multi-source operating data mentioned in S1 includes at least the incoming line voltage, branch current, active power, power factor, total harmonic current distortion rate, three-phase imbalance, ambient temperature, and maintenance event records; the maintenance event records include the time for unpacking, tightening, replacing connectors, and re-inspection confirmation, and are used for resetting the historical sequence of contact degradation index, segmenting training samples, or removing abnormal samples.

[0015] Furthermore, the reference hot spot temperature mentioned in S2 is determined by superimposing the ambient temperature, the Joule heating temperature rise term, and the adjacent node thermal coupling temperature rise term; the Joule heating temperature rise term is calculated based on the square of the branch current, the preset reference contact resistance, and the equivalent thermal resistance coefficient; the adjacent node thermal coupling temperature rise term is obtained by multiplying the difference between the reference hot spot temperature of each adjacent node and the ambient temperature by the corresponding thermal coupling coefficient and then summing them; for cases with multiple thermal coupling nodes, the reference hot spot temperature equations of each connection point are formed into a linear equation system or solved iteratively; the preset reference contact resistance is determined using the factory design value, the minimum temperature rise back-calculated value under historical steady-state operating conditions, or the prototype calibration value; the equivalent thermal resistance coefficient is updated by: collecting the connection point temperature decay curve during shutdown or low-load natural cooling, obtaining the thermal time constant through exponential function fitting, and back-calculating or correcting the equivalent thermal resistance coefficient based on the thermal time constant.

[0016] Furthermore, the voltage drop response coefficient mentioned in S3 is determined based on the ratio of the absolute value of the voltage change to the absolute value of the current change at adjacent sampling times; the load square accumulation is the accumulation of time over time by the square of the ratio of the current to the rated current at each sampling point within a preset time window; the harmonic thermal stress coefficient is determined based on the product of the total harmonic current distortion rate and the square of the current load ratio; and the phase-to-phase unbalanced thermal bias coefficient is determined based on the ratio of the difference between the maximum and minimum values ​​of the three-phase current at the same time to the average value.

[0017] Furthermore, the contact degradation index is obtained by multiplying the voltage drop response coefficient, harmonic thermal stress coefficient, phase-to-phase unbalanced thermal bias coefficient, and load square accumulation by their respective weights and summing them. Each weight is non-negative and the sum is 1. The weights are determined jointly based on the physical prototype loading test data, on-site operation history data, and manual maintenance records, and are normalized to ensure that the contact degradation index of samples with normal contact status is less than 1.

[0018] Furthermore, the lightweight temperature correction model described in S4 adopts a lightweight gradient boosting regression model, consisting of no more than 80 regression trees, with a maximum depth of 2 to 4 for a single tree, a learning rate of 0.02 to 0.10, a minimum number of leaf node samples of 3 to 10, and a total model storage footprint of no more than 256kB.

[0019] Furthermore, the risk index mentioned in S5 is determined by a weighted sum of the temperature exceedance term, the temperature rise slope term, and the degradation exceedance term; the connection points are classified into normal, mild degradation, moderate degradation, and severe degradation levels according to the risk index; the temperature exceedance term is determined based on the amount by which the virtual hotspot temperature exceeds the preset temperature warning threshold, and the preset temperature warning threshold is determined based on the sum of the ambient temperature and the temperature rise warning limit corresponding to the connection point.

[0020] Furthermore, the temperature rise warning limit is set as follows: 45K for metering junction box terminals; 60K for interference fit connectors with a diameter of 6.0 mm or 7.5 mm; and 70K for interference fit connectors with a diameter of 8.5 mm.

[0021] Furthermore, in S5, when multiple consecutive sampling cycles satisfy the condition that the temperature rise slope is greater than the preset slope threshold and the contact degradation index shows an upward trend, a trend failure warning is output even if the virtual hotspot temperature has not yet reached the formal overheating alarm threshold; the number of consecutive sampling cycles ranges from 2 to 20.

[0022] Furthermore, in S6, the maintenance recommendations are sorted by risk index from high to low, outputting the locations of high-risk connection points, and providing corresponding maintenance recommendations based on the connection point type. The method also includes a verification step: pre-embedding thermocouples at key connection points or using an infrared thermal imager to synchronously collect real hotspot temperatures as verification benchmarks. When the average absolute error of the virtual hotspot temperature, the early warning lead time, and the missed detection rate meet the preset index thresholds, it is determined that the lightweight temperature correction model meets the field deployment requirements.

[0023] The present invention provides a virtual temperature measurement and failure early warning method for terminal contact degradation in low-voltage power metering boxes, which has the following advantages: Instead of simply judging terminal risk based on current thresholds or fixed temperature rise thresholds, the present invention first establishes a reference electrothermal proxy model with a clear solution method based on the connection topology of the low-voltage power metering box. Then, it uses voltage drop response, harmonic thermal stress, unbalanced thermal bias, and load square accumulation to jointly construct contact degradation characteristics, and outputs virtual hot spot temperatures through a lightweight temperature correction model. Furthermore, the present invention uses maintenance event records for resetting or segmenting the historical sequence of contact degradation indices, and can use shutdown or low-load natural cooling curves to correct thermal resistance parameters, enabling the virtual temperature measurement model to be updated according to the on-site operation and maintenance status. Compared with methods relying solely on manual inspection, additional sensors, or simple thermal models, the present invention can continuously identify hidden hot spots and contact degradation risks at terminals and connection points without adding dedicated temperature measurement hardware, improving the early warning of failures, reducing the false negative rate, and enhancing engineering usability. Attached Figure Description

[0024] Figure 1 This is a flowchart of the virtual temperature measurement and failure early warning method for terminal contact degradation in low-voltage power metering boxes according to the present invention.

[0025] Figure 2 This is a schematic diagram of the connection point and the virtual temperature measurement node.

[0026] Figure 3 This is a schematic diagram comparing the virtual temperature measurement results of the terminal contact points. The horizontal axis represents time or sampling number, and the vertical axis represents the hot spot temperature. The curves represent the reference hot spot temperature, the virtual hot spot temperature, and the experimentally measured hot spot temperature, respectively.

[0027] Figure 4 This diagram illustrates the changes in the contact degradation index and the temperature rise slope. The horizontal axis represents time or sampling number, and the vertical axis represents the contact degradation index and the temperature rise slope, respectively, to demonstrate the trend warning triggering process.

[0028] Figure 5 This is a diagram comparing the lead time for failure warnings, where the horizontal axis represents different warning methods or operating conditions, and the vertical axis represents the lead time relative to the overheating threshold alarm.

[0029] Figure 6 This diagram illustrates the comparison of terminal degradation recognition effects using different methods. The horizontal axis represents the recognition method or sample type, while the vertical axis represents the recognition accuracy, false negative rate, or comprehensive evaluation index. Detailed Implementation

[0030] To better understand the purpose, structure, and function of this invention, the following description, in conjunction with the accompanying drawings, provides a more detailed account of a virtual temperature measurement and failure early warning method for terminal contact degradation in a low-voltage power metering box.

[0031] like Figure 1 As shown, the present invention provides a virtual temperature measurement and failure early warning method for terminal contact degradation in a low-voltage power metering box. This method is applied to a low-voltage power metering box or a low-voltage power metering box monitoring terminal that has functions for acquiring voltage, current, power, harmonics, ambient temperature, and recording maintenance events. The method includes the following steps:

[0032] In this embodiment, a single-phase multi-position non-metallic low-voltage energy metering box with interference fit connector and a three-phase multi-position non-metallic low-voltage energy metering box with interference fit connector from a certain company are selected as implementation objects. The single-phase prototypes include SXD2 and PXD2, with a rated operating voltage of 380V, a rated branch current of 60A, 9 meter positions, and an IP44 enclosure protection rating. The three-phase prototypes include SXS2 and PXS2, with a rated operating voltage of 380V, a rated branch current of 100A, 2 meter positions, and an IP44 enclosure protection rating. All prototypes undergo full-performance type testing according to Q / GDW 11008-2013 "Technical Specification for Low-Voltage Energy Metering Boxes". The sampling period is preferably 1 minute, and the most recent preset time window is preferably the most recent 30 minutes. The virtual temperature measurement target preferably includes one or more of the following: inlet terminal, meter front terminal, meter rear terminal, busbar connection point, metering junction box terminal, and outlet terminal.

[0033] S1: Collect multi-source operational data related to the connection point. In this embodiment, the incoming line voltage, branch current, active power, power factor, total harmonic current distortion rate, three-phase imbalance, and ambient temperature are collected. For time periods with maintenance or repair activities, maintenance event information such as unpacking, tightening, connector replacement, and re-inspection confirmation times are also recorded simultaneously; the maintenance event information is used to assess the contact degradation index after maintenance. The historical sequence is reset, and the training samples are segmented or abnormal samples with inconsistent states before and after maintenance are removed. Thus, the electrical, environmental, and event quantities required for the thermal state evolution of the connection point can be obtained simultaneously.

[0034] S2: Establish a reference electrothermal proxy model. For example... Figure 2 As shown, based on the connection topology of the low-voltage energy metering box, the incoming line terminal, the terminal before the meter, the terminal after the meter, the busbar connection point, and the outgoing line terminal are abstracted into several connection point nodes. For the first... The connection points at time... according to Solving for reference hotspot temperature .

[0035] in, For ambient temperature, For the first Each connection point corresponds to a branch current. To preset the reference contact resistance, the factory design value, the minimum temperature rise back-calculated value under historical steady-state operating conditions, or the prototype calibration value can be used. This is the equivalent thermal resistance coefficient. In order to be with the first Each connection point has a set of thermally coupled neighboring nodes. is the thermal coupling coefficient between adjacent nodes. The values ​​used are the factory design values, the minimum temperature rise calculated under historical steady-state operating conditions, or the prototype calibration values. For the first The connection points at time... The reference hotspot temperature is used. For cases with multiple thermally coupled nodes, the reference hotspot temperature equations for each connection point can be combined into a linear system of equations or solved iteratively. When shutdown or low-load natural cooling data is available, cooling curves showing the temperature decay of the connection point over time can be collected. The thermal time constant was obtained by fitting the cooling curve. To reverse or correct .

[0036] S3: Extract contact degradation features. In this embodiment, voltage sag response coefficients are preferably extracted. Accumulated load square Harmonic thermal stress coefficient and unbalanced thermal bias coefficient And construct contact degradation feature vectors .

[0037] in, , used to indicate the first The voltage sag response coefficients at each connection point, where... , ; For the first Each connection point corresponds to a branch voltage or terminal voltage drop. For the corresponding branch current, To prevent tiny positive numbers with a denominator of zero;

[0038] , is used to represent the heat accumulation of the square of the current within a preset time window, where The number of sampling points within a preset time window. The sampling interval is... For the first Each connection point corresponds to the rated current of the branch.

[0039] , is used to represent the additional thermal stress caused by harmonic currents, where, The total harmonic current distortion rate is expressed as a decimal.

[0040] This is used to represent the thermal bias caused by three-phase imbalance, for single-phase connection points. Always takes 0.

[0041] Furthermore, the contact degradation index is obtained by weighting and fusing the above characteristics. , to The weight coefficients are non-negative and satisfy the following conditions: And normalized calibration was used to make the samples in normal contact state... Less than 1. The weighting coefficient is determined based on the combined data of the physical prototype loading test, historical field operation data, and manual maintenance records. Figure 4 The diagram illustrates the changes in the contact degradation index and the temperature rise slope over time.

[0042] S4: Perform virtual temperature measurement correction. In this embodiment, it will be performed by... , , , and The input vector is used to input the lightweight temperature correction model, and the output is the virtual hotspot temperature of each connection point. The preferred model is a lightweight gradient boosting regression model, which can consist of 80 CART regression trees with a maximum depth of 2 to 4 (3 in this embodiment), a learning rate of 0.02 to 0.10 (0.05 in this embodiment), a minimum number of leaf node samples of 3 to 10 (5 in this embodiment), and a total model storage footprint of no more than 256kB. Figure 3 A schematic comparison of the hotspot temperature, reference agent temperature, and virtual temperature measurement results obtained in the experiment is shown.

[0043] S5: Perform degradation level judgment and failure warning. In this embodiment, based on the virtual hotspot temperature... Temperature rise slope and contact degradation index Calculate the risk index ,

[0044] ,

[0045] in , Sampling interval. Preset temperature warning threshold. Preferred by Sure, The temperature rise warning limit is the value corresponding to the connection point; for the metering junction box terminals, 45K is preferred; for φ6.0mm or φ7.5mm interference fit connectors, 60K is preferred; for φ8.5mm interference fit connectors, 70K is preferred.

[0046] according to The connection points are classified into normal, mildly degraded, moderately degraded, and severely degraded levels; the risk level thresholds can be determined by validation set ROC curves, manual inspection labels, or prototype loading tests. For example, [the following can be used as a reference:] Values ​​below the first threshold are considered normal; values ​​between the first and second thresholds are considered mild degradation; values ​​between the second and third thresholds are considered moderate degradation; and values ​​above the third threshold are considered severe degradation. When the temperature rise exceeds a preset threshold, an early warning result is output. This warning occurs when the temperature rise slope is greater than a preset slope threshold for N consecutive sampling periods. When it is on an upward trend, even Even if the overheating alarm threshold has not been reached, a trend failure warning is still output; where N is a preset positive integer, ranging from 2 to 20, preferably from 3 to 10, for example, N=3. For connection points with maintenance event records, the contact degradation index is reset in the first effective sampling window after maintenance is completed. Historical sequences or segmented statistics of samples before and after maintenance. Figure 5The results of the advance warning amount of the method of the present invention compared with the traditional threshold warning method are given.

[0047] S6: Output Risk Locations and Maintenance Recommendations. In this embodiment, the locations of high-risk connection points are output in descending order of risk index, along with corresponding maintenance recommendations. For high-risk incoming line terminals, priority is given to checking the terminal fastening condition and contact surface oxidation. For high-risk busbar connection points, priority is given to checking the busbar connection bolts and contact tightening condition. For high-risk terminals before or after the meter, priority is given to checking the metering circuit connectors, terminals, and wire crimping condition.

[0048] S7: Verification Scheme Design. In this embodiment, it is preferable to use full-performance type test data of physical prototypes, on-site operation history data, and manual maintenance records to train and verify the method of the present invention. Normal samples can be collected by SXD2, PXD2, SXS2, and PXS2 prototypes under rated current temperature rise tests: When the ambient temperature of the PXD2 prototype is 22.0℃, the main bus test current is 163A, and the sub-circuit test current is approximately 54.1A to 54.4A, the temperature rise at the input terminal of the C5 circuit energy meter connector is 39.9K and 40.4K, and the temperature rise at the output terminal is 38.0K and 39.6K; when the ambient temperature of the SXD2 prototype is 24.2℃, the main bus test current is 163A, and the sub-circuit test current is approximately 54.1A to 54.2A, the temperature rise at the output terminal of the C5 circuit energy meter connector is approximately 39.9K and 40.4K, and the temperature rise at the output terminal is approximately 38.0K and 39.6K. The temperature rise at the input terminal of the energy meter connector was 37.8K and 39.7K, and the temperature rise at the output terminal was 39.2K and 40.5K. For the PXS2 prototype, at an ambient temperature of 16.4℃ and a main bus test current of 181A, the temperature rise at the input terminal of the C1 circuit energy meter connector was 41.2K to 42.3K, and the temperature rise at the output terminal was 41.8K to 43.5K. For the SXS2 prototype, at an ambient temperature of 24.4℃ and a main bus test current of 180A, the temperature rise at the input terminal of the C1 circuit energy meter connector was 39.6K to 40.7K, and the temperature rise at the output terminal was 43.2K to 44.7K. Temperature rise measurements of interference fit connectors after at least 1000 mating cycles can be used as baseline samples after mating aging. The baseline values ​​for the SXD2 prototype are 46.8K, PXD2 prototype is 46.1K, SXS2 prototype is 43.9K, and PXS2 prototype is 46.5K, all below the corresponding connector temperature rise limits. Degradation samples were collected under conditions of reduced terminal clamping force, contact surface oxidation, connector aging, and harmonic contamination. Thermocouples were pre-embedded at key connection points of the prototypes, or infrared thermal imagers were used to simultaneously collect real hot spot temperatures as verification benchmarks for virtual temperature measurement results. During shutdown or low-load natural cooling tests, the cooling curves of the connection points were simultaneously recorded for identification of thermal time constants and thermal resistance parameters.

[0049] In this embodiment, the contact degradation level label is preferably determined by both the relative temperature rise threshold and the contact state: when the temperature rise at the connection point is not higher than... If no maintenance or confirmation of abnormality is found, mark it as normal; when the temperature at the connection point rises above [a certain value], [the condition is as follows]. and not higher than If the contact resistance increases by 20% to 50% relative to the reference contact resistance, it is marked as mild degradation; when the connection point temperature rises above a certain level... And SS is not higher than If the contact resistance increases by 50% to 100% relative to the reference contact resistance, it is marked as moderate degradation; when the connection point temperature rises above a certain level... If the contact resistance increases by more than 100% relative to the reference contact resistance, or if maintenance event records confirm severe loosening, burning, or connector failure, it is marked as severely degraded. During the verification phase, indicators such as the average absolute error, maximum error, warning lead time, false alarm rate, and missed alarm rate of the virtual hotspot temperature are statistically analyzed. When the average absolute error is less than the preset temperature error threshold, the warning lead time is greater than the preset lead time threshold, and the false alarm rate is lower than the preset false alarm rate threshold, the lightweight temperature correction model is determined to meet the field deployment requirements. Figure 6 The results of the comparison between the method of the present invention and the current threshold method, the fixed temperature rise threshold method, and the method of referring only to the electrothermal model are illustrated.

[0050] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box, characterized in that, For use in low-voltage energy metering boxes or low-voltage energy metering box monitoring terminals with functions of acquiring voltage, current, power, harmonics, ambient temperature, and recording maintenance events, the following steps are included: S1. Collect multi-source operational data related to each connection point of the metering box within the most recent preset time window; S2. Establish a reference electrothermal proxy model for each connection point based on the metering box connection topology, and solve for the reference hot spot temperature. When there is shutdown or low-load natural cooling data, obtain the thermal time constant based on the cooling curve of the connection point temperature decaying over time, so as to back-calculate or correct the equivalent thermal resistance coefficient. S3. Extract the voltage drop response coefficient, load square accumulation, harmonic thermal stress coefficient and phase-to-phase unbalanced thermal bias coefficient of each connection point, and construct the contact degradation feature vector. The voltage drop response coefficient is determined based on the ratio of the absolute value of the voltage change to the absolute value of the current change at adjacent sampling times; the load square accumulation is the accumulation of time over time of the square of the ratio of the current to the rated current at each sampling point within a preset time window. The harmonic thermal stress coefficient is determined by the product of the total harmonic current distortion rate and the square of the current load ratio; the phase imbalance thermal bias coefficient is determined by the ratio of the difference between the maximum and minimum values ​​of the three-phase current at the same moment to the average value. The contact degradation index is obtained by multiplying the voltage drop response coefficient, harmonic thermal stress coefficient, phase-to-phase unbalanced thermal bias coefficient, and load square accumulation by their respective weights and summing them. Each weight is non-negative and the sum is 1. The weights are determined jointly based on the load test data of the physical prototype, the historical data of field operation, and the manual maintenance records, and are normalized to ensure that the contact degradation index of the sample with normal contact status is less than 1. S4. Input the contact degradation feature vector and the reference hotspot temperature into the lightweight temperature correction model to obtain the virtual hotspot temperature of each connection point; The lightweight temperature correction model adopts a lightweight gradient boosting regression model. S5. Calculate the risk index based on the virtual hotspot temperature, temperature rise slope and contact degradation index, and determine the connection point degradation level and failure warning result based on the risk index; S6. Output the location, risk level, and maintenance recommendations for high-risk connection points.

2. The method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box according to claim 1, characterized in that, The multi-source operating data mentioned in S1 includes at least the incoming line voltage, branch current, active power, power factor, total harmonic current distortion rate, three-phase imbalance, ambient temperature, and maintenance event records. The maintenance event records include the time for unpacking, tightening, replacing connectors, and re-inspection confirmation, and are used for resetting the historical sequence of contact degradation index, segmenting training samples, or removing abnormal samples.

3. The method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box according to claim 1, characterized in that, The reference hot spot temperature in S2 is determined by superimposing the ambient temperature, the Joule heating temperature rise term, and the adjacent node thermal coupling temperature rise term. The Joule heating temperature rise term is calculated based on the square of the branch current, the preset reference contact resistance, and the equivalent thermal resistance coefficient. The adjacent node thermal coupling temperature rise term is obtained by multiplying the difference between the reference hot spot temperature of each adjacent node and the ambient temperature by the corresponding thermal coupling coefficient and then summing them. For cases with multiple thermal coupling nodes, the reference hot spot temperature equations of each connection point are combined into a linear equation system or solved iteratively. The preset reference contact resistance is determined using the factory design value, the minimum temperature rise back-calculated value under historical steady-state operating conditions, or the prototype calibration value. The equivalent thermal resistance coefficient is updated by collecting the connection point temperature decay curve during shutdown or low-load natural cooling, obtaining the thermal time constant through exponential function fitting, and back-calculating or correcting the equivalent thermal resistance coefficient based on this thermal time constant.

4. The method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box according to claim 1, characterized in that, The lightweight gradient boosting regression model consists of no more than 80 regression trees, with a maximum depth of 2 to 4 for a single tree, a learning rate of 0.02 to 0.10, a minimum number of leaf node samples of 3 to 10, and a total model storage footprint of no more than 256kB.

5. The method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box according to claim 1, characterized in that, The risk index mentioned in S5 is determined by a weighted sum of the temperature exceedance term, the temperature rise slope term, and the degradation exceedance term. Based on the risk index, the connection points are divided into normal, mild degradation, moderate degradation, and severe degradation levels. The temperature exceedance term is determined based on the amount by which the virtual hotspot temperature exceeds a preset temperature warning threshold. The preset temperature warning threshold is determined based on the sum of the ambient temperature and the temperature rise warning limit corresponding to the connection point.

6. The method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box according to claim 5, characterized in that, The temperature rise warning limit is set as follows: 45K for metering junction box terminals; 60K for interference fit connectors with a diameter of 6.0 mm or 7.5 mm; and 70K for interference fit connectors with a diameter of 8.5 mm.

7. The method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box according to claim 1, characterized in that, In S5, when the temperature rise slope is greater than the preset slope threshold and the contact degradation index shows an upward trend for multiple consecutive sampling cycles, a trend failure warning is output even if the virtual hot spot temperature has not yet reached the formal overheating alarm threshold. The number of consecutive sampling periods ranges from 2 to 20.

8. The method for virtual temperature measurement and failure early warning of terminal contact degradation in a low-voltage power metering box according to claim 1, characterized in that, The maintenance recommendations described in S6 are sorted by risk index from high to low, outputting the locations of high-risk connection points, and providing corresponding maintenance recommendations based on the connection point type; the method also includes a verification step: pre-embedding thermocouples at key connection points or using an infrared thermal imager to synchronously collect real hotspot temperatures as verification benchmarks, and determining that the lightweight temperature correction model meets the field deployment requirements when the average absolute error of the virtual hotspot temperature, the early warning lead time, and the missed detection rate meet the preset index thresholds.

Images