A low-voltage cable branch box end power supply capacity dynamic evaluation method and system

Abstract

The application discloses a kind of low-voltage cable branch box end power supply capacity dynamic evaluation method and system, each branch conductor temperature signal is collected, and the heat resistance and thermal parameter of environment are determined through the thermal analysis under the condition of power failure and light load, and the upper limit of branch current is determined by counting instantaneous most unfavorable fluctuation;Again, the coupling strength of heat conduction and convection between adjacent branches is investigated, the current upper limit is reduced, and the branch margin estimate is obtained by taking into account the influence of adjacent road full load;Further, the margin estimate is continuously verified using the real load transfer event in distribution network, and the temperature margin curve is fitted to form, the sparse section is treated by narrowing extrapolation;Finally, based on the slope of the curve, the sensitivity of each branch margin is obtained, and the network health level is formed by combining the topological cut point identification, and the end power supply margin result is output.The application realizes the real-time matching of actual available capacity and scheduling decision of end network, and provides basis for load dynamic response and optimized allocation.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition assessment technology, and in particular to a method and system for dynamic assessment of the power supply capacity at the end of a low-voltage cable branch box. Background Technology

[0002] The terminal branch box of a low-voltage distribution network is the last-level aggregation node for load access, and its actual carrying capacity directly determines whether the terminal network can flexibly respond to load changes and fault transfer requirements. In current engineering practice, the power supply capacity assessment of each branch in the branch box generally uses the factory rated current as a fixed indicator. This indicator is set based on a uniform environmental assumption and does not consider the impact of actual cable laying conditions, seasonal temperature differences, and differences in heat dissipation inside the box on the conductor's carrying capacity. As a result, there is a significant deviation between the assessment results and the actual operating status of the line.

[0003] Although technologies such as dynamic cable capacity expansion and status awareness have been applied in some scenarios, most existing solutions focus on the temperature-current relationship of a single branch, paying less attention to the heat cross-influence caused by the compact structure of adjacent branches within the branch box. They also fail to use actual load transfer events in the distribution network to verify and correct the assessment results, resulting in a lack of operational verification support for the obtained margin values. At the same time, existing methods lack the means to identify key nodes formed by changes in connection methods within the box, making it difficult to link the local branch margin with the overall network structure risk. This means that the dispatching side can only arrange loads based on conservative fixed capacity, and the surplus power supply capacity in the end network cannot be effectively utilized. Summary of the Invention

[0004] This invention discloses a method and system for dynamically evaluating the power supply capacity at the end of a low-voltage cable branch box. It aims to address the problems of relying on fixed rated parameters, ignoring inter-branch thermal coupling and topology risks, and failing to accurately reflect actual operating conditions when evaluating the power supply capacity of the branch box. By collecting conductor temperature signals to construct a thermal circuit model, calibrating the upper limit of current carrying capacity and margin estimate for each branch, verifying the margin curve by combining real power transfer events, identifying key topological nodes to form a network health level, and outputting end-point power supply margin results that can be used for dynamic scheduling, thereby improving the precision of load scheduling at the end of the distribution network.

[0005] The first aspect of this invention proposes a method for dynamically evaluating the power supply capability at the end of a low-voltage cable branch box, comprising: Acquire the temperature signals of conductors in each branch of the low-voltage cable branch box, screen outage periods based on the conductor temperature signals to obtain the outage calibration window, and determine the initial value of the environmental thermal resistance based on the thermal resistance characteristic analysis of the outage calibration window. The light-load calibration window is obtained by back-calculating the light-load operation period of the branch based on the initial value of the environmental thermal resistance. The thermal resistance characteristics of the branch are analyzed from the light-load calibration window to generate the branch thermal parameters. The branch thermal parameters are used to determine the instantaneous most unfavorable fluctuation to determine the upper limit of the branch current carrying capacity. The cross thermal coupling coefficient is determined by converting the full-load temperature rise of the branch current limit. The contribution of the adjacent full-load temperature rise is analyzed based on the cross thermal coupling coefficient to generate the margin correction coefficient. The margin correction coefficient is used to reduce the branch current limit to obtain the branch margin estimate. Based on the estimated branch load margin, a step transfer of branch load is tracked to form a transfer load list. The transfer load list is used to screen the transfer non-limited sections to obtain measured margin samples. The measured margin samples are then subjected to successive transfer temperature continuous fitting to form a temperature margin curve. The temperature margin curve is statistically analyzed to generate margin sensitivity by calculating the slope of temperature sensitivity of each branch. Based on the margin sensitivity, the terminal topology cut point branches are identified to form a network health level. Based on the network health level, the terminal margin distribution is analyzed to output the power supply margin result.

[0006] A second aspect of this invention provides a dynamic evaluation system for the power supply capability at the end of a low-voltage cable branch box, comprising: The thermal resistance calibration module is used to acquire the temperature signals of the conductors of each branch in the low-voltage cable branch box, screen the power outage period based on the conductor temperature signals to obtain the power outage calibration window, and perform thermal resistance feature analysis based on the power outage calibration window to determine the initial value of the environmental thermal resistance. The load measurement module is used to back-calculate the light-load operation period of the branch based on the initial value of the environmental thermal resistance to obtain the light-load calibration window, analyze the thermal resistance characteristics of the thermal circuit from the light-load calibration window to generate the thermal circuit parameters of the branch, and use the thermal circuit parameters of the branch to determine the instantaneous most unfavorable fluctuation to determine the upper limit of the branch current carrying capacity. The coupling correction module is used to determine the cross thermal coupling coefficient by converting the full-load temperature rise for the upper limit of the branch current carrying capacity, analyze the contribution of the adjacent full-load temperature rise based on the cross thermal coupling coefficient to generate the margin correction coefficient, and use the margin correction coefficient to reduce the upper limit of the branch current carrying capacity to obtain the branch margin estimate. The margin mapping module is used to track the step shift of branch load based on the branch margin estimate to form a transfer load list, screen the transfer load list to obtain the measured margin sample of the transfer non-limited section, and perform successive transfer temperature continuous fitting on the measured margin sample to form a temperature margin curve. The health assessment module is used to statistically analyze the temperature sensitivity slope of each branch of the temperature margin curve to generate margin sensitivity, identify the terminal topology cut point branches based on the margin sensitivity to form a network health level, and perform terminal margin distribution analysis based on the network health level to output power supply margin results.

[0007] The beneficial effects of this invention are reflected in the following points: First, the determination of the branch current-carrying limit no longer relies on the factory rated value or a uniform environmental assumption. Instead, it deduces the specific environmental thermal resistance and heat capacity parameters of each branch from the actual power outage cooling process and light load operation phase experienced by each branch. Furthermore, by distinguishing between fixed periodic interference and actual load fluctuations, the most unfavorable instantaneous fluctuations are included in the temperature budget before the current-carrying capacity is deduced. This ensures that the evaluation results correspond to the actual laying environment and ventilation status of the branch, overcoming the limitation of fixed parameter evaluation being disconnected from the actual on-site conditions. Second, the internal structure of the branch box is compact, and there are two types of heat exchange between adjacent branches: conduction and convection. A fully loaded branch will occupy part of the temperature rise space of the adjacent branch, a phenomenon that has long been ignored in previous single-branch evaluation models. This invention distinguishes between the two types of coupling paths and verifies whether their coupling strength is sustainable. It converts the previously unconsidered worst-case scenario of a fully loaded adjacent branch into a margin correction coefficient, so that the branch margin estimate is no longer an isolated value detached from the state of the adjacent branch. Finally, previous margin models lacked field verification, and relying on active testing to approach branch limits posed risks and costs. This invention, however, utilizes real load surge events such as power transfer switching to verify the margin curve, while simultaneously identifying topology cut points by engaging and disengaging tie switches. The combination of these two methods reveals complex risks that were previously undetectable when electrical topology analysis and thermal margin analysis were conducted independently. Specifically, a branch may be both a critical topology node and have relatively limited margin, providing the dispatching side with key considerations that were difficult to obtain using previous methods. Attached Figure Description

[0008] Figure 1 This is a flowchart of a method for dynamically evaluating the power supply capacity at the end of a low-voltage cable branch box according to the present invention.

[0009] Figure 2 This is a schematic diagram of the terminal distribution network topology and cut point risk classification of the present invention.

[0010] Figure 3 This is a structural block diagram of a dynamic evaluation system for the power supply capacity at the end of a low-voltage cable branch box according to the present invention.

[0011] Among them: 1-Distribution transformer; 2-Low-voltage main feeder; 3-Low-voltage cable branch box; 4-Outgoing cable; 5-End load node; 6-Tie switch; 7-Permanent cut-off point; 8-Temporary cut-off point; 9-Level 1 health zone; 10-Level 3 high-risk zone. Detailed Implementation

[0012] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0013] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0014] The technical solutions of the embodiments of this application will be described below.

[0015] like Figure 1 As shown, this embodiment of the invention provides a method for dynamically evaluating the power supply capability at the end of a low-voltage cable branch box, including the following steps S101-S105: Step S101: Obtain the temperature signals of the conductors of each branch in the low-voltage cable branch box, screen the power outage period based on the conductor temperature signals to obtain the power outage calibration window, and perform thermal resistance characteristic analysis based on the power outage calibration window to determine the initial value of the environmental thermal resistance.

[0016] Specifically, the temperature signals of the conductors in each branch of the low-voltage cable distribution box are acquired. A thin-film temperature sensor is mounted on the conductor of each outgoing cable within the distribution box, near the terminal block. The sensor is tightly attached to the conductor surface and secured with insulating tape to minimize interference from air convection within the box. Each branch's sensor is independently numbered and connected to the same data acquisition unit, with a sampling period of one minute. The temperature along the length of the same conductor is not uniform; due to contact resistance, the temperature rise near the terminal block is typically slightly higher than in the middle section of the cable. The sensors are uniformly mounted in a fixed position near the terminal block to provide a comparable benchmark for readings across different branches. The conductor temperature signals are organized according to two dimensions: branch number and sampling time. The readings of each branch at the same time are arranged and presented to form the overall temperature cross-section of the distribution box. Within the branch box, adjacent branches conduct heat through a shared metal mounting structure and exchange heat via convection through heated airflow. A sudden increase in current in one branch results in a surge in heat generation. This heat travels along these two paths to the adjacent conductor, causing the adjacent conductor's sensor to read a slight, synchronous temperature rise unrelated to its own load. This crosstalk component is embedded in the conductor temperature signal and must be removed from the temperature rise component caused by the branch's own load during subsequent analysis. Heat transfer decreases with increasing conductor spacing, with the most significant impact on adjacent branches, and a substantial reduction beyond a single conductor spacing. Conductor temperature signal acquisition is performed continuously under all operating conditions. Each step extracts a required segment from this continuous time sequence according to the operating conditions, with the extraction conditions being distinct and non-overlapping.

[0017] In some embodiments, obtaining a power outage calibration window by screening power outage periods based on the conductor temperature signal includes: identifying a gradual temperature rise pattern before the power outage based on the conductor temperature signal to obtain a power outage nature sample; verifying discrepancies in the pattern based on the power outage nature sample to generate a power outage type label; calculating the residual heat decay time constant through the power outage type label to obtain the residual heat dissipation delay; and performing a delayed dissipation window interception based on the residual heat dissipation delay to obtain the power outage calibration window.

[0018] Based on conductor temperature signals, a sample of outage characteristics is obtained by identifying the gradual temperature rise pattern before an outage. The moment when the conductor temperature signal reading suddenly drops to near the ambient temperature level inside the enclosure is considered the starting point of a suspected outage event. However, a sudden drop could correspond to a true outage due to a line trip, or it could correspond to the load on this branch being transferred to an adjacent branch after the tie switch is closed, resulting in zero current. A single feature like a sudden drop is insufficient to distinguish between the two; it is necessary to trace the temperature rise trend over a period prior to the sudden drop for corroboration. The conductor temperature signal is taken from the 30 minutes before the start of the sudden drop. During this period, it is checked whether the temperature has entered a gradual rise or remained stable. If the temperature rise rate is below 0.1 degrees Celsius per minute for ten consecutive sampling points, it is considered a gradual temperature rise before the outage. If the temperature is still rising rapidly before the sudden drop, it indicates that the load itself has not yet reached a stable state and is not suitable for calculating a cooling pattern purely driven by environmental factors. If a branch experiences multiple sudden drops during the observation period, each drop should be entered into the outage sample independently, without being merged due to short intervals between drops. If the temperature rise rate before multiple sudden drops is below 0.05℃ per minute and the cooling curves after the drops are highly consistent, it indicates that the steady-state thermal resistance of the branch is relatively stable, and the estimation accuracy of the residual heat decay time constant is also high. If the temperature rise trends before the sudden drops differ significantly, the time constant should be estimated separately according to the season or load range, and the median should be used as the calculation benchmark. The dispersion of the temperature rise rate distribution in the outage sample can also reflect the stability of the branch's heat dissipation conditions: the more concentrated the distribution, the higher the repeatability of the thermal path parameter calibration; when the distribution is more dispersed, the causes should be analyzed in groups based on the corresponding ambient temperatures for each drop.

[0019] Power outage type labels are generated based on discrepancies in the verification patterns of the power outage nature samples. Under normal circumstances, the cooling curve after a sudden drop in the power outage nature samples should exhibit a monotonically decreasing, exponential decay pattern, gradually approaching the ambient temperature inside the chamber. If the curve shows a significant plateau or rebound during the drop, it indicates that the branch was not actually completely unloaded during the sudden drop; this could be due to heat transfer from adjacent branches or a temporary load reduction rather than a true power outage. For each cooling curve in the power outage nature samples, it is checked whether the reading stops decreasing and instead rises by more than 0.3 degrees Celsius within one hour after the sudden drop. Curves exhibiting this type of rebound are labeled as having a discrepancy in pattern, while others are labeled as having a consistent pattern. The time and magnitude of the rebound are recorded simultaneously. The rebound characteristics in the cooling curves of suspected disturbed power outage nature samples can also be verified in reverse: if adjacent branches were indeed under full load during that period, it further confirms that heat transfer is the main cause of the rebound. This cross-verification conclusion is noted in the power outage type label record for convenient subsequent special investigation of thermal coupling intensity. Based on the verification results, power outage type labels are marked as genuine power outages if the form matches, and as suspected disturbances if the form does not match. Each label includes the quantitative basis for triggering the judgment, so as to facilitate subsequent investigation of the source of thermal coupling. The two types of labels are counted separately in the power outage type label record. Subsequent steps will retrieve samples of genuine power outage labels as needed and skip samples of suspected disturbance labels. The two types of samples will not be mixed. If the suspected disturbance power outage type label appears repeatedly on the same branch, it indicates that the thermal coupling of the enclosure is too strong. A special check can be carried out to see if there are any abnormalities in the heat dissipation path of the enclosure, and the insulation or ventilation design can be improved as appropriate.

[0020] The residual heat dissipation time delay is obtained by calculating the residual heat decay time constant through the power outage type label. For only samples labeled as real power outage by the power outage type label, the cooling curve after the sudden drop is taken, the stable temperature rise (the difference between the conductor temperature and the ambient temperature inside the cabinet) before the sudden drop is taken as the initial value, the descending process of the temperature difference over time is fitted according to the exponential decay model, and the temperature difference readings at two successive moments t1 < t2 are taken from the fitted curve for calculating the time constant. The decay time constant is calculated according to the formula: τ_c=(t2-t1) / ln(δT1 / δT2), where t1 and t2 are two successive moments on the cooling curve (t1<t2), δT1 and δT2 are respectively the difference between the conductor temperature and the ambient temperature at the corresponding moments (δT1>δT2), and τ_c is the residual heat decay time constant corresponding to this power outage, which reflects the speed at which the conductor together with its surrounding insulating layer and terminal block releases residual heat to the environment. A larger τ_c indicates that the heat dissipation path of this branch is relatively unobstructed, and the heat dissipates more slowly; the magnitude of τ_c is positively correlated with the cross-sectional area of the conductor and the heat capacity of the insulating layer. A larger cross-sectional area leads to slower heat dissipation. The τ_c of a branch with good heat dissipation can be as low as 20-40 minutes, while the τ_c of a branch with blocked heat dissipation (such as dense trunking or poor ventilation) can exceed 2 hours. The residual heat dissipation time delay is calculated according to the formula: t_delay=τ_c×ln(δT0 / δT_threshold), where δT0 is the stable temperature rise before the sudden drop, δT_threshold is the residual heat dissipation judgment threshold (set to 1°C, which needs to satisfy δT0>δT_threshold), and t_delay is the time required for the temperature difference to decay below this threshold. The residual heat dissipation time delay of a branch with a small τ_c is usually within one hour, and a branch with a large τ_c may take several hours; the τ_c measured for the same branch in different seasons may vary, and the results of each residual heat dissipation time delay are recorded truthfully without averaging across seasons; for branches whose τ_c is systematically higher with seasons, the corresponding residual heat dissipation time delay in winter is shorter, and the starting point of the power outage calibration window can be relatively advanced to improve the utilization rate of effective data in winter.

[0021] The power outage calibration window is obtained by extracting a delayed heat dissipation window based on the residual heat dissipation delay. Within the given duration of the residual heat dissipation delay, the conductor temperature still carries residual traces of the load heat generated before the sudden drop. At this time, the reading deviates from the pure environmental steady state. If this reading is directly used to calculate the environmental thermal resistance, the residual load heat will also be included in the environmental heat dissipation effect, making the initial value too low. Starting from the point of the sudden drop, skipping the time corresponding to the residual heat dissipation delay, the readings of two hours thereafter are extracted as the power outage calibration window. The lower limit is that the short-cycle temperature disturbance energy in the box can be fully averaged, and the upper limit is that it does not cross the range of the next load change disturbance. The power outage calibration window records the start and end times of the window corresponding to the actual power outage and the conductor temperature readings within the window. The start time is strictly later than the time corresponding to the residual heat dissipation delay. If the duration of the same power outage is less than two hours, the power outage calibration window extracts the actually usable segment and marks the actual extracted duration. If the power outage is interrupted prematurely due to the connection of a new load, the power outage calibration window still includes this incomplete window and does not forcibly discard the existing readings. If multiple actual power outages occur on the same branch during the observation period, the corresponding power outage calibration windows for each outage are captured independently and do not overlap with each other. If the residual heat dissipation delay corresponding to a power outage exceeds the duration of the power outage itself, then that power outage will not generate a valid power outage calibration window, will be recorded as unusable and skipped, and will not affect the calibration of other outages. The start and end times and valid duration of each power outage calibration window are all included in the record, which facilitates subsequent grouping and verification of the stability of the initial environmental thermal resistance value according to seasons or different temperature conditions, identifies whether there is a systematic shift with seasonal changes, and provides a reference for differentiated calibration.

[0022] The initial value of the ambient thermal resistance is determined by analyzing the thermal resistance characteristics based on the power outage calibration window. Within the power outage calibration window, the conductor has been freed from the residual effects of load heating, and the reading should fluctuate slightly around the ambient temperature inside the enclosure. This reading provides the ambient reference temperature, and it is necessary to retrieve the conductor temperature and load current readings from before the power outage, when the load had reached a stable state, as a reference. The conductor temperature rise during the stable load phase before the sudden drop (the difference between the conductor temperature and the ambient reference temperature measured within the power outage calibration window, denoted as ΔT_ref) and the corresponding load current I are taken, and P_ref = I. 2The heat dissipation power is calculated by multiplying R_c by the cable's nameplate DC resistance value (adjusted for operating temperature) and the laying length. The initial environmental thermal resistance is calculated using the formula: R_th0 = ΔT_ref / P_ref, where R_th0 represents the increase in conductor temperature relative to the environment per unit heat dissipation power. Differences in R_th0 directly reflect the varying heat dissipation conditions of each branch; branches with limited heat dissipation will accumulate a higher steady-state temperature rise under the same load. This value only represents the heat dissipation level under the current calculation conditions and does not differentiate between seasonal variations. The initial value of environmental thermal resistance is calculated separately for each branch. Different branches installed in the same batch have different initial values ​​of environmental thermal resistance due to differences in their location in the box, wire cross-section, etc. If the R_th0 values ​​obtained from multiple power outage calibration windows for the same branch are close to each other, the arithmetic mean can be taken to improve the estimation accuracy of the initial value of environmental thermal resistance. If the difference is too large, it is necessary to check whether the sensor mounting has been displaced or whether the contact resistance has changed significantly. The standard deviation of the multiple calculation results is also recorded. The larger the standard deviation, the more sensitive the heat dissipation conditions of the branch are to changes in operating conditions.

[0023] Step S102: Based on the initial value of the environmental thermal resistance, the light-load operation period of the branch is deduced to obtain the light-load calibration window. The thermal resistance characteristics of the thermal circuit are analyzed from the light-load calibration window to generate the thermal circuit parameters of the branch. The branch thermal circuit parameters are used to determine the instantaneous most unfavorable fluctuation to determine the upper limit of the branch current carrying capacity.

[0024] Specifically, the light-load calibration window is obtained by inversely calculating the light-load operating period of the branch based on the initial value of the environmental thermal resistance. The initial value of the environmental thermal resistance R_th0 characterizes the thermal conductivity between the conductor of the branch and the surrounding environment. If the measured load current I for a certain period is substituted into P_ref=I... 2The heat dissipation power P_ref is calculated by multiplying R_c by R_th0, and then the theoretical steady-state temperature rise ΔT_predict = P_ref × R_th0 is obtained. This predicted temperature rise is compared with the measured conductor temperature rise during the same period. If the two are close, it indicates that the heat dissipation behavior of this branch conforms to the pattern described by R_th0, and no additional heat source or heat dissipation anomaly has occurred. The temperature rise residual is calculated using the formula: ΔT_res = ΔT_actual - ΔT_predict, where ΔT_actual is the measured conductor temperature rise during the same period, and ΔT_predict is the theoretical temperature rise calculated based on the initial value of the environmental thermal resistance. The closer the residual is to zero, the better the measured behavior matches the benchmark model. Periods where the load current is consistently below 30% of the branch's nameplate rated current I_N, and the fluctuation range of ΔT_res for two consecutive hours does not exceed 0.5℃, are identified. During these periods, the heat dissipation power is limited, and the temperature rise residual is more sensitive to subtle changes in the heat dissipation path, making it easier to identify deviations from the benchmark. The time periods that meet the conditions form a light load calibration window. Each window records the start and end times, the load current sequence and the temperature rise residual sequence during the period. Candidate time periods with a duration of less than two hours are not included. If the same branch has multiple light load time periods that meet the conditions, the light load calibration window records each time period separately. If the root mean square of ΔT_res remains within 0.1℃ for a long time within the light load calibration window, it indicates that the reference heat dissipation model given by the initial value of the environmental thermal resistance is always consistent. Once ΔT_res continuously shows a positive deviation of more than 0.3℃, the branch thermal parameters should be recalibrated in advance even if the load has not reached the upper limit, in order to prevent the degradation of the heat dissipation path from being ignored.

[0025] The branch thermal parameters are generated by analyzing the thermal resistance characteristics of the thermal circuit within the light load calibration window. Although the load current within the light load calibration window is not large, it is not zero. The conductor continuously heats up while the environment continuously dissipates heat. During the conductor's heating process, some heat flows continuously to the environment (corresponding to the steady-state heat dissipation characteristic R_th), while the other part is absorbed and stored by the conductor and insulation layer and then slowly released (corresponding to the thermal inertia C_th). The superposition of these two heat transfer paths drives the temperature to gradually rise. They need to be separated to calibrate R_th and C_th in the branch thermal parameters separately. The temperature rise residual verification of R_th was completed when identifying the light load calibration window. In this step, the steady-state thermal resistance inferred from the period that passed the verification is used as the value of R_th. The conductor temperature trend before and after a small step change in the load current within the light load calibration window is taken. After the step, the temperature slowly approaches the new steady state according to an exponential curve. The time constant of this curve is the thermal response time constant τ_th, which characterizes the thermal response speed required for the conductor to catch up with the new steady state after a sudden load change. It needs to be calibrated independently. The heat capacity parameter is calculated using the formula: C_th = τ_th / R_th, where R_th is the steady-state thermal resistance derived from the period without anomalies confirmed by the temperature rise residual within the light-load calibration window, and C_th is the heat capacity parameter in the branch thermal circuit parameters, reflecting the heat absorption capacity of the conductor and insulation layer. If no step event meeting the conditions is detected within the current light-load calibration window, C_th uses the previous valid calibration result and is marked as pending update; R_th and C_th for each branch are complete thermal circuit parameters describing the dynamic temperature response of that branch, and the latest batch of calibration results is updated continuously; once R_th is successfully calibrated, it replaces the initial value of the ambient thermal resistance R_th0, and R_th is used as the reference thermal resistance throughout the entire process, and R_th0 is no longer referenced separately.

[0026] In some embodiments, determining the branch current carrying capacity limit by measuring the instantaneous worst-case fluctuation using the branch thermal circuit parameters includes: tracking instantaneous temperature jumps based on the branch thermal circuit parameters to form an instantaneous fluctuation sample set; identifying fixed-amplitude repetitive jump interference features through the instantaneous fluctuation sample set to obtain an effectiveness discrimination parameter; eliminating sensor noise jumps based on the effectiveness discrimination parameter to generate effective fluctuation samples; and performing the worst-case fluctuation value measurement based on the effective fluctuation samples to determine the branch current carrying capacity limit.

[0027] Instantaneous fluctuation sample sets are formed by tracking instantaneous temperature jumps based on branch thermal parameters. The R_th and C_th values ​​provided by the branch thermal parameters jointly determine the upper limit of the branch's temperature response speed to load changes. Based on these parameters, the time required for temperature to transition from one steady state to another will not be less than a certain number of time constants. Temperature jumps occurring much shorter than this expected response time cannot be physically caused solely by normal changes in load current; therefore, they can be considered instantaneous jump events. Instantaneous jumps are recorded when the temperature difference between adjacent sampling points in the conductor temperature signal exceeds the maximum reasonable temperature rise per unit time calculated based on the branch thermal parameters. These jumps may correspond to short-term impact loads or simply sensor or signal link noise. The judgment threshold is set separately for each branch based on its own branch thermal parameters. Branches with larger heat capacity and stronger thermal inertia have correspondingly lower allowable temperature rise thresholds per unit time. Each branch's threshold value is independent and cannot be shared across branches. Each detected jump event is recorded individually, with each event listed separately, and the time sequence and temperature information are fully preserved. The density of jump events in the same branch during periods of high load activity is significantly higher than during periods of calm. The density distribution of jump events itself is also a reflection of the load behavior characteristics of the branch. If the instantaneous fluctuation sample set of a branch is empty for a long period of time, it indicates that there has been no significant short-term impact in the history of the branch. The upper limit of the branch current carrying capacity can be directly determined by the steady-state temperature rise constraint, skipping the instantaneous fluctuation sample set statistical step to improve analysis efficiency and avoid meaningless interference feature identification steps that consume computational resources.

[0028] Effectiveness criteria parameters are obtained by identifying the characteristics of repetitive jump interference with fixed amplitude using a set of instantaneous fluctuation samples. In the instantaneous fluctuation sample set, the jump amplitudes of several samples are highly similar, and the time intervals between jumps also exhibit a similar periodicity. This type of pattern originates from fixed-frequency electrical interference coupled through the sensor signal line. Inverter carrier leakage or switching power supply ripple is injected into the sampling link at a fixed period. The interference amplitude is determined by the characteristics of the interference source itself and is independent of the actual load current of the branch, thus exhibiting a fixed amplitude and repetitive occurrence at an approximate period. The distribution of amplitude jumps in a set of instantaneous fluctuation samples is statistically analyzed. Samples whose amplitudes fall within a narrow range and whose coefficient of variation for the interval between occurrences is below a set threshold are identified. The amplitude range and interval period corresponding to this batch of samples are recorded as a set of interference features. The validity discrimination parameters incorporate the extracted interference features, including the upper and lower bounds of the interference amplitude range and the corresponding repetition period range, to determine whether a newly occurring jump falls within the known interference feature range. Each set of interference features operates independently; adding or removing features only requires appending or removing the corresponding record, without affecting each other. After the validity discrimination parameters are updated, samples from the instantaneous fluctuation sample set previously judged as noise can be re-included, while previously retained samples that fall within the newly added feature range are reclassified and removed, achieving bidirectional updating. Taking a certain enclosure as an example, when the model of nearby frequency converter equipment changes, the interference amplitude range shifts accordingly, and the old version of the validity discrimination parameters becomes invalid. It needs to be updated based on the amplitude distribution statistically analyzed from the new batch of instantaneous fluctuation sample sets. After the update, jump samples that were originally considered real fluctuations but fall into the new interference range will be reclassified, ensuring that the validity fluctuation samples always maintain high purity.

[0029] Based on the validity discrimination parameters, sensor noise jumps are eliminated to generate valid fluctuation samples. For each jump sample in the instantaneous fluctuation sample set, its jump amplitude and occurrence interval are compared to whether they fall within any set of interference characteristics given by the validity discrimination parameters. Samples falling within the range are judged as sensor noise jumps and will not be included in subsequent analysis; samples not falling within any known interference characteristic range are judged as samples that may reflect real load fluctuations and are retained. Only when both amplitude and interval fall within the same set of interference characteristics are they judged as noise jumps. Samples with only one matching are still included in the valid fluctuation samples to avoid misjudging and eliminating real load impact events. Samples judged as noise jumps are archived separately and labeled with the interference characteristic number that triggered the judgment. If the validity discrimination parameters subsequently remove this set of interference characteristics, this batch of samples will be reactivated and included in the valid fluctuation samples. The jump samples retained through two rounds of screening form valid fluctuation samples. Each sample still retains the original jump amplitude, occurrence time, and temperature readings before and after the jump. Valid fluctuation samples for the same branch accumulate continuously over the observation period. The larger the sample size, the more stable and reliable the estimation result of the most unfavorable fluctuation. When calibrating the branch current carrying capacity limit, the most unfavorable jump amplitude is directly read from the valid fluctuation samples without re-querying the original instantaneous fluctuation sample set, ensuring the unidirectional and traceable data transfer between the two steps. After the valid fluctuation samples accumulate to a certain number, the statistical result of the most unfavorable fluctuation tends to stabilize, and the corresponding branch current carrying capacity limit estimate converges accordingly. After each round of validity discrimination parameter update, archived noise samples that no longer match due to the removal of interference features are automatically moved back to the valid fluctuation samples. If the number of valid fluctuation samples is consistently too small, the maximum value of the most unfavorable fluctuation is used from the existing samples until a new measured value appears.

[0030] The branch current carrying capacity limit is determined by measuring the most unfavorable fluctuation value based on the effective fluctuation sample. Each real load fluctuation in the effective fluctuation sample will push the conductor temperature to an instantaneous peak value in a short period of time. Although this peak value is short-lived and may not reach the level of steady-state temperature rise, the thermal breakage reaction of the insulation molecular chain caused by the short-term high temperature is cumulative. Each impact leaves irreversible chemical damage to the insulation layer. After multiple superpositions, the insulation withstand voltage capability degrades rapidly. After multiple high-temperature impacts, the dielectric loss angle of the insulation layer gradually increases and the power frequency withstand voltage decreases. Even if the steady-state load never exceeds the rated value, the insulation may still deteriorate prematurely. This is the physical logic behind taking the historical most unfavorable fluctuation amplitude rather than the average amplitude as the basis for setting the branch current carrying capacity limit. Therefore, the setting of the branch current carrying capacity limit cannot only consider whether the steady-state temperature rise meets the standard, but also needs to include the cumulative damage caused by such instantaneous fluctuations in the constraints. The effective fluctuation samples are compared and ranked from highest to lowest according to their jump amplitude. The samples with the highest values ​​are selected as the most unfavorable fluctuation scenarios historically experienced by that branch. This most unfavorable fluctuation amplitude is then superimposed on the temperature rise corresponding to the steady-state load to obtain the equivalent peak temperature rise considering instantaneous fluctuations. The branch current carrying capacity limit is calculated by reverse engineering based on the condition that the equivalent peak temperature rise does not exceed the maximum allowable continuous operating temperature T_max of the conductor insulation material (90℃ for XLPE insulation and 70℃ for PVC insulation). If the steady-state load has already occupied most of the temperature rise space, the margin left for instantaneous fluctuations is narrowed, and the branch current carrying capacity limit is adjusted accordingly. Taking XLPE insulated cable as an example, T_max = 90℃. If the current ambient temperature inside the box is 35℃, the total temperature rise budget is 55℃. If the steady-state load has consumed 45℃ of this, the allowable superimposed instantaneous fluctuation peak temperature rise is only 10℃, and the equivalent branch current carrying capacity limit will therefore be significantly lower than the nameplate rating.

[0031] Step S103: Determine the cross thermal coupling coefficient by converting the full-load temperature rise of the branch current limit, analyze the contribution of the adjacent full-load temperature rise based on the cross thermal coupling coefficient to generate the margin correction coefficient, and use the margin correction coefficient to reduce the branch current limit to obtain the branch margin estimate.

[0032] In some embodiments, determining the cross-thermal coupling coefficient by performing full-load temperature rise conversion based on the upper limit of the branch current carrying capacity includes: generating an equivalent temperature rise sample by performing full-load temperature rise conversion processing on the upper limit of the branch current carrying capacity; analyzing the proportion of the metal shell conduction and the air convection path in the box to determine the coupling path type based on the equivalent temperature rise sample; analyzing the long-term non-attenuation characteristics of the coupling strength based on the coupling path type to obtain the coupling strength classification; and determining the cross-thermal coupling coefficient by calibrating the coefficients based on the coupling strength classification.

[0033] For example, the step of generating an equivalent temperature rise sample by performing full-load temperature rise conversion processing through the branch current carrying limit includes: using the branch current carrying limit to eliminate the influence of current carrying difference to obtain a current carrying normalization factor; performing a top branch heat dissipation anomaly verification based on the current carrying normalization factor to obtain a height heat dissipation characteristic value; analyzing the upper and lower layer temperature rise gradient distribution based on the height heat dissipation characteristic value to obtain a spatial gradient parameter; and performing gradient compensation according to the vertical position based on the spatial gradient parameter to generate an equivalent temperature rise sample.

[0034] The current-carrying normalization factor is obtained by eliminating the influence of current-carrying differences using the branch current-carrying upper limit. The branch current-carrying upper limit varies due to the different thermal parameters of each branch. If the measured temperature rise data of each branch are directly compared for cross-thermal coupling, the temperature rise difference caused by current-carrying differences will be mixed with the temperature rise differences caused by location and coupling path, making it impossible to isolate the coupling effect. Therefore, it is necessary to first convert the measured temperature rise to a unified full-load reference before comparison. The current-carrying normalization factor is calculated using the formula: η=(I_max / I). 2 Where I_max is the upper limit of the current carrying capacity of the branch, I is the current measured load current, and η is the dimensionless normalization coefficient; the heating power is proportional to the square of the current. During the light load period, I is much lower than I_max, so η is obviously greater than one. Multiplying it by the measured temperature rise ΔT_actual at that moment gives the full-load equivalent temperature rise ΔT_equiv=η×ΔT_actual. The current-carrying normalization factor is calculated separately for each measured data point of each branch. The load current of the same branch varies at different times, and the corresponding current-carrying normalization factor also varies accordingly. The normalization calculation results are not suitable for direct comparison across branches because the systematic differences caused by the vertical position have not yet been eliminated. It is necessary to wait until the subsequent gradient compensation step is completed before the equivalent temperature rise samples from different branches and different layers can be placed in the same coordinate system for coupling strength comparison, so as to ensure that the comparison results only reflect the thermal coupling between branches and not other interference factors. The equivalent temperature rise samples exclude extreme light load data with I below 0.1×I_max, because the η in this range is too large, which causes the conversion noise to be severely amplified and is not suitable for cross-branch coupling comparison. The specific value of the current-carrying normalization factor η will change with the load level of the branch at that time. When lightly loaded, η is much greater than one and the conversion ratio is large. When fully loaded, η approaches one and the conversion ratio is small. Therefore, it is required that the equivalent temperature rise samples have a sufficient number of samples at each load level.

[0035] The heat dissipation characteristic value of the top branch was obtained by verifying the heat dissipation anomaly of the top branch based on the current-carrying normalization factor. Multiple branches in the branch box are arranged vertically in layers. Heated air in the box flows upwards. Even if the load of the upper branch is the same as that of the lower branch, it is constantly surrounded by warm air rising from below, resulting in naturally weaker heat dissipation. If this disadvantage due to its location is not identified, it can easily be misjudged as an abnormality in the branch's thermal characteristics, leading to an incorrect downward adjustment of the branch's current-carrying capacity. The equivalent full-load temperature rise of each branch, calculated using the current-carrying normalization factor, was taken and arranged from bottom to top according to the branch's location. The system was checked whether the calculated temperature rise of the topmost branches was systematically higher than that of the middle and lower branches, and whether this higher value was consistently observed in multiple observations. If there was no systematic difference in the calculated temperature rise between layers, it indicates that the airflow channels in the box are unobstructed, and the heat dissipation characteristic value is approximately equal across layers, requiring no additional positional correction. The height heat dissipation characteristic value quantifies the average height difference of each layer relative to the bottom layer into a set of values, with the bottom layer as the zero base and increasing layer by layer, reflecting the strength of thermal stratification inside the cabinet. If active cooling measures such as top cooling fans are added to the same branch cabinet, the difference between layers in the height heat dissipation characteristic value will be significantly narrowed. This difference itself can reflect whether the cabinet's heat dissipation design is appropriate. After the internal installation structure of the cabinet is adjusted, the height heat dissipation characteristic value needs to be re-collected to update the parameter basis used for subsequent gradient compensation. If the difference between layers in the height heat dissipation characteristic value of the same branch cabinet varies greatly in different seasons, a set of parameters needs to be maintained for each season. The same set of parameters cannot be applied all year round to avoid introducing seasonal errors.

[0036] Spatial gradient parameters are obtained by analyzing the temperature rise gradient distribution between upper and lower layers based on the height heat dissipation characteristic values. The height heat dissipation characteristic values ​​only provide an isolated set of values ​​representing the average deviation of each layer relative to the bottom layer; a continuous curve that can be interpolated for any layer is not yet formed. If the number of layers in the enclosure is large, recording the deviation of each layer individually will lack a unified conversion basis for subsequent layer-by-layer compensation. The height heat dissipation characteristic values ​​are arranged in layer order, and a distribution curve of the deviation variation with layer position is fitted. In enclosures with fewer layers and unobstructed airflow channels, the curve is close to a straight line; in enclosures with more layers and where airflow tends to stagnate in the upper part, a steeper rise often occurs in the top one or two layers. During fitting, if the height heat dissipation characteristic value of a certain layer deviates significantly from the overall trend, it is necessary to check whether there is heat dissipation obstruction or installation abnormality in that layer. Only after confirmation can it be included in the fitting data. After fitting, it should be checked whether the deviation of the curve between layers increases monotonically. If a non-monotonic anomaly occurs, it indicates that the sampling of the height heat dissipation characteristic value of that layer is not representative, and the measured data of that layer needs to be supplemented for refitting. After fitting, the deviation of each layer is fixed as a spatial gradient parameter. The compensation amount corresponding to the layer can be obtained by looking up the table according to the layer position. After the internal structure of the box is adjusted (such as adding partitions or replacing cooling fans), the height heat dissipation characteristic value needs to be re-collected, and the corresponding spatial gradient parameter is also updated accordingly. If there is a significant difference in the height heat dissipation characteristic value between summer and winter, the spatial gradient parameter should be saved separately for each season for retrieval and use in the corresponding season. The same set of parameters should not be applied all year round to introduce seasonal systematic errors. It should be updated in a timely manner at the seasonal transition point.

[0037] Equivalent temperature rise samples are generated based on gradient compensation according to vertical position using spatial gradient parameters. For each branch, the full-load equivalent temperature rise is calculated using the current-carrying normalization factor. The corresponding higher amplitude h_layer is extracted from the spatial gradient parameters according to its stratum, and the compensated temperature rise is calculated according to the formula: ΔT_comp = ΔT_equiv - h_layer, where ΔT_comp is the equivalent temperature rise sample value after removing stratum factors. Theoretically, the equivalent temperature rise sample only reflects the thermal characteristics of the branch itself and the coupling status with its neighboring branches, and no longer includes systematic differences caused by different strata. After normalization and compensation, the temperature rise values ​​from different branches and different layers are restored to the baseline state of full load and no location difference. The temperature difference that still exists under this baseline can be attributed to thermal coupling between branches rather than differences in installation layer or load level. If the equivalent temperature rise sample of the branch located on the top layer with a large upward deviation given by the spatial gradient parameter is still significantly higher after compensation, it means that there is an additional heat source or coupling contribution outside the layer of the branch, which needs to be investigated separately. The accuracy of gradient compensation directly affects the reliability of subsequent thermal coupling analysis: if the spatial gradient parameter underestimates the difference in heat dissipation between layers, the equivalent temperature rise sample of the top branch will be systematically higher, and the Pearson correlation coefficient with the adjacent branch will be artificially high, and the cross thermal coupling coefficient will be overcalibrated; if the difference between layers is overestimated, the overcompensation will cause the equivalent temperature rise sample of the top layer to be systematically lower, and the coupling strength will be underestimated. Both types of deviations will transmit systematic errors to the margin correction coefficient. After gradient compensation is completed, it is necessary to verify whether the mean of the equivalent temperature rise samples of each layer has eliminated the gradient residuals that are monotonically correlated with the layer. If the residuals still exist, it is indicated that the spatial gradient parameters need to be re-acquired and calibrated.

[0038] The coupling path type was determined by analyzing the proportion of heat conduction through the metal shell and air convection within the chamber in the equivalent temperature rise sample. Heat transfer between two adjacent branches mainly occurs through two paths: one is direct conduction through shared metal mounting rails, busbars, and other structures, and the other is heat transfer from the periphery of one branch to the periphery of the other branch through air convection within the chamber. Take a sample of the equivalent temperature rise of a branch circuit during the period when there is a significant increase. At the same time, check the operation status of the fan in the box (provided by the power distribution automation operation log or fan auxiliary switch signal) and the opening and closing status of the box door (provided by the access control sensor or maintenance personnel inspection record) during the same period. The operation of the fan or the opening and closing of the box door will significantly change the strength of the convection path. If the increase changes significantly with the start and stop of the fan or the opening and closing of the box door, it indicates that the heat transfer of the branch circuit is dominated by the convection path. If the increase remains basically unchanged before and after the start and stop of the fan, it indicates that the metal conduction path is dominant. If the branch box has neither a fan nor an access control signal, take multiple periods of increase and compare them with the ambient wind speed of the nearest weather station. When the wind speed is high, the increase narrows and is judged to be dominated by convection. When the increase is stable, it is judged to be dominated by conduction. The coupling path type is determined for each pair of adjacent branches and labeled as conduction-dominated, convection-dominated, or mixed. The determination of the coupling path type affects the subsequent judgment on whether the coupling strength is stable in the long term. Conduction paths depend on fixed physical structures and are usually more stable than convection paths that depend on airflow conditions. This judgment provides a preliminary directional basis for identifying the long-term non-attenuation characteristics of the equivalent temperature rise sample. Once the coupling path type is determined, it usually remains stable for a long period of time. It only needs to be re-determined when the ventilation structure of the box is physically changed (such as adding or removing fans or partitions). The status of the fan and the box door is recorded simultaneously for each equivalent temperature rise sample.

[0039] The coupling strength is graded based on the long-term non-attenuation characteristics of coupling strength analysis according to coupling path type. For branches identified as conduction-dominated, the coupling strength mainly depends on the fixed metal structure, and the coupling strength usually remains stable over extended observation periods. For branches identified as convection-dominated, the coupling strength is easily affected by factors such as the frequency of door opening and closing and seasonal ventilation differences, and may fluctuate and be unstable over extended observation periods. For each branch with identified coupling path type, the Pearson correlation coefficient r of its equivalent temperature rise sample is calculated across different observation batches. Branches with r consistently above 0.6 for multiple observation periods without a significant decreasing trend are identified as long-term non-attenuating couplings. Branches with r continuously decreasing over the observation period are identified as attenuating couplings, which may be due to the narrowing of convection channels with the seasons or a decrease in thermal conductivity of the metal structure due to oxidation. The coupling strength classification categorizes long-term non-attenuating branch pairs into three levels based on their r-values: strong (r≥0.8), medium (0.6≤r<0.8), and weak (0.5≤r<0.6). Attenuating coupling branch pairs are classified separately, and their correlation decreasing over time is indicated. The thresholds used for coupling strength classification are set according to the number of observation batches. When the number of batches is small, the thresholds are appropriately widened to avoid a large number of branch pairs being misclassified as attenuating due to insufficient samples. The coupling strength classification results are saved in seasonal groups to track whether the coupling characteristics exhibit regular seasonal changes. If there is a significant difference in the coupling strength classification results for the same branch pair between summer and winter, it indicates that the path between the branches is mainly convective, and subsequent coefficient calibration should provide k values ​​for each season.

[0040] The cross-thermal coupling coefficient is determined by calibrating the coefficients through coupling strength classification. For branch pairs marked as strong, medium, and weak in the coupling strength classification, the time period in which the equivalent temperature rise sample of the source branch shows a significant increase is selected. The increase amplitude of the equivalent temperature rise sample of the adjacent branch in the same time period is compared. Data from multiple increase periods are calculated separately and then the arithmetic mean is taken to reduce the impact of single occasional fluctuations on the calibration results. The cross-thermal coupling coefficient is calculated using the formula: k_ij = mean(ΔT_j,induced / ΔT_i,equiv), where ΔT_i,equiv is the equivalent temperature rise of source branch i, ΔT_j,induced is the temperature rise of adjacent branch j within the same time period, and k_ij represents the heat transfer efficiency between the two branches. Its magnitude directly determines the contribution of the source branch to the temperature rise of the adjacent branch when it is fully loaded, and is reflected in the product form in the margin correction coefficient calculation. The value of k_ij is closely related to the distance between the two branches, the cross-sectional area of ​​the metal connectors, and the airflow state inside the box. The closer the distance, the more complete the metal connection, and the weaker the airflow, the higher k_ij is usually. The difference in k_ij between each branch indirectly reflects the uniformity of the box installation layout and heat dissipation design. Cross-thermal coupling coefficients are calibrated separately for each branch pair. The k-values ​​of different adjacent branches of the same branch as a source branch pair are usually not the same. The k-values ​​of branches on both sides are often higher than those of branches separated by more than one road. For branch pairs classified as attenuating coupling in the coupling strength classification, the cross-thermal coupling coefficient is taken as the weighted average of the most recent batches of observations (the recent batches have higher weights). At the same time, the attenuation trend is marked in the record to prompt the dispatching side to evaluate whether the branch pair can be reclassified. Attenuating coupling is more common in convection-dominated branch pairs. After the ventilation conditions are improved, convection is enhanced and k_ij gradually weakens. For conduction-dominated branch pairs, because the physical structure remains unchanged, k_ij usually remains stable in the long term. The difference between the two types of trends can verify the determination of the coupling path type.

[0041] The margin correction coefficient is generated based on the analysis of the adjacent full-load temperature rise contribution using the cross-thermal coupling coefficient. The calibrated cross-thermal coupling coefficient k_ij is substituted into the calculation for the most unfavorable operating condition: assuming all adjacent branches operate simultaneously at their respective branch current limits, this condition is not uncommon when the terminal loads rise synchronously during high-temperature seasons. The sum of the adjacent branch full-load temperature rise contributions is calculated using the formula: ΔT_neighbor = Σ jk_ij×ΔT_j,equiv (j iterates through all adjacent branches of this branch), where the cross-thermal coupling coefficient k_ij is the heat transfer efficiency of adjacent branch j to this branch i, ΔT_j,equiv is the equivalent temperature rise of branch j under full load, and ΔT_neighbor is the total additional temperature rise borne by this branch under the condition that all adjacent branches are fully loaded; the ratio of ΔT_neighbor to the temperature rise of this branch itself reflects the degree of thermal coupling 'parasitic' interference of this branch. The higher the ratio, the more severely the available temperature rise space of this branch is squeezed by adjacent branches. The allowable temperature rise budget of this branch itself is equal to the difference between the maximum continuous operating temperature T_max allowed by the conductor insulation material (90℃ for XLPE insulation and 70℃ for PVC insulation) and the current ambient temperature, denoted as ΔT_total; the equivalent temperature rise of itself under full load is denoted as ΔT_self, and the room for adjacent branches to contribute temperature rise is ΔT_budget=ΔT_total-ΔT_self. The margin correction factor is calculated using the formula: γ = (ΔT_budget - ΔT_neighbor) / ΔT_total, where the denominator is the total allowable temperature rise ΔT_total. γ represents the proportion of the remaining usable temperature rise space of this branch after the adjacent branch is fully loaded. When γ is close to a certain value, the thermal contribution of the adjacent branch is negligible. The lower the γ, the stronger the thermal coupling effect of the adjacent branch. When the simultaneous full load of the adjacent branch is the normal operating condition, the low γ characteristic of this branch should be listed as a key monitoring target. When γ is too small or even negative, the full load of the adjacent branch has exhausted the temperature rise budget of this branch, and this branch can hardly operate at full load independently anymore. The margin correction factor γ is a quantitative characterization of the degree to which the thermal coupling effect of the entire adjacent branch squeezes the usable temperature rise space of this branch. It is updated in conjunction with the change of the current carrying capacity limit of any adjacent branch, ensuring that the margin estimate always reflects the current actual thermal load level of the adjacent branch.

[0042] The branch current carrying capacity estimate is obtained by subtracting the branch current carrying capacity limit from the margin correction factor. The branch current carrying capacity limit is the current carrying capacity obtained by considering only the branch's own thermal characteristics and instantaneous fluctuations, without considering the influence of neighboring branches. In actual operation, the temperature rise caused by the full load of adjacent branches occupies part of the temperature rise budget. Therefore, the current that the branch can safely carry is lower than when calculated separately. The margin correction factor γ is used to describe the proportion of this budget that is occupied. The heating power is proportional to the square of the current. The temperature rise budget is reduced proportionally by γ, which means that the allowable current can only be reduced by γ. This ratio is reduced accordingly. The estimated branch margin is calculated using the formula: Where I_max is the upper limit of branch current carrying capacity, γ is the margin correction coefficient, and I_margin is the estimated margin of branch current carrying capacity, which reflects the upper limit of current that the branch can safely carry after taking into account the most unfavorable full load situation of the adjacent branch. The difference between the estimated margin of branch current carrying capacity and the upper limit of branch current carrying capacity is equal to the portion of current carrying capacity that thermal coupling deprives from the temperature rise budget of this branch. The larger the difference, the stronger the dependence of this branch on the adjacent branch to maintain a low load and the narrower its independent scheduling space. The branch margin estimate is linked in real time with each update of the current carrying capacity limit of adjacent branches: when the current carrying capacity limit of the adjacent branch is lowered, the branch margin estimate of this branch increases; when the current carrying capacity limit of the adjacent branch is raised, the branch margin estimate of this branch narrows. If a branch has strong thermal coupling with multiple adjacent branches at the same time, its branch margin estimate is affected by the linkage of multiple branches. Under the most unfavorable working condition where all adjacent branches are fully loaded, the branch margin estimate may be much lower than the branch current carrying capacity limit. In this case, current limiting protection should be given priority to this branch. When γ<0, the branch margin estimate is zero and an alarm is triggered: under the working condition where the adjacent branches are fully loaded at the same time, the temperature rise of this branch has exceeded T_max by the heat conduction of the adjacent branches alone. It is necessary to evaluate the heat dissipation modification or add heat insulation measures from the overall branch box level, rather than simply treating it as zero margin.

[0043] Step S104: Based on the branch load margin estimate, track the step transfer of branch load to form a transfer load list. Screen the transfer load list to obtain measured margin samples of the transfer unrestricted sections. Perform successive transfer temperature fitting on the measured margin samples to form a temperature margin curve.

[0044] Specifically, the branch margin estimate is used to track the step transfer of branch loads to form a list of transferred loads. The branch margin estimate provides the theoretical current carrying capacity calculated by the model. The best way to verify whether this capacity is real and reliable is to see how the branch performs when it experiences a real high load impact in actual operation. The most common high load impact scenario in the distribution network is load transfer, that is, a branch needs to be taken out of operation due to maintenance or failure, and the load it originally carried is temporarily switched to an adjacent branch. The current of the branch receiving the load will show a significant step increase in a short period of time. Retrieve the load current sequence of each branch and find the moment when the current suddenly rises between two adjacent samplings and the current level after the rise remains at least 3 times the thermal circuit time constant τ_th for a duration (rather than a momentary drop). τ_th is the most recent valid calibration value. If τ_th is marked as pending and more than a quarter has passed since the last valid calibration, the median of τ_th of all branches in the branch box where the branch is located is temporarily used as the replacement. At the same time, the recalibration process of the thermal capacity parameter C_th of the branch is triggered. This trend is different from short-term impact load and is consistent with the characteristics of long-term load increase after the transfer of power supply. Detected step transfer events are recorded one by one in the load transfer list. Each entry retains the start and end times of the transfer (t_start and t_end, used to calculate Δt and the rate of change v), the current levels before and after the transfer, and the estimated branch margin at that time. The same branch may experience multiple transfers during the observation period, and each transfer is recorded separately. The load transfer list marks the direction of events for load increase and load decrease. The subsequent analysis focuses on events in the direction of load increase, while events in the direction of load decrease are kept for future reference. The records in the load transfer list are arranged in chronological order to facilitate subsequent steps in filtering the corresponding branch margin estimate version by time period, ensuring that the latest model estimate is used for each event.

[0045] In some embodiments, the step of obtaining measured margin samples by screening non-over-limit sections of the transfer load list includes: generating high-load candidate samples by verifying the transfer load ratio intervals of the transfer load list; obtaining an effective sample set by verifying the sample validity and over-limit status using the high-load candidate samples; forming a filtered sample set by parsing the transfer-out path restrictions based on the effective sample set; and obtaining measured margin samples by performing remaining available margin calculation based on the filtered sample set.

[0046] For example, the step of generating high-load candidate samples for verifying the proportion of transferred load in the transferred load list includes: obtaining load jump samples by tracking the load change rate before and after the transfer based on the transferred load list; forming a high-load screening threshold by statistically analyzing the proportion distribution of jump amplitude using the load jump samples; obtaining composite operating condition labels by screening the proportion superimposed over-limit sections through the high-load screening threshold; and generating high-load candidate samples by setting the composite operating condition labels as superimposed load priority classification.

[0047] Load jump samples are obtained by tracking the load change rate before and after the transfer based on the load transfer list. For each step transfer event recorded in the load transfer list, the time Δt during which the current rises from the pre-transfer level to the post-transfer level is not entirely consistent. Some transfers result in a one-time rise to the current level immediately after the switching operation, while others show a stepped rise due to the load being connected in batches. This time reflects the transfer operation method and load characteristics and needs to be extracted separately for subsequent judgment on whether the jump is clean and decisive. The load change rate is calculated using the formula: v = ΔI / Δt, where ΔI is the difference in current readings before and after the transfer recorded in the load transfer list, Δt is the rise time from t_start to t_end, and v is the load change rate. The load jump sample includes a change rate v exceeding v_min = 0.05 × I_N·min. -1 For load transfer events, events with a rate lower than v_min are considered natural load increases rather than load transfer switching and have no reference value for margin verification. If the interval between two load transfers is very short and both point to the same receiving branch, and the second one is superimposed before the previous current has fully stabilized, the load jump sample will treat such superimposed events as a composite jump, and the jump amplitude will be calculated as the sum of the two increases. Multiple load transfers occurring within the observation period of the same branch will be recorded separately in the load jump sample, and will be grouped by branch in subsequent statistics. Load jump samples from different branches will not be merged. The load jump sample will continue to increase as the load transfer list continues to accumulate. When the number of load jump samples of a certain branch reaches the statistical threshold, it can be used for the calculation of the subsequent high load screening threshold. If the threshold is not reached, the average level of all load transfer events will be used as a temporary reference.

[0048] High load screening thresholds are formed by statistically analyzing the distribution of the percentage of load jump amplitudes in load jump samples. The magnitude of each jump in the load jump samples varies. Some power transfers only increase the rated current of the branch by about 10%, while others can cause the current to jump to 70% or 80% of the rated current. If all jump samples are included as high load candidates in the subsequent analysis without distinction, a large number of power transfers with slight amplitudes will drown out the truly valuable high load events. The current level after each load jump in the load change samples is converted into a proportion of the rated current of that branch. The distribution of all samples on this proportion is statistically analyzed, and the lower limit of the proportion corresponding to the samples in the upper quarter of the distribution is retrieved. This lower limit is set as the high load screening threshold. The high load screening threshold is statistically analyzed separately for each branch, and is re-analyzed every 20 new load jump samples or every quarter (whichever comes first). When the load characteristics of the area served by the branch undergo structural changes, the high load screening threshold will be adjusted accordingly with the statistical results of the new batch of load jump samples. If no new power transfer events occur on a branch for a long period of time, the high load screening threshold will continue to use the previous round. The statistical results will be updated on a rolling basis once new samples appear. For branches with fewer than 20 total jump samples during the observation period, the high-load screening threshold will be the average of the current proportions after all existing jump events as a temporary threshold. Once the number of load jump samples reaches 20, the statistics will be recalculated using the quarter-weighted method. If there are no new samples that meet the conditions when the high-load screening threshold statistical period expires, the results of the previous period will be continued directly without forcibly triggering a recalculation. The high-load screening threshold reflects the actual high load level of each branch. Different branches have different thresholds due to the different load characteristics of the areas they serve. During the statistics, the thresholds will be strictly grouped by branch and will not be used across branches.

[0049] Composite operating condition labels are obtained by screening the proportion of overloaded sections using a high-load screening threshold. Branches connected to power transfer services often already bear their original load before the transfer. The current increase brought about by the transfer is superimposed on the original load. Simply checking whether the absolute current level after the transfer exceeds the high-load screening threshold can easily overlook situations where the original load proportion was already high, and the transfer only adds a small portion, causing the total proportion to suddenly exceed the limit. For each transfer event, the original load proportion before the transfer and the newly added load proportion after the transfer are taken and added together to obtain the total proportion. Events where the total proportion exceeds the high-load screening threshold, but the newly added proportion brought about by the transfer itself is not high on its own, are identified as superimposed overload events; events that individually exceed the high-load screening threshold are treated as simple high-load events, and the two categories are statistically analyzed separately. Even if the final load percentages of the two types of events are similar, their verification value is not equivalent: the former, where the conductor has already accumulated a high thermal background temperature due to continuous operation of the original load, is superimposed with a power transfer impact, verifying the superimposed bearing capacity under a sustained high thermal background; the latter, where a large amount of power is directly transferred from a lower initial temperature state, verifies the instantaneous bearing capacity of an impact load under a low thermal background. The thermal stress accumulation trajectory experienced by the insulation material is different in the two cases. The composite operating condition label is used to mark superimposed overload events, distinguishing them from simple high-load events, for subsequent grouping and prioritization. After the superimposed overload verification is completed, whether or not the event has a composite operating condition label is written into the high-load candidate sample as a fixed attribute. After this step, the two types of events are distinguished by labels and are no longer merged.

[0050] Composite load condition labels are set to prioritize the generation of high-load candidate samples based on superimposed loads. Transfer events tagged with composite load conditions reflect the actual bearing capacity of branch margins under the combined effect of the original load and the transferred load. This type of performance most closely resembles the most unfavorable superimposed scenarios that may occur in actual operation, and is more valuable than samples where the original load is very light and the transferred load surges alone. Therefore, it should be prioritized in subsequent analysis. All candidate events that pass the high-load screening threshold are divided into two groups based on whether they have composite load condition labels. The tagged group is prioritized for inclusion, while the untagged group is added as a supplement. Within each group, events are further ranked from highest to lowest based on their total percentage after superposition, with events having a higher percentage appearing earlier within the group. High-load candidate samples are collected into two groups of events according to the above priority order, and each load percentage item and composite condition label is saved, along with a record index. The original time series can be retrieved by using this index during the validity verification process. If the high-load candidate samples with composite condition labels are sufficient to cover all load intervals, unlabeled events are still included to supplement the high-percentage single-impact intervals that cannot be covered by simply stacking over-limit samples. Specifically, if the composite condition high-load events have covered the 0.7-0.9 percentage interval but there are no samples in the 0.6-0.7 interval, then the single-impact data in that interval are preferentially supplemented from the unlabeled events. The priority-based design of high-load candidate samples ensures that each round of margin verification can make full use of the most representative high-load events, so that the accumulated high-load measured information is not diluted or covered by a large number of subsequent lightly transferred samples. After the priority is arranged, the high-load candidate samples at the top of the queue are given priority to enter the validity verification.

[0051] The validity of the samples and the occurrence of exceedances are verified using high-load candidate samples to obtain a valid sample set. Some events in the high-load candidate samples may be limited by sensor failures, communication interruptions, etc., resulting in missing or significantly abnormal jumps in temperature or current readings before and after the power transfer, which can be retrieved through the record index. If such data quality issues are not investigated in advance, they will directly contaminate the margin calculation results derived from these samples. For each event in the high-load candidate samples, the continuity and completeness of temperature and current readings before and after the power transfer are checked through the record index. Events with a missing proportion exceeding the set upper limit are judged as incomplete data and excluded from subsequent analysis. After the data integrity check is completed, it is further checked whether the conductor temperature during the power transfer exceeds the maximum allowable operating temperature T_max of the insulation material. Events exceeding the limit are separately categorized and marked. Such events prove that the previous branch margin estimate is too optimistic under this condition, but cannot be included in the valid sample set as normal samples for measuring the remaining available margin. The effective sample set includes events with complete data that do not exceed the limits. Events exceeding the limits and events with incomplete data are classified into two separate checklists. The checklists for events exceeding the limits periodically trigger a review of the accuracy of the branch margin estimation model to ensure that information exceeding the limits is not completely discarded due to being excluded from the effective sample set. If a large number of events in a batch of highly loaded candidate samples are excluded due to data quality issues, resulting in a small effective sample set, appropriate conservatism must be exercised in subsequent discretionary calculations to avoid using a small number of extreme events as the sole basis for representative conclusions. The effective sample set that passes the validity verification is of reliable quality and can be directly called upon in subsequent discretionary calculation steps without the need to repeatedly verify the original data.

[0052] Based on the analysis of the effective sample set, a filtered sample set is formed by considering the path limitations on the sending side. For the transfer events included in the effective sample set, the actual amount of new load borne by the receiving branch is not entirely determined by the capacity of the receiving branch itself. The path capacity and switching capacity of the sending side (the branch or line that originally carried this load and now needs to transfer it out) also limit the upper limit of the load that can be transferred out. If the sending side itself has limited capacity and can only transfer out a small portion of the load, the current observed by the receiving branch after the transfer, even if it doesn't appear very high, may not truly reflect the upper limit of the receiving branch's own margin, because the sending side simply doesn't have any more load to transfer. For each event in the valid sample set, the load level of the corresponding line or branch on the sending side before the transfer occurred and its own capacity limit are checked to calculate the remaining transferable load space on the sending side at that time. If this space is less than both the estimated branch margin of the receiving branch and 50% of the rated current I_N of the branch, it is determined that the sending side is constrained and excluded from the filtered sample set. If it is less than the former but not less than the latter, it is retained and marked that the constraints on the sending side may not have been fully released. The M_actual of this measured margin sample should be interpreted with caution. The filtered sample set includes events where the remaining transferable space on the sending side is sufficient to cover the estimated branch margin of the receiving branch, ensuring that the samples used for subsequent margin calculations are effective measured points that truly test the carrying capacity of the receiving branch, rather than conservative tests constrained by the sending side. Each record in the filtered sample set is annotated with the transferable capacity level on the sending side at that time, which facilitates the statistical analysis of verification events with different degrees of sufficiency according to capacity range.

[0053] The remaining available margin is calculated based on the filtered sample set to obtain the measured margin sample. For each transfer event in the filtered sample set, the current level reached by the receiving branch after the transfer represents the actual load point measured in this event. This directly reflects whether there is a gap between the theoretical estimate and the actual load capacity compared with the previously calculated branch margin estimate. The remaining available margin is calculated according to the formula: M_actual = I_margin - I_transfer, where I_margin is the branch margin estimate before the transfer occurs, and I_transfer is the measured current level reached after the transfer. The sign of M_actual reveals the direction of the deviation between the theoretical model and the actual measurement: when M_actual is significantly positive, the margin estimate is conservative, and the branch still has actual margin that has not been reached; when M_actual approaches zero or is negative, it indicates that the measured load has reached or even exceeded the theoretical boundary, and the accuracy of the branch margin estimate model needs to be reviewed. The calculated M_actual, together with the basic information of the transfer event, T_before (conductor stable temperature before transfer) and T_after (conductor stable temperature after transfer, ΔT=T_after-T_before), constitutes the measured margin sample, which can be directly read when constructing the temperature sample point set. The distribution of M_actual can reflect the systematic deviation of the branch margin estimation: when M_actual is positive and concentrated in a certain range, it indicates that the model is consistently conservative by that magnitude; when M_actual approaches zero or is negative, the measured margin sample has confirmed that the measured load is approaching or exceeding the model boundary. Taking the measured margin sample with M_actual=-5A as an example, the measured current after the transfer has exceeded the branch margin estimation at that time, and the upper limit of the branch current carrying capacity must be checked immediately. This is the direct triggering basis for re-evaluating the accuracy of the branch margin estimation model.

[0054] In some embodiments, the step of successively fitting the measured margin sample to the temperature to form a temperature margin curve includes: generating a temperature sample point set based on capturing the temperature change before and after the transfer using the measured margin sample; calculating the time and temperature intervals between adjacent sample points in the temperature sample point set to obtain sample sparsity; using the sample sparsity to screen for sample temperature coverage blind spots to determine unverified intervals; and narrowing the blind spot extrapolation range for the unverified intervals to form a temperature margin curve.

[0055] A temperature sample point set is generated by capturing temperature changes before and after the power transfer based on measured margin samples. Each record in the measured margin samples is labeled with two stable temperature readings, T_before and T_after. The temperature change is calculated using the formula: ΔT = T_after - T_before. This expands the samples, which originally only reflected current margin, to also reflect temperature performance, achieving an extended description from the current dimension to the temperature dimension. With L (per unit value, i.e., the ratio of measured current to rated current after power transfer) as the horizontal axis and ΔT as the vertical axis, the scatter points of each measured margin sample constitute the temperature sample point set. Each sample point also carries the source measured margin sample number and the occurrence time t_event of the corresponding power transfer event, so that the time interval Δtw in the subsequent sample sparsity calculation can be directly used, eliminating the step of reverse lookup of the original event record. If multiple sample points on the same branch have similar load proportions but significantly different ΔT values, it indicates that there are other factors affecting temperature changes besides load proportions (such as seasonal ambient temperature differences or the linkage contribution of adjacent branches). These differences are retained in the temperature sample point set without artificial smoothing or removal. When fitting the temperature margin curve later, the statistical model will automatically treat this scattered distribution as interval uncertainty and output a curve with a confidence range rather than a single deterministic curve. The temperature sample point set forms a two-dimensional scatter plot with load proportion L (horizontal axis) and temperature change ΔT (vertical axis). The range of L values ​​covered gradually extends as the measured margin samples accumulate. The density of the sample points on the L axis directly determines the reliability of each segment of the subsequent temperature margin curve.

[0056] The sparsity of the temperature sample points is obtained by statistically analyzing the time and temperature intervals between adjacent sample points. The scattered points in the temperature sample point set are not uniformly distributed across the entire load percentage range. The occurrence of power transfer events depends on the actual needs of power grid operation and dispatch. Some load percentage intervals may accumulate dense sample points, while other intervals may have no power transfer events occurring for a long period. The sample points in the temperature sample point set are sorted by load percentage, and the sample sparsity is calculated using the formula: ρ=max(Δtw / D_ref,ΔLw / L_ref), where Δtw is the time interval between adjacent sample points (calculated from the t_event field of each point), ΔLw is the load percentage interval, D_ref is taken as 180 days (approximately half a year, covering the complete summer and winter load cycle; data beyond this period will have reduced representativeness due to seasonal drift), and L_ref is taken as 0.2 as a reference normalization value. The larger of the two values ​​is chosen to ensure that a large value in either dimension will trigger a high sparsity determination. The sample sparsity ρ is given separately for each segment between every two adjacent sample points in the temperature sample point set, covering the entire load percentage range. In sections with high sample sparsity, even if there is one measured data point at each end, the thermal response pattern in the middle section has not been verified by real transfer events. If there is a sudden change or nonlinear acceleration in the actual thermal response within this load range, the interpolation will produce a systematic bias in both direction and magnitude that cannot be verified. The sample sparsity value is the risk basis that must be quantified when delineating unverified intervals. Highly sparse sections need to be marked separately when displaying curves to indicate weak evidence. The higher the ρ, the sparser or older the historical transfer events in this section are, indicating insufficient evidence. ρ=0 indicates that there is sufficient and recent measured support in this section, and it can be directly fitted. ρ transitions smoothly between the two levels according to a continuous formula, and the extrapolation narrowing ratio increases in the same direction as ρ.

[0057] Sample sparsity is used to screen for temperature coverage blind spots and identify unverified intervals. Segments with sample sparsity exceeding the set upper limit ρ_max=1.0 indicate that too few or too distant actual transfer events have occurred within this load percentage range, insufficient to provide a reliable judgment on the temperature change pattern of that segment. These segments constitute coverage blind spots on the temperature margin curve. Several adjacent segments with continuously exceeding sample sparsity limits are retrieved and merged into a continuous coverage blind spot. The start and end load percentages of the blind spot are taken as the percentage values ​​corresponding to the two ends of the exceeding segment. When a single isolated exceeding segment is flanked by segments with normal sparsity, it is still marked as a separate, narrower blind spot to accurately reflect the evidence gap at that load percentage point. If an unverified interval happens to fall within a high-load zone, its risk is higher than that of a blind zone in a low-load zone: In high-load zones, as T_max approaches, the thermal conductivity of the insulation layer decreases with temperature rise, leading to reduced heat dissipation efficiency and causing a larger temperature rise for the same current increment, resulting in a faster upward curve; low-load zones remain in the linear segment, with relatively smaller interpolation errors. Therefore, the absolute level of the load percentage in the unverified interval record must be indicated simultaneously. A more conservative narrowing strategy is adopted for unverified intervals in high-load zones, while a more relaxed approach is used in low-load zones. Identified coverage blind zones are integrated into unverified intervals. The start and end percentage range of each segment and the values ​​of the nearest measured sample points at both ends of the interval are used as anchoring benchmarks for extrapolation. The risk of unverified intervals in high-load zones (L>0.8) requires particular attention because the insulation thermal stress in this zone is close to the upper limit, and the actual curve slope is often steeper than the extrapolation prediction for low-load zones.

[0058] For unverified intervals, a narrowing of the blind zone extrapolation range is implemented to form a temperature margin curve. For sections with sufficient sample support, a curve segment showing the temperature change as a function of load percentage is directly fitted. For unverified intervals, interpolation or extrapolation is performed using the nearest measured sample points at both ends of the interval as anchor points, but the extrapolation range is actively narrowed to avoid extending into high-load intervals far exceeding the anchor point coverage, thus preventing sections without measured data from showing a seemingly continuous but unsubstantiated trend. The narrowing ratio is tiered according to the width of the unverified interval: for large blind zones with a width exceeding 20% ​​of the rated percentage, the extrapolation coverage does not exceed 50% of the interval width; for narrow blind zones with a width less than 10% of the rated percentage, full interpolation coverage is possible; there is a linear transition between the two tiers. During extrapolation, the slope of the tangent line at the anchor point is used. If the curve near the anchor point shows an accelerating upward trend, the extrapolation inherits this steep trend; if it is flat, it extends with a flat extension. The quality of extrapolation is highly sensitive to the measured sample density near the anchor point, and the edge of the densely measured sample section is preferentially used as the starting point for extrapolation. The temperature margin curve is composed of a measured fitted segment and a narrowed extrapolated segment. The extrapolated segment is separately marked on the curve to distinguish it from the measured fitted segment during display. The measured fitted segment values ​​can be directly used for decision-making, while the extrapolated segment values ​​are recommended to be overlaid with a temperature rise safety margin of at least 10% before use. No reference values ​​are provided for temporarily missing segments; a prompt is made to wait for subsequent measured margin samples to be added before evaluation. The temperature margin curve continuously iterates as measured margin samples accumulate. Previously marked as temporarily missing unverified intervals will be converted to measured fitted segments once a new valid transfer event occurs in that load interval and passes the screening process. When all unverified intervals are filled with measured data, the temperature margin curve consists of the entire measured fitted segment. At this point, the curve has the highest reliability and can be directly referenced by the dispatcher without additional safety margin.

[0059] Step S105: Calculate the slope of temperature sensitivity of each branch on the temperature margin curve to generate margin sensitivity. Based on the margin sensitivity, identify the cut-point branches at the end of the topology to form the network health level. Based on the network health level, perform end margin distribution analysis and output the power supply margin result.

[0060] Specifically, the temperature margin curve is used to statistically analyze the slope of the temperature sensitivity of each branch to generate the margin sensitivity. The temperature margin curve shows the change in temperature as a function of the load percentage. The steepness of the curve varies across different load ranges. For example, if the load percentage increases by one percentage point, the temperature change may be small in the flat section of the curve, but quite significant in the steep section. Looking only at the absolute value of the temperature change at a single point on the curve cannot reflect how quickly the temperature will deteriorate as the load on that branch continues to increase. The current actual load percentage of the branch is taken at the position on the temperature margin curve, and two adjacent points are taken in the direction of higher load percentage. The sensitivity slope is calculated using the formula: s = (ΔT2 - ΔT1) / (L2 - L1), where L1 and L2 are the two adjacent load percentage values ​​near the current operating point (per unit, range 0-1), ΔT1 and ΔT2 are the corresponding temperature changes, and s is the margin sensitivity in °C / pu. When s is large, even a small increase in load can cause a significant temperature rise, leaving little remaining scheduling flexibility, requiring priority to be given to current limiting; when s is small, the curve is flat, and the risk of short-term load increases is relatively controllable. Margin sensitivity is calculated separately for each branch's current operating point. The load of the same branch changes daily, and the position of the corresponding operating point on the temperature margin curve also shifts accordingly. Margin sensitivity needs to be reread according to the actual load level. If the margin sensitivity falls near the unverified range of the temperature margin curve, the slope here is derived from interpolation. If the actual thermal response in this range has entered an accelerated upward phase, the interpolation will systematically underestimate the slope. When using this value, it is necessary to distinguish it from the values ​​within the measured fitted segment to avoid making overly optimistic scheduling arrangements based on weak evidence slopes.

[0061] In some embodiments, the step of identifying end-topology cut-point branches based on the margin sensitivity to form a network health level includes: generating real-time topology samples by tracking branch deployment and deployment topology changes based on the margin sensitivity; using the real-time topology samples to statistically analyze the proportion of temporary and permanent cut-points to obtain cut-point nature categories; verifying the cut-point nature categories against temporary cut-point duration thresholds to establish a valid cut-point set; and performing topology criticality classification based on the valid cut-point set to form a network health level.

[0062] Real-time topology samples are generated based on the tracking of branch connection / disconnection topology changes using margin sensitivity. Each time a switch status changes, the margin sensitivity value of each involved branch is recorded, and this value, along with a snapshot of the connection relationships before and after the switch change, forms a real-time topology sample. For branches with tie switches in the branch box, when a branch is taken out of service due to planned maintenance outages or line faults, the tie switch closes, and the electrical connection relationship of the receiving branch changes accordingly. The margin sensitivity determines the intensity of the thermal response after the connection is established, and therefore must be recorded synchronously in the real-time topology sample. The time when the switch status changes is retrieved, and combined with the corresponding margin sensitivity of each branch at that time, the changes in the connection relationships between which branches changed before and after the change are recorded. Branches with higher margin sensitivity will exhibit a more drastic temperature response to load changes once they take on the newly added downstream load after the tie switch is established; this is also noted when recording topology changes. Each snapshot of the connection relationships before and after a switch state change forms a real-time topology sample, recording the time of the topology change, the connection relationships before and after, and the margin sensitivity level of the involved branches. If multiple branches are changed simultaneously in the same commissioning / decommissioning operation, they are included in the same real-time topology sample for unified recording. The snapshot frequency of the real-time topology sample is consistent with the update frequency of distribution automation remote signaling, usually accurate to the minute level. Cut point identification performs graph search on each snapshot separately, rather than analyzing the entire sequence as a whole, ensuring that each topology state is independently calculated. If a branch box has not undergone commissioning / decommissioning operations for a long period, only the beginning and end snapshots are retained, and intermediate periods are not recorded repeatedly. Once a branch with high margin sensitivity becomes a connection channel, its hot-state risk increases sharply. This key attribute is recorded through the correlation between real-time topology samples and margin sensitivity, so that the impact of each topology change on network hot-state security can be quantitatively tracked.

[0063] The nature of cut points is categorized by statistically analyzing the proportion of temporary and permanent cut points using real-time topology samples. Cut point retrieval is performed on each connection snapshot in the real-time topology samples, and the percentage of times each branch is identified as a cut point in all snapshots is calculated. Branches with a percentage close to 100% consistently act as cut points regardless of switch orientation and are classified as permanent cut points. Branches with a percentage significantly lower than 100%, exhibiting cut point characteristics only in partial switch orientation states, are classified as temporary cut points. The resulting cut point occurrence ratio reflects the frequency with which each branch serves as the sole path in the historical operating topology. A higher ratio indicates that the branch is almost always at a cut point position regardless of switch orientation combinations, and its failure impact should be prioritized during upgrades. Branches with a lower ratio only trigger their cut point role under specific maintenance combinations, and their temporary risk is directly linked to the maintenance plan. The cut point nature category is labeled for each branch. For temporary cut points, the start and end times (t_onset, t_recovery) of each cut point state segment are recorded. The state transition time is extracted from the snapshot sequence of the real-time topology sample, and the duration is t_recovery - t_onset. After adding a new connection path, the real-time topology sample brought by the new path needs to be included in the statistics and recalculated to confirm whether the cut point nature category of the original permanent cut points has changed. If the branch box topology is extremely simple and there are no connection switches between branches, then all branches are cut points in all snapshots, and the cut point nature category will be labeled as permanent cut points, indicating that the branch box has no backup connection path and the topology redundancy is extremely low.

[0064] Establish a valid cutpoint set by verifying the duration threshold of temporary cutpoints based on their nature category. Branches marked as permanent cutpoints in the cutpoint nature category should be included in the subsequent critical assessment, regardless of their duration, because they reflect the inherent vulnerability of the branch box's topology. Branches marked as temporary cutpoints, on the other hand, only have their cutpoint status under specific combination of commissioning and decommissioning. If this combination only lasts for a short time and then returns to normal, its actual impact on network health is quite limited. However, if this combination lasts for a long time (such as when a branch fails to recover due to long-term maintenance), the risk posed by temporary cutpoints during this period is almost the same as that of permanent cutpoints. For each cutpoint state marked as temporary in the cutpoint nature category, the duration (in days) from t_onset to t_recovery is calculated. Temporary cutpoint states lasting longer than seven days are retrieved. For medium-voltage planned maintenance outages not exceeding five working days, outages exceeding seven days indicate fault repair or equipment replacement. During this period, the risk persistence of the temporary cutpoint is equivalent to that of a permanent cutpoint, and it is considered as having the same effect as a permanent cutpoint, thus being included in the subsequent evaluation scope. In practice, t_onset and t_recovery are extracted for each cutpoint state according to the time series of the cutpoint nature category. The duration is calculated as duration = t_recovery - t_onset (in days). Temporary cutpoint states with duration ≤ 7 days are skipped, while those with duration > 7 days are merged with all permanent cutpoint records to form a valid cutpoint set. Permanent cut points and temporary cut point states that exceed the duration threshold are included in the effective cut point set. Each record includes the cut point nature, the downstream affected branch range, and the margin sensitivity value s of the branch at that time (directly taken from the real-time topology sample). If the same branch experiences multiple temporary cut point states that exceed the threshold, the effective cut point set records each one separately, with each instance corresponding to a different time period and different deployment / retreat combinations.

[0065] Network health levels are determined by classifying topological criticality based on the effective cut point set. For example... Figure 2As shown, the distribution transformer 1 supplies power to each low-voltage cable branch box 3 via the low-voltage main feeder 2, and the low-voltage cable branch box 3 distributes power to the terminal load node 5 via the outgoing cable 4; a tie switch 6 is provided between adjacent branch boxes 3, which is disconnected during normal operation and only activated during switching operations. Effective cut points are concentrated; permanent cut point 7 constitutes a cut point in all operating topology states, while temporary cut point 8 only exhibits cut point characteristics under specific tie switch 6 activation / deactivation combinations; permanent cut point 7 is located in the level 3 high-risk area 10, where the downstream impact range of branches is wide and the margin sensitivity is high; failure in this area will cause all downstream load nodes 5 to lose power; branches located in the level 1 health area 9 do not constitute cut points, and the network health level is assessed as level 1. For each record in the effective cut point set, count the number of branches n affected downstream of its corresponding branch, and combine this with the margin sensitivity value s of that branch at that time. Records with n≤2 and s≤s_threshold (it is recommended that s_threshold be the median of the margin sensitivity of all branches in the entire branch box) are rated as Level 1, with low risk; records with n>2 or s>s_threshold but not both are rated as Level 2; records with n>2 and s>s_threshold are rated as Level 3, which is a high-risk situation that is both a topological critical node and has relatively tight margin. These records should be given priority in the inspection plan and additional backup channels should be considered. The network health level is determined by the record with the highest rating among the effective cut points within the branch box. If the rating of an individual cut point is too high, the overall health level of the branch box will be lowered. For example, if the margin sensitivity s=42℃ / pu of a permanent cut point branch exceeds the median s_threshold=30℃ / pu of the entire branch box, and the number of downstream affected branches n=4, both conditions are met simultaneously, directly triggering a level 3 assessment. The network health level of the entire branch box will then be raised to level 3, triggering priority inspection and modification response.

[0066] Power supply margin results are output based on the distribution analysis of end-point margins according to network health levels. Network health levels reflect whether there are vulnerable single points in the internal topology of the branch box. This structural risk and the individual margin levels of each branch are two independent yet interconnected issues. Even if a branch has ample margin estimates, if it is also classified as a high-risk cut point in the network health level assessment, its downstream impact will still be widespread if it experiences an anomaly. Conversely, even if a branch does not constitute a cut point, if its margin is already strained, it requires separate attention. For all end-point branches within the branch box, the following three pieces of information are summarized: branch margin estimates, margin sensitivity values, and whether it is listed as a high-risk cut point in the network health level assessment. These are organized separately for each branch, and no recalculation is performed during the summarization; only a consolidated presentation is provided. The power supply margin results first provide the overall health level of the entire branch box according to the network health level. Then, the margin level and margin sensitivity of each end branch are listed one by one. Branches that simultaneously meet the two conditions of margin tension (I_margin / I_max<0.85, that is, the thermal coupling of the adjacent road has squeezed more than 15% of the temperature rise budget and the remaining elasticity is insufficient to cope with a medium-amplitude power transfer impact) and topological critical (included in the effective cut point set) are highlighted for maintenance personnel to prioritize inspection or modification. Taking the power supply margin result output of a 6-way branch box as an example: the current network health level of the entire box is level 2, and the margin of 5 branches is normal; branch 3 simultaneously meets the conditions of tight margin (I_margin=78A, I_max=96A, ratio 0.81 is lower than the threshold of 0.85) and topological critical (marked as a permanent cut point by the effective cut point set, margin sensitivity s=38℃ / pu is higher than the median of the entire box), and is marked as a high-risk branch in the power supply margin result. It is recommended to complete the capacity expansion or the addition of a new communication channel within this month.

[0067] To execute the methods described in the above embodiments and achieve the corresponding functions and technical effects, this application also provides the following system. See [link to system]. Figure 3 , Figure 3 This paper presents a structural block diagram of a dynamic evaluation system 300 for the power supply capacity at the end of a low-voltage cable branch box, according to an embodiment of this application. The system includes: The thermal resistance calibration module 301 is used to acquire the temperature signal of each branch conductor in the low-voltage cable branch box, screen the power outage period based on the conductor temperature signal to obtain the power outage calibration window, and perform thermal resistance feature analysis based on the power outage calibration window to determine the initial value of the environmental thermal resistance. The load measurement module 302 is used to back-calculate the light load operation period of the branch based on the initial value of the environmental thermal resistance to obtain the light load calibration window, analyze the thermal resistance characteristics of the thermal circuit from the light load calibration window to generate the thermal circuit parameters of the branch, and use the thermal circuit parameters of the branch to determine the instantaneous most unfavorable fluctuation to determine the upper limit of the branch current carrying capacity. The coupling correction module 303 is used to determine the cross thermal coupling coefficient by converting the full-load temperature rise for the upper limit of the branch current carrying capacity, analyze the contribution of the adjacent full-load temperature rise based on the cross thermal coupling coefficient to generate the margin correction coefficient, and use the margin correction coefficient to reduce the upper limit of the branch current carrying capacity to obtain the branch margin estimate. The margin mapping module 304 is used to track the step transfer of branch load based on the branch margin estimate to form a transfer load list, screen the transfer load list to obtain the measured margin sample of the transfer non-limited section, and perform successive transfer temperature continuous fitting on the measured margin sample to form a temperature margin curve. The health assessment module 305 is used to statistically analyze the temperature sensitivity slope of each branch of the temperature margin curve to generate margin sensitivity, identify the terminal topology cut point branches based on the margin sensitivity to form a network health level, and perform terminal margin distribution analysis based on the network health level to output power supply margin results.

[0068] The aforementioned dynamic evaluation system 300 for the power supply capacity at the end of a low-voltage cable branch box can implement one of the methods described in the above embodiments for dynamic evaluation of the power supply capacity at the end of a low-voltage cable branch box. The options described in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining contents of this application's embodiments can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0069] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

Claims

1. A method for dynamically evaluating the power supply capacity at the end of a low-voltage cable branch box, characterized in that, include: Acquire the temperature signals of conductors in each branch of the low-voltage cable branch box, screen outage periods based on the conductor temperature signals to obtain the outage calibration window, and determine the initial value of the environmental thermal resistance based on the thermal resistance characteristic analysis of the outage calibration window. The light-load calibration window is obtained by back-calculating the light-load operation period of the branch based on the initial value of the environmental thermal resistance. The thermal resistance characteristics of the branch are analyzed from the light-load calibration window to generate the branch thermal parameters. The branch thermal parameters are used to determine the instantaneous most unfavorable fluctuation to determine the upper limit of the branch current carrying capacity. The cross thermal coupling coefficient is determined by converting the full-load temperature rise of the branch current limit. The contribution of the adjacent full-load temperature rise is analyzed based on the cross thermal coupling coefficient to generate the margin correction coefficient. The margin correction coefficient is used to reduce the branch current limit to obtain the branch margin estimate. Based on the estimated branch load margin, a step transfer of branch load is tracked to form a transfer load list. The transfer load list is used to screen the transfer non-limited sections to obtain measured margin samples. The measured margin samples are then subjected to successive transfer temperature continuous fitting to form a temperature margin curve. The temperature margin curve is statistically analyzed to generate margin sensitivity by calculating the slope of temperature sensitivity of each branch. Based on the margin sensitivity, the terminal topology cut point branches are identified to form a network health level. Based on the network health level, the terminal margin distribution is analyzed to output the power supply margin result.

2. The method according to claim 1, characterized in that, The process of obtaining the power outage calibration window by screening the power outage period based on the conductor temperature signal includes: Based on the conductor temperature signal, identify the gradual temperature rise pattern before the power outage to obtain a sample of the power outage characteristics; Power outage type labels are generated based on discrepancies in the verification morphology of the power outage nature samples. The residual heat dissipation delay is obtained by calculating the residual heat decay time constant using the power outage type label. The power outage calibration window is obtained by extracting the delayed dissipation window based on the residual heat dissipation time delay.

3. The method according to claim 1, characterized in that, The determination of the branch current carrying capacity limit by measuring the instantaneous most unfavorable fluctuation using the branch thermal circuit parameters includes: Based on the branch thermal circuit parameters, instantaneous temperature jumps are tracked to form an instantaneous fluctuation sample set; The validity discrimination parameters are obtained by identifying the fixed amplitude repetitive jump interference features through the instantaneous fluctuation sample set; Based on the validity discrimination parameters, sensor noise jumps are eliminated to generate valid fluctuation samples; The upper limit of branch current carrying capacity is determined by performing the most unfavorable fluctuation value determination based on the effective fluctuation sample.

4. The method according to claim 1, characterized in that, The determination of the cross thermal coupling coefficient by performing full-load temperature rise conversion based on the upper limit of the branch current carrying capacity includes: Equivalent temperature rise samples are generated by performing full-load temperature rise conversion processing based on the upper limit of the branch current carrying capacity. The equivalent temperature rise sample was analyzed to determine the ratio of conduction paths in the metal shell to convection paths in the chamber, and to identify the coupling path type. Based on the long-term non-attenuation characteristic of the coupling strength analyzed by the aforementioned coupling path type, the coupling strength classification is obtained; The cross-thermal coupling coefficient is determined by calibrating the coefficients through the coupling strength classification.

5. The method according to claim 1, characterized in that, The step of obtaining measured margin samples by screening the non-over-limit sections of the transferred load list includes: For the aforementioned load transfer list, verify the load transfer ratio range to generate high-load candidate samples; The validity of the samples and the case of exceeding the limit are verified by using the high-load candidate samples to obtain the effective sample set; Based on the valid sample set, the outgoing path restriction is parsed to form a filtered sample set; The remaining available margin is calculated based on the filtered sample set to obtain the measured margin sample.

6. The method according to claim 1, characterized in that, The step of successively transferring the measured margin sample to the temperature and continuously fitting it to form a temperature margin curve includes: A temperature sample point set is generated based on the measured margin sample to capture the temperature change before and after the transfer. The sample sparsity is obtained by statistically analyzing the time and temperature intervals between adjacent sample points in the temperature sample point set. The sample sparsity is used to screen for sample temperature coverage blind spots and determine unverified intervals. For the unverified interval, the blind zone extrapolation range is narrowed to form a temperature margin curve.

7. The method according to claim 1, characterized in that, The process of identifying end-topology cut-point branches based on the margin sensitivity to form a network health level includes: Based on the margin sensitivity, real-time topology samples are generated by tracking branch topology changes during deployment and withdrawal. The proportion of temporary and permanent cut points is statistically analyzed using the real-time topology samples to obtain the cut point property categories; Verify the duration threshold of temporary cut points based on the described cut point property categories to establish a valid cut point set; Based on the set of effective cut points, a topological criticality classification is performed to form a network health level.

8. The method according to claim 4, characterized in that, The process of generating an equivalent temperature rise sample by converting the full-load temperature rise based on the upper limit of the branch current carrying capacity includes: The current carrying capacity upper limit of the branch is used to eliminate the influence of current carrying capacity differences to obtain the current carrying capacity normalization factor; Based on the current-carrying normalization factor, the anomaly verification of the top branch heat dissipation is performed to obtain the heat dissipation characteristic value. Spatial gradient parameters are obtained by analyzing the temperature rise gradient distribution of the upper and lower layers based on the aforementioned heat dissipation characteristic values. Based on the spatial gradient parameters, gradient compensation is performed according to the vertical position to generate an equivalent temperature rise sample.

9. The method according to claim 5, characterized in that, The step of generating high-load candidate samples for verifying the proportion of transferred load in the transferred load list includes: Based on the aforementioned load transfer list, the load change rate before and after the transfer is tracked to obtain load jump samples; A high-load screening threshold is formed by statistically analyzing the percentage distribution of load jump amplitudes in the aforementioned load jump samples. A composite working condition label is obtained by superimposing the screening ratio of the high-load screening threshold and the over-limit section. The composite operating condition label is set to prioritize and grade the superimposed load to generate high-load candidate samples.

10. A dynamic evaluation system for the power supply capacity at the end of a low-voltage cable branch box, characterized in that, include: The thermal resistance calibration module is used to acquire the temperature signals of the conductors of each branch in the low-voltage cable branch box, screen the power outage period based on the conductor temperature signals to obtain the power outage calibration window, and perform thermal resistance feature analysis based on the power outage calibration window to determine the initial value of the environmental thermal resistance. The load measurement module is used to back-calculate the light-load operation period of the branch based on the initial value of the environmental thermal resistance to obtain the light-load calibration window, analyze the thermal resistance characteristics of the thermal circuit from the light-load calibration window to generate the thermal circuit parameters of the branch, and use the thermal circuit parameters of the branch to determine the instantaneous most unfavorable fluctuation to determine the upper limit of the branch current carrying capacity. The coupling correction module is used to determine the cross thermal coupling coefficient by converting the full-load temperature rise for the upper limit of the branch current carrying capacity, analyze the contribution of the adjacent full-load temperature rise based on the cross thermal coupling coefficient to generate the margin correction coefficient, and use the margin correction coefficient to reduce the upper limit of the branch current carrying capacity to obtain the branch margin estimate. The margin mapping module is used to track the step shift of branch load based on the branch margin estimate to form a transfer load list, screen the transfer load list to obtain the measured margin sample of the transfer non-limited section, and perform successive transfer temperature continuous fitting on the measured margin sample to form a temperature margin curve. The health assessment module is used to statistically analyze the temperature sensitivity slope of each branch of the temperature margin curve to generate margin sensitivity, identify the terminal topology cut point branches based on the margin sensitivity to form a network health level, and perform terminal margin distribution analysis based on the network health level to output power supply margin results.

Images