Intelligent metal comprehensive distribution box compensation loop hot spot virtual temperature measurement and early warning method

Abstract

This invention relates to the field of thermal status sensing and early warning technology for power distribution equipment, and particularly to a method for virtual temperature measurement and early warning of hotspots in the compensation circuit of an intelligent metal integrated distribution box. The method includes: Step 1: Collecting limited measurement point information and operating status information of the compensation circuit; Step 2: Establishing a reference thermal network model of the compensation circuit, solving for the reference temperature of each hidden node in the model, and thus obtaining the overall hotspot reference temperature; Step 3: By comparing the measured measurement points with the reference temperature, identifying abnormal measurement points with missing or drifting states, and performing corresponding compensation filling or retaining measured values; Step 4: Constructing a virtual temperature measurement feature vector, and obtaining the reference temperature correction amount through a temperature estimation model, and then using the reference temperature correction amount to reconstruct the hidden node temperature and the overall hotspot temperature; Step 5: Outputting an over-temperature early warning result based on the reconstructed overall hotspot temperature and its temperature rise trend. This invention can improve the continuity of hotspot sensing and the timeliness of over-temperature early warning.

Description

Technical Field

[0001] This invention relates to the field of thermal status sensing and early warning technology for low-voltage complete sets of power distribution equipment, and in particular to a method for virtual temperature measurement and early warning of hot spots in the compensation circuit of an intelligent metal integrated power distribution box. Background Technology

[0002] In addition to the main power distribution circuit, the JP intelligent metal integrated distribution box also includes a compensation circuit. The busbar connection points, incoming and outgoing terminals of the compensation circuit breaker, and the wiring terminals of the reactive power compensation module in the compensation circuit may all develop localized hot spots during long-term operation. In actual engineering projects, independent temperature sensors are typically not installed at each hot spot location, thus presenting the following problems:

[0003] First, direct temperature measurement at only a small number of points is insufficient to cover all key hot spots and may result in the omission of hidden hot spots.

[0004] Second, sensors may experience missing measurements, drift, or short-term inaccuracies. If the actual measured values ​​are relied upon directly, it can easily affect the over-temperature warning results.

[0005] Third, the thermal state of the compensation circuit is not only affected by the current, but also by harmonics, branch health status and environmental conditions. If the temperature is inferred based solely on a simple linear estimate, the accuracy will be insufficient.

[0006] Fourth, if the actual measurement point has not yet reached the direct alarm threshold, but the hidden hot spot is approaching the dangerous temperature, the traditional measurement point direct reporting method will be unable to issue an early warning in a timely manner.

[0007] Therefore, there is an urgent need to provide a technical solution that can reconstruct the hot spot temperature of the compensation loop under limited measurement points and further output the over-temperature early warning result. Summary of the Invention

[0008] To address the aforementioned technical problems in existing technologies, this invention proposes a method for virtual temperature measurement and early warning of hotspots in the compensation circuit of an intelligent metal integrated power distribution box. The specific technical solution is as follows:

[0009] A method for virtual temperature measurement and early warning of hotspots in the compensation circuit of an intelligent metal integrated distribution box, comprising:

[0010] Step 1: Collect limited measurement point information and operating status information of the compensation loop;

[0011] Step 2: Establish a reference thermal network model for the compensation loop, solve for the reference temperature of each hidden node in the model, and then obtain the overall hot spot reference temperature;

[0012] Step 3: By comparing the measured points with the reference temperature, identify abnormal measuring points that are missing or drifting, and perform corresponding compensation or retention operations.

[0013] Step 4: Construct a virtual temperature measurement feature vector, obtain a reference temperature correction amount through a temperature estimation model, and then use the reference temperature correction amount to reconstruct the hidden node temperature and the overall hot spot temperature;

[0014] Step 5: Based on the overall hot spot temperature of the reconstruction and its temperature rise trend, output the over-temperature warning result.

[0015] Furthermore, in step 1, the limited measurement point information includes the temperature of the busbar connection point of the compensation cabinet, the temperature of the output terminal of the compensation circuit breaker, the main circuit current, the compensation current, the total harmonic current distortion rate, the ambient temperature, and the current of each compensation branch, switching commands, and feedback status output by the distribution monitoring and compensation control unit of the distribution box; the operating status information is as follows: The method of obtaining, among which The health status of the m-th compensation branch at time k; Let be the current deviation rate of the m-th compensated branch, used to characterize the degree of deviation of the measured branch current from the reference current; This is the switching anomaly indication for the m-th compensation branch, when a switching command is issued. With feedback status If the sampling periods are inconsistent for N consecutive periods, the value is 1; otherwise, the value is 0. λ1 and λ2 are the weighting coefficients of the branch current deviation term and the switching anomaly term, respectively. The upper limit threshold for the health status of the branch; Indicates the input quantity Limited to the range Inside.

[0016] Furthermore, the reference thermal network model in step 2 includes the air node inside the distribution box, the busbar connection point node of the compensation cabinet, the incoming terminal node of the compensation circuit breaker, the outgoing terminal node of the compensation circuit breaker, the wiring terminal node of the reactive power compensation module, the cavity node of the compensation controller, and the box shell node.

[0017] Furthermore, each hidden node in step 2 satisfies a heat dissipation balance relationship: ,in, The sampling period; For the first Each node at time... temperature, For the first The heat generation power of each node, For the first The node and the first Equivalent thermal resistance between nodes For the first The equivalent heat transfer resistance between each node and the environment. For the first The equivalent heat capacity of each node, Ambient temperature;

[0018] No. The heat capacity of each node is calculated according to Calculate; thermal resistance between adjacent nodes according to Calculation; thermal resistance between nodes and the environment according to Calculate; where, The mass of the conductor or structural component corresponding to the node. For the specific heat capacity of the material, This is the equivalent heat transfer path length. The equivalent thermal conductivity is For the equivalent heat transfer cross-sectional area, To consider the overall heat transfer coefficient, This refers to the heat exchange area.

[0019] Furthermore, the first The heating power of each node is related to the square of the compensation current and is modulated by the health status quantity and harmonic influence coefficient of the compensation branch.

[0020] Furthermore, in step 2, the initial equivalent contact resistance of each hidden node, including the busbar connection point of the compensation cabinet, the incoming terminal of the compensation circuit breaker, the outgoing terminal of the compensation circuit breaker, and the wiring terminal of the reactive power compensation module, is given an initial value based on the rated operating conditions and structural dimensions. Using calibration samples covering no-load, rated load, staged switching, harmonic disturbance, and ambient temperature change conditions, the square error between the measured temperature and the output temperature of the thermal network is minimized by the iterative least squares method with boundary constraints. The initial equivalent contact resistance, the equivalent thermal resistance between nodes, and the equivalent heat transfer thermal resistance between nodes and the environment are jointly corrected to solve for the reference temperature of each node and form a reference thermal network parameter table.

[0021] Furthermore, in step 2, the maximum value among the reference temperatures of the busbar connection point of the compensation cabinet, the incoming terminal of the compensation circuit breaker, the outgoing terminal of the compensation circuit breaker, and the wiring terminal of the reactive power compensation module is taken as the overall hot spot reference temperature.

[0022] Furthermore, in step 3, the residual between the measured temperature and the reference temperature at the bus connection point is tracked. If the residual continues to deviate within the most recent preset time window and the difference between the slope of the measured temperature change and the slope of the reference temperature change exceeds a threshold, it is identified as a drift state and the measured temperature is retained. The drift state quantity is also used as an additional feature input into the lightweight gradient boosting regression correction model in subsequent steps. For the measurement point at the outgoing end of the compensation circuit breaker, if more than 3 sampling points are missing consecutively within a preset time interval, it is marked as a missing measurement state and the reference temperature is used to compensate and fill it.

[0023] Furthermore, in step 4, the virtual temperature measurement feature vector consists of the following elements: compensation temperature of the busbar connection point of the compensation cabinet, compensation temperature of the output terminal of the compensation circuit breaker, missing and drift state quantities corresponding to the measured points, ambient temperature, total harmonic current distortion rate, main circuit current, compensation current, maximum branch health state quantity, overall hot spot reference temperature change rate of the most recent preset time period, and the circuit breaker input terminal temperature, module terminal temperature and overall hot spot reference temperature output by the reference thermal network model.

[0024] Furthermore, in step 4, the temperature estimation model consists of three single-output lightweight gradient boosting regression sub-models corresponding to the incoming terminal of the compensation circuit breaker, the terminal of the reactive power compensation module, and the overall hotspot temperature, respectively. Calculate the reference temperature correction amount, and according to The final estimated temperature is obtained, where The corresponding reference temperature output in step 2. This is the i-th single-output lightweight gradient boosting regression sub-model corresponding to the target temperature output;

[0025] The lightweight gradient boosting regression sub-model consists of 60 to 100 CART regression trees, each with a maximum depth of 2 to 4, a learning rate of 0.02 to 0.10, and a minimum number of leaf node samples of 3 to 10. The squared error is used as the loss function. The training samples are mainly measured data of the physical prototype under normal operating conditions, low power factor fluctuation conditions, harmonic pollution conditions, branch aging disturbance conditions, and abnormal switching conditions, supplemented by boundary condition augmentation data generated by the digital prototype.

[0026] Beneficial effects: This invention does not simply use adjacent measuring points for linear interpolation. Instead, it first establishes a reference thermal network that conforms to the structure of the JP distribution box compensation circuit, and then couples the branch health status quantities, harmonic influence coefficients, and abnormal state quantities of measuring points into the equivalent contact resistance and residual correction process. Compared with methods that rely solely on a limited number of measuring points or solely on a reference thermal network, this invention can more effectively reconstruct hidden hot spots in the compensation circuit and provide a more reliable basis for over-temperature early warning for the compensation circuit. Attached Figure Description

[0027] Figure 1 This is a flowchart of the main process of a method for virtual temperature measurement and early warning of hot spots in a smart metal integrated distribution box compensation circuit according to this embodiment.

[0028] Figure 2 This is a schematic diagram of the arrangement of key measuring points and virtual nodes in the compensation loop of this embodiment.

[0029] Figure 3 This is a comparison diagram of hotspot temperature reconstruction in the compensation loop of this embodiment.

[0030] Figure 4This is a diagram showing the effect of hotspot temperature compensation under the missing measurement condition in this embodiment.

[0031] Figure 5 This is a comparison chart of virtual temperature measurement error and over-temperature warning lead time in this embodiment. Detailed Implementation

[0032] To make the objectives, technical solutions, and technical effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0033] This embodiment discloses a method for virtual temperature measurement and early warning of hot spots in a compensation circuit, applicable to JP intelligent metal integrated distribution boxes with reactive power compensation modules, compensation circuit breakers, busbar connection points of compensation cabinets, and power distribution monitoring and compensation controllers.

[0034] The JP intelligent metal integrated power distribution box includes a box body and a signal acquisition unit, a power distribution monitoring and compensation control unit, a thermal status analysis unit, a storage unit, and an early warning output unit housed within the box body. The signal acquisition unit includes a temperature acquisition unit, a current acquisition unit, and a harmonic acquisition unit. The thermal status analysis unit is used to call programs in the storage unit to execute, for example... Figure 1 The method for virtual temperature measurement and early warning of hotspots in the compensation circuit of an intelligent metal integrated distribution box, as shown, specifically includes the following steps:

[0035] Step 1: Collect limited measurement point information and operating status information of the compensation circuit in the JP intelligent metal integrated distribution box.

[0036] In this embodiment, reference Figure 2 As shown, the temperature acquisition unit retains only two actual measurement points for long-term installation: the busbar connection point of the compensation cabinet and the output terminal of the compensation circuit breaker. The input terminal of the compensation circuit breaker, the wiring terminal of the reactive power compensation module, and the overall hot spot temperature in the compensation circuit are used as virtual temperature measurement targets. During the development phase, additional temperature sensors can be temporarily deployed for calibration and verification.

[0037] The limited measurement point information includes at least the temperature of the busbar connection point of the compensation cabinet. Compensation circuit breaker output terminal temperature Main circuit current Compensation current Total harmonic current distortion rate Ambient temperature and the current of each compensation branch output by the power distribution monitoring and compensation control unit. Throwing and cutting instructions With feedback status .

[0038] The operating status information is as follows The method of obtaining, among which, The health status of the m-th compensation branch at time k; Let be the current deviation rate of the m-th compensated branch, used to characterize the degree of deviation of the measured branch current from the reference current; This is the switching anomaly indication for the m-th compensation branch, when a switching command is issued. With feedback status If the sampling periods are inconsistent for N consecutive periods, the value is 1; otherwise, the value is 0. λ1 and λ2 are the weighting coefficients of the branch current deviation term and the switching anomaly term, respectively. The upper limit threshold for the health status of the branch; Indicates the input quantity Limited to the range Inside.

[0039] Step 2: Establish a reference thermal network model for the compensation loop and solve for the reference temperature of each hidden node in the model.

[0040] A reference thermal network model is established based on the structural relationship of the compensation circuit. The reference thermal network model includes at least the air node inside the box, the bus connection point node of the compensation cabinet, the incoming terminal node of the compensation circuit breaker, the outgoing terminal node of the compensation circuit breaker, the wiring terminal node of the reactive power compensation module, the cavity node of the compensation controller, and the box shell node.

[0041] In this embodiment, each node satisfies the heat dissipation balance relationship. .in, The sampling period is 1 minute. For the first Each node at time... temperature, For the first The heat generation power of each node, For the first The node and the first Equivalent thermal resistance between nodes For the first The equivalent heat transfer resistance between each node and the environment. For the first The equivalent heat capacity of each node, The ambient temperature.

[0042] The reference heat network parameters are determined as follows: The heat capacity of each node is calculated according to Calculate; thermal resistance between adjacent nodes according to Calculation; thermal resistance between nodes and the environment according to Calculation. Among them, The mass of the conductor or structural component corresponding to the node. For the specific heat capacity of the material, This is the equivalent heat transfer path length. The equivalent thermal conductivity is For the equivalent heat transfer cross-sectional area, To consider the overall heat transfer coefficient, This refers to the heat exchange area.

[0043] The initial equivalent contact resistance of key contact nodes (i.e., hidden nodes) including busbar connection points of compensation cabinets, incoming terminals of compensation circuit breakers, outgoing terminals of compensation circuit breakers, and wiring terminals of reactive power compensation modules. First, initial values ​​are given based on rated current, conductor cross-section, and installation structure. Then, using calibration samples covering no-load, 25% rated compensation current, 50% rated compensation current, 75% rated compensation current, 100% rated compensation current, staged switching, harmonic disturbances, and ambient temperature changes, the square error between the measured temperature and the thermal network output temperature is minimized using an iterative least squares method with boundary constraints. , and Joint calibration is performed; the iteration stops when the relative decrease rate of the objective function is less than 1% for 5 consecutive iterations, and the calibration validity is determined by the average absolute error of the measurement points not exceeding 2 degrees Celsius on the independent verification samples, thereby obtaining a reference thermal network parameter table for online solution.

[0044] The overall hot spot reference temperature is taken as the maximum value among the reference temperatures of the busbar connection point of the compensation cabinet, the incoming terminal of the compensation circuit breaker, the outgoing terminal of the compensation circuit breaker, and the wiring terminal of the reactive power compensation module.

[0045] In this embodiment, the heat generation power at the incoming terminal, outgoing terminal, and reactive power compensation module terminal of the compensation circuit breaker is related to the square of the compensation current and is further modulated by the branch health status quantity and harmonic influence coefficient. For the i-th heat generation node, its exemplary heat generation power is expressed as:

[0046] ,in .

[0047] in, Let be the initial equivalent contact resistance of the i-th heat-generating node under normal operating conditions. Let be the sensitivity coefficient of the i-th hot spot node to the health status of the branch. Let be the sensitivity coefficient of the i-th heating node to harmonic effects. For the health status measurement of each compensation branch The maximum value in. Let the first... Each compensation branch is at any time The health status is measured as follows:

[0048] ,

[0049] Branch current deviation rate, This is an abnormal indication value that takes a value of 1 if the switching command and the feedback status are inconsistent for three consecutive sampling periods, and 0 otherwise. Take the health status of each compensation branch The maximum value in.

[0050] Let be the reference current of the m-th compensation branch under the current target switching state. Let be the rated current of the m-th compensated branch; the harmonic influence coefficient at time k is defined as... . This represents the per-unit value of the total harmonic distortion rate. This represents the upper limit of the harmonic influence coefficient. In one exemplary implementation, it can be set to... , , , .

[0051] Step 3: By comparing the measured temperature with the reference temperature, identify the drift and missing measurement status of the measurement point, i.e., identify abnormal measurement points, and perform compensation and filling for abnormal measurement points.

[0052] The measured temperature at the actual measurement point is compared with the reference temperature to identify whether the measurement point is in a normal, drifting, or missing state, and the corresponding abnormal state quantity is updated as follows:

[0053] The residual between the measured temperature and the reference temperature at the bus connection point is tracked. If the residual continuously deviates within the most recent preset time window and the difference between the slope of the measured temperature change and the slope of the reference temperature change exceeds a threshold, it is identified as a drift state, and the measured temperature is retained. The drift state quantity is also used as an additional feature input into the lightweight gradient boosting regression correction model in subsequent steps. For the measurement points at the outgoing end of the compensated circuit breaker, if more than three sampling points are missing consecutively within a preset time interval, it is marked as a missing measurement state, and the reference temperature is used to compensate and fill in the missing points in subsequent steps.

[0054] In this embodiment, the drift state quantity of the q-th measuring point Update as follows:

[0055] .

[0056] in, This represents the average deviation between the measured temperature and the reference temperature within the most recent 15-minute window. , and These are the slopes of change between the measured temperature and the reference temperature, calculated using the difference between the first and last points within the most recent 15-minute window. In this embodiment, a smoothing coefficient can be used. Average deviation threshold Celsius, slope difference threshold Celsius / minute, maximum drift state quantity .

[0057] Step 4: Construct a virtual temperature measurement feature vector, and obtain the reference temperature correction amount through a lightweight gradient boosting regression correction model. Then, use the reference temperature correction amount to reconstruct the hidden node temperature and the overall hotspot temperature.

[0058] The following parameters were measured: busbar connection point compensation temperature, circuit breaker output terminal compensation temperature, missing measurement markers and drift status quantities corresponding to two measured points, ambient temperature, total harmonic current distortion rate, main circuit current, compensation current, maximum branch health status quantity, and overall hot spot reference temperature change rate over the past 15 minutes. The virtual temperature measurement feature vector is composed of the circuit breaker input temperature, module terminal temperature, and overall hot spot reference temperature output from the reference thermal network. .in, The unit is degrees Celsius per minute. This is the reference temperature for the overall hotspot.

[0059] For the three objectives of the inlet temperature of the compensation circuit breaker, the terminal temperature of the reactive power compensation module, and the overall hot spot temperature, a single-output lightweight gradient boosting regression sub-model is established respectively. This constructs a lightweight gradient boosting regression correction model, using the difference between the actual target temperature and the output temperature of the reference thermal network as the training label. Calculate the reference temperature correction, i.e.: lightweight gradient boosting regression correction model Inputting a virtual temperature measurement feature vector outputs a deviation correction, including temperature corrections for the incoming line of the compensation circuit breaker, the terminals of the reactive power compensation module, and the overall hot spot temperature, rather than direct temperature values ​​detached from physical constraints. Then press... To obtain the final estimated temperature, the reference temperature correction amount is added to the corresponding reference temperature obtained in step 2. The above data shows the temperature at the incoming terminal of the compensation circuit breaker, the temperature at the terminal of the reactive power compensation module, and the overall hot spot temperature.

[0060] Each single-output lightweight gradient boosting regression sub-model consists of 60 to 100 CART regression trees, with a maximum depth of 2 to 4, a learning rate of 0.02 to 0.10, and a minimum number of leaf node samples of 3 to 10.

[0061] In this embodiment, each sub-model uses 80 CART regression trees, a maximum depth of 3, a learning rate of 0.05, and a minimum number of leaf node samples of 5. The squared error is used as the loss function, and training stops when the validation set loss does not decrease for 10 consecutive rounds. The training samples are mainly measured data obtained by sampling the physical prototype for 1 minute under normal operating conditions, low power factor fluctuation conditions, harmonic pollution conditions, branch aging disturbance conditions, and abnormal switching conditions, supplemented by boundary condition augmentation data generated by the digital prototype. During the development phase, additional temperature sensors are temporarily deployed at the incoming end of the compensation circuit breaker, the wiring end of the reactive power compensation module, and the overall hot spot location to obtain training labels. The samples are divided into training set, validation set, and test set in proportions of 70%, 15%, and 15%, respectively.

[0062] Step 5: Based on the overall hot spot temperature of the reconstruction and its temperature rise trend, output the over-temperature warning result.

[0063] When the overall hotspot temperature obtained from the reconstruction An over-temperature warning will be issued when the temperature is greater than or equal to 68 degrees Celsius; or when the overall hotspot temperature is greater than or equal to 65.5 degrees Celsius and the estimated temperature change rate of the overall hotspot in the last 15 minutes is... When the temperature rises by 0.08 degrees Celsius or more per minute, a trend-based overheat warning is also issued. .

[0064] If the measured temperature has not yet reached the direct alarm temperature, but the hidden hot spot is close to the threshold, an early warning can still be issued.

[0065] Step 6: Verify and analyze the results using both physical and digital prototypes to generate hotspot locations, measurement point health status, and maintenance recommendations for use by the operation and maintenance system.

[0066] In the development phase of this embodiment, joint verification was conducted using both physical prototype testing and a system-level digital prototype built in a Python environment. Time-series verification was performed for normal operating conditions, low power factor fluctuation conditions, harmonic pollution conditions, and branch aging disturbance conditions. Physical prototype data was used for model calibration and real performance verification, while digital prototype data was used to supplement boundary conditions that were inconvenient to cover long-term and to illustrate the relative improvement trend of the solution.

[0067] In this embodiment, a progressive drift error is introduced to the bus connection point measurement point under harmonic pollution conditions; under branch aging disturbance conditions, a continuous missing measurement interval is introduced to the measurement point at the output end of the compensation circuit breaker, in order to verify the robustness of the present invention to the two types of anomalies, drift and missing measurement.

[0068] In the joint test set corresponding to this embodiment, the average absolute errors of the reconstruction of the temperature at the incoming terminal of the compensation circuit breaker, the terminal of the reactive power compensation module, and the overall hot spot temperature using the method of the present invention are approximately 0.1499 degrees Celsius, 0.1279 degrees Celsius, and 0.1499 degrees Celsius, respectively, while the average absolute error of the direct estimation using the reference thermal network is on the order of approximately 1.84 degrees Celsius. Furthermore, in the constructed real over-temperature trend test sample, the average early warning lead time of the method of the present invention compared to the direct reporting method at the measurement points is approximately 26.75 minutes. The above results demonstrate that the scheme of the present invention has good hot spot reconstruction and early warning capabilities under the set parameters and operating conditions.

[0069] like Figure 3 As shown, the hotspot temperature reconstructed by this invention can closely match the actual hotspot temperature changes; for example... Figure 4 As shown, even when there are missing measurement intervals at the measured points, this invention can still continuously output relatively smooth and accurate hotspot temperature estimation results; for example... Figure 5 As shown, the present invention outperforms methods that rely solely on a reference thermal network or rely solely on a limited number of measurement points in terms of both reconstruction error and early warning lead time.

[0070] Therefore, the intelligent metal integrated distribution box of the present invention can improve the continuity of hotspot perception and the timeliness of over-temperature early warning when the number of measurement points in the compensation circuit is limited.

[0071] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any way. Although the implementation process of the present invention has been described in detail above, those skilled in the art can still modify the technical solutions described in the foregoing examples or make equivalent substitutions for some of the technical features. All modifications and equivalent substitutions made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart metal integrated distribution box compensation loop hot spot virtual temperature measurement and early warning method, characterized in that, include: Step 1: Collect limited measurement point information and operating status information of the compensation loop; Step 2: Establish a reference thermal network model for the compensation loop, solve for the reference temperature of each hidden node in the model, and then obtain the overall hot spot reference temperature; Step 3: By comparing the measured points with the reference temperature, identify abnormal measuring points that are missing or drifting, and perform corresponding compensation or retention operations. Step 4: Construct a virtual temperature measurement feature vector, obtain a reference temperature correction amount through a temperature estimation model, and then use the reference temperature correction amount to reconstruct the hidden node temperature and the overall hot spot temperature; Step 5: Based on the overall hot spot temperature of the reconstruction and its temperature rise trend, output the over-temperature warning result.

2. The method as described in claim 1, characterized in that, In step 1, the limited measurement point information includes the temperature of the bus connection point of the compensation cabinet, the temperature of the output terminal of the compensation circuit breaker, the main circuit current, the compensation current, the total harmonic current distortion rate, the ambient temperature, and the current of each compensation branch, switching commands and feedback status output by the power distribution monitoring and compensation control unit of the distribution box. The operating status information is as follows The method of obtaining, among which The health status of the m-th compensation branch at time k; Let be the current deviation rate of the m-th compensated branch, used to characterize the degree of deviation of the measured branch current from the reference current; This is the switching anomaly indication for the m-th compensation branch, when a switching command is issued. With feedback status If the sampling periods are inconsistent for N consecutive periods, the value is 1; otherwise, the value is 0. λ1 and λ2 are the weighting coefficients of the branch current deviation term and the switching anomaly term, respectively. The upper limit threshold for the health status of branch circuits; This indicates the input quantity. Limited to the range Inside.

3. The method as described in claim 1, characterized in that, The reference thermal network model in step 2 includes the air node inside the distribution box, the bus connection point node of the compensation cabinet, the incoming terminal node of the compensation circuit breaker, the outgoing terminal node of the compensation circuit breaker, the wiring terminal node of the reactive power compensation module, the cavity node of the compensation controller, and the box shell node.

4. The method as described in claim 1, characterized in that, Each hidden node in step 2 satisfies the heat dissipation balance relationship: ,in, The sampling period; For the first Each node at time... temperature, For the first The heat generation power of each node, For the first The node and the first Equivalent thermal resistance between nodes For the first The equivalent heat transfer resistance between each node and the environment. For the first The equivalent heat capacity of each node, The ambient temperature; No. The heat capacity of each node is calculated according to Calculate; thermal resistance between adjacent nodes according to Calculation; thermal resistance between nodes and the environment according to Calculate; where, The mass of the conductor or structural component corresponding to the node. For the specific heat capacity of the material, This is the equivalent heat transfer path length. The equivalent thermal conductivity is For the equivalent heat transfer cross-sectional area, To consider the overall heat transfer coefficient, This refers to the heat exchange area.

5. The method as described in claim 4, characterized in that, The first The heating power of each node is related to the square of the compensation current and is modulated by the health status quantity and harmonic influence coefficient of the compensation branch.

6. The method as described in claim 1, characterized in that, In step 2, the initial equivalent contact resistance of each hidden node, including the busbar connection point of the compensation cabinet, the incoming terminal of the compensation circuit breaker, the outgoing terminal of the compensation circuit breaker, and the wiring terminal of the reactive power compensation module, is given an initial value based on the rated operating conditions and structural dimensions. Using calibration samples covering no-load, rated load, staged switching, harmonic disturbance, and ambient temperature change conditions, the square error between the measured temperature and the output temperature of the thermal network is minimized by the iterative least squares method with boundary constraints. The initial equivalent contact resistance, the equivalent thermal resistance between nodes, and the equivalent heat transfer thermal resistance between nodes and the environment are jointly corrected to solve for the reference temperature of each node and form a reference thermal network parameter table.

7. The method as described in claim 6, characterized in that, In step 2, the maximum value among the reference temperatures of the busbar connection point of the compensation cabinet, the incoming terminal of the compensation circuit breaker, the outgoing terminal of the compensation circuit breaker, and the wiring terminal of the reactive power compensation module is taken as the overall hot spot reference temperature.

8. The method as described in claim 1, characterized in that, In step 3, the residual between the measured temperature and the reference temperature at the bus connection point is tracked. If the residual continues to deviate within the most recent preset time window and the difference between the slope of the measured temperature change and the slope of the reference temperature change exceeds a threshold, it is identified as a drift state and the measured temperature is retained. The drift state quantity is also used as an additional feature to input the lightweight gradient boosting regression correction model in subsequent steps. For the measurement point at the outgoing end of the compensation circuit breaker, if more than 3 sampling points are missing consecutively within a preset time interval, it is marked as a missing measurement state and the reference temperature is used to compensate and fill it.

9. The method as described in claim 1, characterized in that, In step 4, the virtual temperature measurement feature vector consists of the following elements: compensation temperature of the busbar connection point of the compensation cabinet, compensation temperature of the output terminal of the compensation circuit breaker, missing and drift state quantities corresponding to the measured points, ambient temperature, total harmonic current distortion rate, main circuit current, compensation current, maximum branch health state quantity, overall hot spot reference temperature change rate of the most recent preset time period, and the circuit breaker input terminal temperature, module terminal temperature and overall hot spot reference temperature output by the reference thermal network model.

10. The method as described in claim 1, characterized in that, In step 4, the temperature estimation model consists of three single-output lightweight gradient boosting regression sub-models corresponding to the incoming terminal of the compensation circuit breaker, the terminal of the reactive power compensation module, and the overall hotspot temperature, respectively. Calculate the reference temperature correction amount, and according to The final estimated temperature is obtained, where The corresponding reference temperature output in step 2. This is the i-th single-output lightweight gradient boosting regression sub-model corresponding to the target temperature output; The lightweight gradient boosting regression sub-model consists of 60 to 100 CART regression trees, each with a maximum depth of 2 to 4, a learning rate of 0.02 to 0.10, and a minimum number of leaf node samples of 3 to 10. The squared error is used as the loss function. The training samples are mainly measured data of the physical prototype under normal operating conditions, low power factor fluctuation conditions, harmonic pollution conditions, branch aging disturbance conditions, and abnormal switching conditions, supplemented by boundary condition augmentation data generated by the digital prototype.

Images