A power distribution cabinet condensation active defense method and system

Abstract

The present application relates to the technical field of power distribution cabinet environment control, and specifically discloses a power distribution cabinet condensation active defense method and system, which comprises the following steps: collecting the running state of the anti-condensation equipment, and collecting the temperature and humidity measured values of multiple measuring points inside and outside the power distribution cabinet by using sensors; judging the data state and the sensor state, and determining the temperature and humidity actual values of each measuring point inside the power distribution cabinet; calculating the dew point temperature estimation value and estimation uncertainty of each grid inside the power distribution cabinet, and the average absolute humidity change rate inside the cabinet; calculating the safety margin of the grid; calculating the condensation risk index; and generating the control instruction of the anti-condensation equipment to control the anti-condensation equipment to work. The present application forms a complete closed-loop defense chain from the sensor level diagnosis, to the grid level prediction, to the decision level fusion, and to the equipment level control, and realizes the paradigm upgrade of the power distribution cabinet condensation defense from passive response to active defense.

Description

Technical Field

[0001] This invention relates to the field of environmental control technology for power distribution cabinets, and in particular to an active defense method and system for condensation in power distribution cabinets. Background Technology

[0002] Distribution cabinets are core equipment in power systems responsible for power distribution, control, and protection. Their internal temperature and humidity environment directly affects the safe operation of the equipment. When changes in temperature and humidity inside and outside the distribution cabinet cause the surface temperature of a local area inside the cabinet to fall below the air dew point temperature, condensation will occur in critical parts such as insulation components, busbars, and terminals. This can lead to reduced insulation strength and corona discharge, or even phase-to-phase short circuits, equipment burnout, or large-scale power outages.

[0003] Currently, there are three main technical approaches to anti-condensation technology for power distribution cabinets: The first approach is passive anti-condensation, which involves installing heaters, dehumidifiers, or fans in the distribution cabinet and controlling their start and stop using fixed thresholds. A common drawback of this approach is its crude control logic, which fails to differentiate between the severity and spatial distribution of condensation risks, often resulting in ineffective equipment operation, high energy consumption, and poor anti-condensation performance.

[0004] The second approach is predictive anti-condensation based on dew point temperature. This involves calculating the air dew point temperature and comparing it with the equipment surface temperature to determine condensation risk. For example, Chinese invention patent CN115811133A discloses a condensation prediction and control method, module, and system for switchgear. It calculates the dew point temperature by acquiring relative humidity, surface temperature, and dry-bulb temperature, and compares it with the surface temperature to determine condensation risk. Chinese invention patent CN121558125A discloses an online monitoring method and system for condensation risk in electrical cabinets. It calculates the dew point temperature using the Magnus formula, integrates wall temperature gradient and airflow state to calculate the equivalent dew point temperature, and then calculates the real-time safety margin and constructs a condensation risk trend index. However, the aforementioned patents have key shortcomings: First, the calculation of dew point temperature is based on deterministic physical formulas, which cannot quantify the uncertainty of the prediction results. Under conditions of high sensor noise, drastic changes in equipment status, or complex flow fields inside the cabinet, the output of the deterministic model may deviate from the actual situation, which can easily lead to misjudgment. Second, the condensation risk assessment lacks spatial resolution capabilities and is only characterized by a binary judgment of whether there is risk or not inside the cabinet or by a single index. It cannot identify the spatial distribution of risk inside the cabinet and is difficult to support differentiated and precise control.

[0005] The third approach is intelligent anti-condensation using multi-sensor fusion. This approach attempts to overcome the limitations of single-sensor solutions, but it also introduces new technical challenges: the detection and fault tolerance of sensor failures. Distribution cabinets operate in harsh environments, and temperature and humidity sensors inevitably drift, fail, or even malfunction completely. When a sensor in a multi-sensor system fails, using the fault data indiscriminately for condensation prediction can lead to localized false alarms or, more seriously, masking the true risk of condensation. For example, if a sensor in the cabinet detects persistently high humidity due to moisture, existing solutions may misjudge this as a condensation risk and incorrectly activate anti-condensation equipment; conversely, if a certain area does exhibit a condensation tendency but the sensor data is lowered due to a fault, existing solutions may miss the detection.

[0006] In addition, existing technologies have the following common shortcomings: the operating status of anti-condensation equipment itself will change the temperature and humidity distribution inside the cabinet, but existing prediction models do not take the operating status of the equipment as an input variable, resulting in a significant decrease in prediction accuracy under transient conditions; when there are contradictions between data from multiple sensors, there is a lack of systematic judgment logic to determine the value to be adopted; in the transition stage from suspicious to confirmed fault of sensor status, there is a lack of a smooth actual value transition mechanism, and directly switching the data source may cause a step jump in the actual value, resulting in control oscillation.

[0007] In summary, existing anti-condensation technologies for distribution cabinets suffer from key deficiencies such as the lack of sensor fault detection and fault tolerance mechanisms, inability to quantify prediction uncertainty, single risk assessment dimension, insufficient spatial resolution, and disconnect between equipment operating status and prediction models. There is an urgent need for an active defense method that can achieve fault detection and fault tolerance at the sensor level, grid-based uncertainty quantification at the spatial level, and multi-dimensional risk fusion judgment at the decision level. Summary of the Invention

[0008] The present invention aims to solve the above-mentioned problems. To this end, the present invention provides a method and system for active protection against condensation in power distribution cabinets.

[0009] This invention provides an active defense method for condensation in power distribution cabinets, the technical solution of which includes: S1: Collect the operating status of the anti-condensation equipment and use sensors to collect the measured temperature and humidity values ​​at multiple measuring points inside and outside the power distribution cabinet; among them, the operating status of the anti-condensation equipment includes heater power, dehumidifier duty cycle, and fan speed; S2: Utilize the spatial correlation of measuring points to perform cross-prediction of the measured temperature and humidity values ​​at each measuring point within the distribution cabinet, and determine the data status; simultaneously, based on physically constrained virtual sensors, calculate the virtual predicted temperature and humidity values ​​for each measuring point according to the measured temperature, measured humidity, and the operating status of the condensation equipment, and determine the sensor status; based on the data status and sensor status, determine the actual temperature and humidity values ​​at each measuring point within the distribution cabinet; S3: Divide the internal space of the distribution cabinet into multiple three-dimensional grids, and calculate the estimated dew point temperature and estimation uncertainty of each grid based on the actual temperature and humidity values ​​of each measuring point inside the distribution cabinet, the measured temperature values ​​of the measuring points outside the distribution cabinet, and the operating status of the anti-condensation equipment; at the same time, calculate the average absolute humidity change rate inside the cabinet. S4: Calculate the safety margin of the mesh based on the estimated dew point temperature, the estimation uncertainty, and the mesh surface temperature; S5: Based on the safety margin and estimated uncertainty, determine the risk grid and uncertainty grid; combine the average absolute humidity change rate inside the cabinet to calculate the condensation risk index; S6: Generate control instructions for the anti-condensation equipment based on the condensation risk index, and control the operation of the anti-condensation equipment.

[0010] Furthermore, in step S2, the measured values ​​of multiple sensors with the highest correlation are used to perform weighted prediction to obtain the anomaly reconstruction prediction value of the target sensor. If the absolute value of the deviation between the predicted value and the measured value exceeds the deviation threshold, the data status of the measured value is abnormal. The measured values ​​include measured temperature and measured humidity; the anomaly reconstruction prediction values ​​include anomaly reconstruction prediction values ​​for temperature and anomaly reconstruction prediction values ​​for humidity.

[0011] Furthermore, the correlation between sensors is calculated based on historical data using the Pearson correlation coefficient.

[0012] Furthermore, in step S2, the physical constraint virtual sensor adopts a regression model with a three-layer fully connected structure, taking the measured temperature and humidity values ​​of other sensors and the operating status of the anti-condensation equipment as inputs, and outputting the virtual predicted temperature and humidity values ​​of the target sensor. Each sensor in the cabinet corresponds to a physical constraint virtual sensor.

[0013] Furthermore, if the absolute value of the deviation between the virtual predicted value and the measured value exceeds the dynamic threshold, the sensor status is suspected of being faulty; if it exceeds the dynamic threshold for multiple consecutive cycles, the sensor status is faulty. The virtual forecast values ​​include virtual temperature forecast values ​​and virtual humidity forecast values.

[0014] Furthermore, the formula for calculating the dynamic threshold is: in, The dynamic threshold at time t, For smoothing coefficients, The dynamic threshold at time t-1 It is the absolute value of the deviation between the virtual predicted value and the measured value at time t-1.

[0015] Furthermore, when the data status is normal, the measured value is taken as the actual value; When the data status is abnormal but the sensor status is normal, the measured value is taken as the actual value. When the data status is abnormal and the sensor status is suspected of being faulty, the actual value is calculated by weighting the abnormal reconstruction prediction value and the virtual prediction value. When the data status is abnormal and the sensor status is faulty, the virtual predicted value is used as the actual value.

[0016] Furthermore, in step S4, when the safety margin of the grid is less than the trigger threshold, the estimated uncertainty is greater than the uncertainty threshold, or the periodic verification condition is reached, a fine calculation of thermal-humid coupling is triggered to update the dew point temperature estimate and the estimated uncertainty of the grid.

[0017] Furthermore, in step S5, the formula for calculating the condensation risk index is: in, This is the condensation risk index. To obtain the maximum value, The minimum safety margin in the risk grid. As the trigger threshold, The average absolute humidity change rate inside the cabinet. The threshold for the rate of change of humidity. For the number of risk grids, For an uncertain number of grid cells, The total number of grid cells. For safety margin weighting, As weighted by the rate of change of humidity, For risk grid weights, The grid weights are uncertain.

[0018] This invention also provides an active condensation prevention system for power distribution cabinets, the technical solution of which is as follows: including: The data acquisition module is used to collect the operating status of the anti-condensation equipment and to collect the measured temperature and humidity values ​​at multiple measuring points inside and outside the power distribution cabinet using sensors. The operating status of the anti-condensation equipment includes heater power, dehumidifier duty cycle, and fan speed. The data verification module is used to perform cross-prediction of the measured temperature and humidity values ​​of each measuring point in the distribution cabinet by utilizing the spatial correlation of the measuring points, and to determine the data status. At the same time, based on the physical constraint virtual sensor, it calculates the virtual predicted temperature and humidity values ​​of each measuring point according to the measured temperature, measured humidity values ​​and the operating status of the condensation equipment, and determines the sensor status. Based on the data status and sensor status, it determines the actual temperature and humidity values ​​of each measuring point in the distribution cabinet. The dew point temperature estimation module is used to divide the internal space of the distribution cabinet into multiple three-dimensional grids, and calculate the estimated dew point temperature and estimation uncertainty of each grid based on the actual temperature and humidity values ​​of each measuring point inside the distribution cabinet, the measured temperature values ​​of the measuring points outside the distribution cabinet, and the operating status of the anti-condensation equipment; at the same time, it calculates the average absolute humidity change rate inside the cabinet. The safety margin calculation module is used to calculate the safety margin of the mesh based on the estimated dew point temperature, the estimation uncertainty, and the mesh surface temperature. The condensation risk index calculation module is used to determine the risk grid and uncertainty grid based on the safety margin and estimated uncertainty; and calculates the condensation risk index by combining the average absolute humidity change rate inside the cabinet. The control module generates control commands for the anti-condensation equipment based on the condensation risk index, and controls the operation of the anti-condensation equipment.

[0019] The above-described one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: 1. This invention effectively distinguishes between sensor fault detection and abnormal condensation data, improving the system's robustness and data reliability under sensor fault conditions. This invention employs two independent detection mechanisms: an anomaly reconstruction detection path and a virtual sensor detection path. The former utilizes the spatiotemporal correlation between multiple sensors for cross-prediction to determine the data state of the measured value, while the latter combines the operating status of the anti-condensation equipment to perform self-calibration prediction on the target sensor to determine its state. Through these two states, this invention can differentiate between sensor faults and genuine local condensation events, avoiding false alarms and missed alarms caused by confusing the two in existing technologies. During the transition phase where the data state is abnormal and the sensor state is suspected of being faulty, a weighted transition mechanism gradually transitions the output actual value from the anomaly reconstruction prediction value to the virtual prediction value, avoiding the abrupt jump in actual value caused by data source switching at the moment of fault confirmation and the impact on downstream control links.

[0020] 2. This invention incorporates prediction uncertainty into the condensation risk assessment of distribution cabinets, achieving a leap from deterministic judgment to probabilistic decision-making, significantly improving the reliability of risk assessment and the ability to make conservative decisions. This invention uses a Gaussian process regression model to predict the dew point temperature of each grid, outputting not only the estimated dew point temperature value but also the estimated uncertainty (prediction standard deviation). When calculating the grid safety margin, the uncertainty is incorporated into the safety margin formula as a penalty term, improving the system's adaptability to complex operating conditions.

[0021] 3. This invention constructs a multi-dimensional condensation risk index, solving the problem that existing technologies have a single risk representation dimension and cannot comprehensively reflect the overall picture of condensation risk. The calculation of the condensation risk index integrates four independent dimensions: the minimum safety margin in the risk grid (representing the severity of the risk), the number of risk grids (representing the spatial breadth of the risk), the number of uncertain grids (representing the knowledge uncertainty of the current prediction), and the rate of change of the average absolute humidity inside the cabinet (representing the temporal evolution trend of the risk), taking into account both the comprehensiveness of risk assessment and the operability of control decisions.

[0022] 4. This invention uses the operating status of the anti-condensation equipment as an explicit input to the prediction model, improving the prediction accuracy under transient conditions. In both the virtual sensor detection path and dew point temperature prediction processes, the operating status of the anti-condensation equipment, such as heater power, dehumidifier duty cycle, and fan speed, is used as model input. This allows the model to perceive the active influence of equipment start-up, shutdown, and power switching on the temperature and humidity field inside the cabinet, overcoming the shortcomings of existing technologies where the model and equipment status are disconnected, leading to decreased prediction accuracy under transient conditions.

[0023] 5. This invention forms a complete closed-loop defense chain, from sensor-level diagnosis to grid-level prediction, to decision-level fusion, and finally to equipment-level control. Starting with raw data acquisition in S1, the invention obtains reliable actual values ​​through sensor fault tolerance in S2, obtains spatially distributed dew point temperature estimates and uncertainties through gridded GPR prediction in S3, achieves conservative decision-making through safety margin calculation in S4, outputs a condensation risk index through multi-dimensional risk fusion in S5, and finally transforms these into hierarchical control commands for anti-condensation equipment in S6. Each step is organically connected and progressively advanced, realizing a paradigm shift from passive response to active defense in distribution cabinet condensation protection.

[0024] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a flowchart of the method provided by the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but should not be used to limit the scope of this invention.

[0028] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0029] The following is combined with Figure 1 The present invention will be further described in detail below, including a method and system for active protection against condensation in power distribution cabinets: In this embodiment, as Figure 1 As shown, a method for active protection against condensation in a power distribution cabinet is provided, including the following steps: S1: Collect the operating status of the anti-condensation equipment and use sensors to collect the actual temperature and humidity values ​​at multiple measuring points inside and outside the power distribution cabinet.

[0030] In this embodiment, a temperature and humidity sensor is installed at each measuring point. Multiple temperature and humidity sensors are deployed inside and outside the distribution cabinet to collect temperature and humidity data from multiple measuring points in real time. Generally, a large number of temperature and humidity sensors are deployed inside the distribution cabinet, and the measuring point locations need to cover the cabinet door, busbars, and incoming / outgoing line areas, while only one sensor is deployed outside the distribution cabinet. Commonly used anti-condensation devices include heaters, dehumidifiers, and fans.

[0031] In this embodiment, eight temperature and humidity sensors are deployed, seven of which are located inside the cabinet: two near the inner wall of the cabinet door, two near the busbar insulation, one in the top area, one in the bottom cable entry / exit area, and one in the side wall area; one temperature and humidity sensor is installed outside the cabinet. Each temperature and humidity sensor is connected to the main control unit via an RS485 bus, with a sampling period of one second. The start / stop status and power status of the heater, dehumidifier, and fan are also collected by the main control unit.

[0032] After the main control unit collects the data, it aligns the timestamps of each data point.

[0033] S2: Utilize the spatial correlation of measuring points to perform cross-prediction of the measured temperature and humidity values ​​at each measuring point within the distribution cabinet, and determine the data status; simultaneously, based on physically constrained virtual sensors, calculate the virtual predicted temperature and humidity values ​​for each measuring point according to the measured temperature, measured humidity, and the operating status of the condensation equipment, and determine the sensor status; based on the data status and sensor status, determine the actual temperature and humidity values ​​at each measuring point within the distribution cabinet.

[0034] This embodiment sets up two detection paths: an anomaly reconstruction detection path and a virtual sensor detection path.

[0035] (1) Anomaly reconstruction detection path: Based on the measured temperature and humidity values ​​of each measuring point, the spatial correlation of the measuring points is used to cross-predict the measured temperature and humidity values ​​of each measuring point in the distribution cabinet, and the anomaly reconstruction prediction values ​​of temperature and humidity of each measuring point are calculated. Then, based on the anomaly reconstruction prediction value, the measured value and the deviation threshold, the data status of the measured value is determined. The data status includes normal and abnormal.

[0036] Spatial correlation is represented by a spatiotemporal correlation coefficient matrix among multiple sensors, constructed based on historical data. The elements of the matrix are calculated using the Pearson correlation coefficient. In this embodiment, this matrix is ​​updated every 10 minutes. If any sensor malfunctions, the spatiotemporal correlation coefficient matrix is ​​no longer updated.

[0037] When calculating the anomaly reconstruction prediction value of any sensor, the actual values ​​of the multiple sensors with the highest correlation are selected for weighted prediction based on the spatiotemporal correlation coefficient matrix. In this embodiment, for any temperature and humidity sensor inside the cabinet, the current actual values ​​of the three temperature and humidity sensors with the highest correlation inside the cabinet are used for weighted prediction. The calculation formula can be expressed as: in, The abnormal reconstruction prediction value of temperature and humidity sensor k includes the abnormal reconstruction prediction value of temperature and the abnormal reconstruction prediction value of humidity. The two values ​​are calculated separately. The measured values ​​are those of the highly correlated temperature and humidity sensor m, including the measured temperature and humidity values. Let be the correlation coefficient between temperature and humidity sensor k and temperature and humidity sensor m, which is an element in the spatiotemporal correlation coefficient matrix. Temperature and humidity sensors identified as faulty cannot be used as highly correlated temperature and humidity sensors in the calculation of anomaly reconstruction prediction values.

[0038] Finally, the absolute value of the deviation between the anomaly reconstruction prediction value and the measured value is calculated. If it exceeds the deviation threshold, the measured value is suspected of being abnormal, and the data status is marked as abnormal; otherwise, the data status is marked as normal. In this embodiment, the temperature deviation threshold is set to 1.0℃, and the humidity deviation threshold is set to 5%RH.

[0039] It should be noted that temperature and humidity need to be calculated and determined separately.

[0040] (2) Virtual sensor detection path: A regression model is constructed as a physical constraint virtual sensor. For a target sensor, the actual temperature and humidity values ​​of other sensors and the operating status of the anti-condensation equipment are used as inputs, and the virtual predicted values ​​of temperature and humidity of the target sensor are output. The model output follows the physical laws of heat transfer and humid air migration in the cabinet.

[0041] In this embodiment, the regression model uses a three-layer fully connected layer with ReLU activation function. During training, temperature and humidity data from the past 48 hours and the operating status of anti-condensation equipment are collected as the training set. The Adam optimizer is used with an initial learning rate of 0.001 and a batch size of 256. The mean squared error loss function converges to below 0.015 after 120 iterations.

[0042] In this embodiment, for any temperature and humidity sensor in a cabinet, the measured temperature and humidity values ​​from the other seven temperature and humidity sensors, along with the heater power, dehumidifier duty cycle, and fan speed, are input into the corresponding physical constraint virtual sensor to calculate the virtual predicted temperature and humidity values ​​for that sensor. The operating status of the anti-condensation equipment includes heater power, dehumidifier duty cycle, and fan speed. Each temperature and humidity sensor in a cabinet corresponds to one physical constraint virtual sensor.

[0043] The outputs of the physically constrained virtual sensors (virtual predicted values ​​of temperature and humidity) serve as the reference baseline for the self-calibration of the temperature and humidity sensors. The absolute value of the deviation between the virtual predicted values ​​and the measured values ​​is calculated, and then the state of the temperature and humidity sensors is determined based on dynamic thresholds, including normal, suspected fault, and fault. In this embodiment, temperature and humidity need to be calculated and determined separately. If one or both exceed the corresponding dynamic threshold, it is recorded as exceeding the threshold, and the sensor state is marked as suspected fault. If the threshold is exceeded for multiple consecutive cycles (5 cycles in this embodiment), the state of the temperature and humidity sensors is marked as fault. For example, when the absolute value of the deviation between the virtual predicted temperature value and the measured temperature value exceeds the dynamic threshold of temperature, regardless of whether the absolute value of the deviation between the virtual predicted temperature value and the measured humidity value exceeds the dynamic threshold of humidity, this cycle will be recorded as exceeding the threshold.

[0044] In this embodiment, the dynamic threshold is determined using an exponentially weighted moving average method, and the calculation formula is as follows: in, The dynamic threshold at time t, In this embodiment, the smoothing coefficient is used. , Let be the absolute value of the deviation between the virtual predicted value and the measured value at time t-1. The dynamic threshold at time t-1. Initial value, dynamic threshold. The sensor's factory accuracy specifications are set as follows: temperature ±0.5℃, humidity ±3%RH.

[0045] After determining the data status (normal or abnormal) and sensor status (normal, suspected fault, and fault), the actual value is determined based on the measured value, the anomaly reconstructed predicted value, and the virtual predicted value. The specific determination rules are as follows: When the data status is normal, and the sensor status is normal, suspected fault, or fault, the measured value is taken as the actual value. A normal data status indicates that the sensor readings are highly consistent with other sensors, suggesting the sensor is likely functioning correctly. A suspected fault sensor status may be caused by transient disturbances in the local operating condition. A fault sensor status may be due to insufficient generalization of the model under that operating condition.

[0046] When the data status is abnormal but the sensor status is normal, the measured value is taken as the actual value.

[0047] When the data status is abnormal and the sensor status is suspected of being faulty, the actual value is calculated by weighting the abnormal reconstruction prediction value and the virtual prediction value. In this situation, the sensor is highly suspicious. To avoid a step jump in the actual value at the moment of fault confirmation, this embodiment designs a weighted transition mechanism, so that the output actual value gradually transitions from the abnormal reconstruction prediction value to the virtual prediction value as the threshold is exceeded. The calculation formula is: in, This is the actual value. As transitional weights, For abnormal reconstruction prediction values, This is a virtual predicted value. In this embodiment, the transition weight gradually increases from 0.2 to 1 (increasing by 0.2 each time).

[0048] When the data status is abnormal and the sensor status is faulty, the virtual predicted value is used as the actual value, and a sensor fault alarm is generated simultaneously. Fault recovery mechanism: When the sensor is in a faulty state, if the data status is normal and the virtual sensor detection path does not exceed the threshold for 30 consecutive sampling cycles, the sensor is considered to have recovered.

[0049] S3: Divide the internal space of the distribution cabinet into multiple three-dimensional grids, and calculate the estimated dew point temperature and estimation uncertainty of each grid based on the actual temperature and humidity values ​​of each measuring point inside the distribution cabinet, the measured temperature values ​​of the measuring points outside the distribution cabinet, and the operating status of the anti-condensation equipment; at the same time, calculate the average absolute humidity change rate inside the cabinet.

[0050] This embodiment divides the interior of the distribution cabinet into a door sealing area, a busbar insulation area, a bottom inlet / outlet area, a side wall area, and a ventilation interface area, and further divides each area into multiple three-dimensional grids. For example, if the internal space of the distribution cabinet is 1200mm × 800mm × 400mm, it can be divided into a 50mm × 50mm × 50mm grid.

[0051] This embodiment calculates the estimated dew point temperature and estimation uncertainty of the grid using a dew point temperature prediction model.

[0052] The dew point temperature prediction model can employ a Gaussian process regression model, a random forest regression model, or a lightweight neural network model. In this embodiment, the dew point temperature prediction model uses a Gaussian process regression model. The inputs include the grid location, the location of each measuring point within the distribution cabinet and their actual temperature and humidity values, heater power, dehumidifier duty cycle, fan speed (anti-condensation equipment operating status), and the measured temperature values ​​of measuring points outside the distribution cabinet. This process calculates the estimated dew point temperature and its uncertainty for each grid. The Gaussian process regression model uses a squared exponential kernel function, and the hyperparameters are determined by maximizing the logarithmic marginal likelihood (LML). Optimization uses the L-BFGS algorithm, and the initial length scale is set according to the empirical range of each input dimension. The estimated dew point temperature for each grid is the predicted mean of the Gaussian process regression, and the uncertainty is the predicted standard deviation. The Gaussian process regression model can use offline training of induced points or local sub-models; in the online phase, only prediction and uncertainty calculations are performed, and a complete Gaussian process retraining is not performed on all samples in the main control loop.

[0053] In this embodiment, the actual humidity values ​​of each measuring point in the distribution cabinet are converted into absolute humidity. Then, the three-dimensional spatial kriging interpolation algorithm is used to calculate the absolute humidity interpolation estimate of the center point of each grid. The average absolute humidity in the cabinet is then calculated, and finally the rate of change of the average absolute humidity in the cabinet is calculated.

[0054] S4: Calculate the safety margin of the mesh based on the estimated dew point temperature, the estimation uncertainty, and the mesh surface temperature.

[0055] In this embodiment, the formula for calculating the safety margin of the mesh is: in, This represents the safety margin of grid g. This represents the surface temperature of grid g. This represents the estimated dew point temperature of grid g. This represents the estimation uncertainty of grid g. This is the uncertainty amplification factor. To provide a preset safety margin. In this embodiment, =2, =0.5℃. The surface temperature of the grid was obtained by interpolating the actual temperature values ​​of each measuring point inside the cabinet using three-dimensional kriging interpolation.

[0056] For any grid, when the grid's safety margin is less than the trigger threshold, the estimated uncertainty is greater than the uncertainty threshold, or the periodic verification condition is met, a fine calculation of the thermal-humid coupling is triggered to update the grid's dew point temperature estimate and estimated uncertainty. A more refined dew point temperature estimate and estimated uncertainty for the grid are calculated, and then the safety margin is calculated.

[0057] In this embodiment, precise calculations can be performed using simplified thermal-humidity coupling equations, finite difference models, finite element offline model lookup tables, or calibrated reduced-order models. The trigger threshold is 0.5℃; the uncertainty threshold is 3.0℃; and the verification period is 30 minutes.

[0058] S5: Based on the safety margin and estimated uncertainty, determine the risk grid and uncertainty grid; combine the average absolute humidity change rate inside the cabinet to calculate the condensation risk index.

[0059] The specific process for this step is as follows: S5.1: Based on the safety margin and estimated uncertainty, determine whether a grid is a risky grid or an uncertain grid. The determination rule is: when the safety margin is less than the trigger threshold, the grid is a risky grid. When the estimated uncertainty is greater than the uncertainty threshold, the grid is an uncertain grid. Count the number of risky grids and the number of uncertain grids.

[0060] S5.2: Extract the minimum value from the safety margins of all risk grids to obtain the minimum safety margin in the risk grid.

[0061] S5.3: The condensation risk index is calculated based on the minimum safety margin in the risk grid, the number of risk grids, the number of uncertain grids, and the average absolute humidity change rate inside the cabinet. The calculation formula is: in, This is the condensation risk index. To obtain the maximum value, The minimum safety margin in the risk grid. As the trigger threshold, The average absolute humidity change rate inside the cabinet. The threshold for the rate of change of humidity. For the number of risk grids, For an uncertain number of grid cells, The total number of grid cells. For safety margin weighting, As weighted by the rate of change of humidity, For risk grid weights, The grid weights are uncertain. In this embodiment, = , =0.5, =0.3, =0.1, =0.1.

[0062] Furthermore, for different areas within the cabinet (cabinet door sealing area, busbar insulation area, bottom inlet / outlet area, side wall area, and ventilation interface area), the trigger threshold and humidity change rate threshold are not fixed values, but are dynamically adjusted according to the area's location. For areas where temperature and humidity changes more frequently and drastically, such as the cabinet door sealing area and ventilation interface area, the threshold is increased by 30% from the baseline value; for other areas, such as the busbar insulation area, bottom inlet / outlet area, and side wall area, the threshold is decreased by 20% from the baseline value.

[0063] S6: Generate control instructions for the anti-condensation equipment based on the condensation risk index, and control the operation of the anti-condensation equipment.

[0064] This embodiment divides the condensation risk into four condensation risk levels based on the condensation risk index, and sets different control commands for anti-condensation equipment accordingly. The condensation risk level is high, and anti-condensation equipment is operating at high power. The condensation risk level is medium risk, and the anti-condensation equipment is operating at medium power. The condensation risk level is low, and the anti-condensation equipment is operating at low power. The condensation risk level is no risk, and the anti-condensation equipment is not working.

[0065] In this embodiment, the power control commands for the heater are: high risk: 500W, medium risk: 300W, and low risk: 150W; the control commands for the dehumidifier are: high risk: 80% duty cycle, medium risk: 60% duty cycle, and low risk: 40% duty cycle; and the speed control commands for the fan are: high risk: full speed, medium risk: half speed, and low risk: 1 / 4 speed.

[0066] The startup sequence for the anti-condensation equipment is: heater, dehumidifier, fan. During operation, the condensation risk level is assessed every 5 minutes, meaning steps S1 to S6 are repeated every 5 minutes. If necessary, the heater, dehumidifier, and fan will be turned off in sequence, and the system will return to normal monitoring.

[0067] This embodiment verifies the effectiveness of the method (periodic verification was not enabled in this verification): The test platform uses an embedded controller based on ARM Cortex-A72, with a main frequency of 1.5GHz, 2GB RAM, and running a Linux operating system. The simulation model of the distribution cabinet is 1200mm×800mm×400mm in size, divided into a 50mm×50mm×50mm grid, with a total of 8 temperature and humidity sensors (2 near the inner wall of the cabinet door, 2 near the busbar insulation, 1 in the top area, 1 in the bottom inlet / outlet area, 1 in the side wall area, and 1 outside the cabinet). Each sensor communicates with the main control unit via an RS485 bus, with a sampling frequency of 1Hz. A 48-hour continuously changing temperature and humidity scenario is set up, covering typical operating conditions such as diurnal temperature and humidity fluctuations, temperature fluctuations inside the cabinet caused by changes in equipment load, and the intrusion of external humid air introduced by the brief opening of the cabinet door.

[0068] The experimental results are as follows: (1) The proportion of the grid that triggers the solution of the partial differential equation of thermal-humid coupling is 7.8%; (2) The average absolute error of dew point prediction under single-point sensor failure is 0.28℃.

[0069] In this method, only about 7.8% of the 3072 grids triggered fine-grained calculations of thermal-humidity coupling; the remaining approximately 92.2% of the grids obtained dew point values ​​that met the control accuracy requirements through approximate estimation using a dew point temperature prediction model. It should be noted that this 7.8% trigger rate depends on the preset trigger threshold and the temperature and humidity fluctuations during the test conditions; the trigger rate will vary depending on the cabinet and operating conditions. A single-point fault test of the sensor was artificially injected in the 12th hour: a constant bias of 2.5℃ was superimposed on the sensor output located on the inner wall of the front door to simulate an output drift fault. This method automatically detected the anomaly in the 8th sampling cycle after the fault occurred and switched to the virtual sensor output. The average absolute error of the dew point prediction was 0.28℃ over 48 hours.

[0070] This embodiment also provides an active condensation prevention system for power distribution cabinets, the technical solution of which is as follows: including: The data acquisition module is used to collect the operating status of the anti-condensation equipment and to collect the measured temperature and humidity values ​​at multiple measuring points inside and outside the power distribution cabinet using sensors. The data verification module is used to perform cross-prediction of the measured temperature and humidity values ​​of each measuring point in the distribution cabinet by utilizing the spatial correlation of the measuring points, and to determine the data status. At the same time, based on the physical constraint virtual sensor, it calculates the virtual predicted temperature and humidity values ​​of each measuring point according to the measured temperature, measured humidity values ​​and the operating status of the condensation equipment, and determines the sensor status. Based on the data status and sensor status, it determines the actual temperature and humidity values ​​of each measuring point in the distribution cabinet. The dew point temperature estimation module is used to divide the internal space of the distribution cabinet into multiple three-dimensional grids, and calculate the estimated dew point temperature and estimation uncertainty of each grid based on the actual temperature and humidity values ​​of each measuring point inside the distribution cabinet, the measured temperature values ​​of the measuring points outside the distribution cabinet, and the operating status of the anti-condensation equipment; at the same time, it calculates the average absolute humidity change rate inside the cabinet. The safety margin calculation module is used to calculate the safety margin of the mesh based on the estimated dew point temperature, the estimation uncertainty, and the mesh surface temperature. The condensation risk index calculation module is used to determine the risk grid and uncertainty grid based on the safety margin and estimated uncertainty; and calculates the condensation risk index by combining the average absolute humidity change rate inside the cabinet. The control module generates control commands for the anti-condensation equipment based on the condensation risk index, and controls the operation of the anti-condensation equipment.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for active protection against condensation in power distribution cabinets, characterized in that, include: S1: Collect the operating status of the anti-condensation equipment and use sensors to collect the measured temperature and humidity values ​​at multiple measuring points inside and outside the power distribution cabinet; among them, the operating status of the anti-condensation equipment includes heater power, dehumidifier duty cycle, and fan speed; S2: Utilize the spatial correlation of measuring points to perform cross-prediction of the measured temperature and humidity values ​​at each measuring point within the distribution cabinet, and determine the data status; simultaneously, based on physically constrained virtual sensors, calculate the virtual predicted temperature and humidity values ​​for each measuring point according to the measured temperature, measured humidity, and the operating status of the condensation equipment, and determine the sensor status; based on the data status and sensor status, determine the actual temperature and humidity values ​​at each measuring point within the distribution cabinet; S3: Divide the internal space of the distribution cabinet into multiple three-dimensional grids, and calculate the estimated dew point temperature and estimation uncertainty of each grid based on the actual temperature and humidity values ​​of each measuring point inside the distribution cabinet, the measured temperature values ​​of the measuring points outside the distribution cabinet, and the operating status of the anti-condensation equipment; at the same time, calculate the average absolute humidity change rate inside the cabinet. S4: Calculate the safety margin of the mesh based on the estimated dew point temperature, the estimation uncertainty, and the mesh surface temperature; S5: Based on the safety margin and estimated uncertainty, determine the risk grid and uncertainty grid; combine the average absolute humidity change rate inside the cabinet to calculate the condensation risk index; S6: Generate control instructions for the anti-condensation equipment based on the condensation risk index, and control the operation of the anti-condensation equipment.

2. The active condensation prevention method for power distribution cabinets as described in claim 1, characterized in that, In step S2, the measured values ​​of multiple sensors with the highest correlation are used to perform weighted prediction to obtain the anomaly reconstruction prediction value of the target sensor. If the absolute value of the deviation between the predicted value and the measured value exceeds the deviation threshold, the data status of the measured value is abnormal. The measured values ​​include measured temperature and measured humidity; the anomaly reconstruction prediction values ​​include anomaly reconstruction prediction values ​​for temperature and anomaly reconstruction prediction values ​​for humidity.

3. The active condensation prevention method for power distribution cabinets as described in claim 2, characterized in that, The correlation between sensors is calculated based on historical data using the Pearson correlation coefficient.

4. The active condensation prevention method for power distribution cabinets as described in claim 2, characterized in that, In step S2, the physical constraint virtual sensor adopts a regression model with a three-layer fully connected structure. It takes the actual temperature and humidity values ​​of other sensors and the operating status of the anti-condensation equipment as inputs and outputs the virtual temperature and humidity prediction values ​​of the target sensor. Each sensor in the cabinet corresponds to a physical constraint virtual sensor.

5. A method for active protection against condensation in a power distribution cabinet as described in claim 2 or 4, characterized in that, If the absolute value of the deviation between the virtual predicted value and the measured value exceeds the dynamic threshold, the sensor status is suspected of being faulty; if it exceeds the dynamic threshold for multiple consecutive periods, the sensor status is faulty. The virtual forecast values ​​include virtual temperature forecast values ​​and virtual humidity forecast values.

6. The active condensation prevention method for power distribution cabinets as described in claim 5, characterized in that, The formula for calculating the dynamic threshold is: in, The dynamic threshold at time t, For smoothing coefficients, The dynamic threshold at time t-1 It is the absolute value of the deviation between the virtual predicted value and the measured value at time t-1.

7. The active condensation prevention method for power distribution cabinets as described in claim 5, characterized in that, When the data status is normal, the measured value is taken as the actual value; When the data status is abnormal but the sensor status is normal, the measured value is taken as the actual value. When the data status is abnormal and the sensor status is suspected of being faulty, the actual value is calculated by weighting the abnormal reconstruction prediction value and the virtual prediction value. When the data status is abnormal and the sensor status is faulty, the virtual predicted value is used as the actual value.

8. The active condensation prevention method for power distribution cabinets as described in claim 1, characterized in that, In step S4, when the safety margin of the grid is less than the trigger threshold, the estimated uncertainty is greater than the uncertainty threshold, or the periodic verification condition is reached, the thermal-humid coupling fine calculation is triggered to update the dew point temperature estimate and the estimated uncertainty of the grid.

9. The active condensation prevention method for power distribution cabinets as described in claim 1, characterized in that, In step S5, the formula for calculating the condensation risk index is: in, This is the condensation risk index. To obtain the maximum value, The minimum safety margin in the risk grid. As the trigger threshold, The average absolute humidity change rate inside the cabinet. The threshold for the rate of change of humidity. For the number of risk grids, For an uncertain number of grid cells, The total number of grid cells. For safety margin weighting, As weighted by the rate of change of humidity, For risk grid weights, The grid weights are uncertain.

10. A power distribution cabinet condensation active defense system, characterized in that, A method for implementing an active condensation prevention method for a distribution cabinet as described in any one of claims 1 to 9, comprising: The data acquisition module is used to collect the operating status of the anti-condensation equipment and to collect the measured temperature and humidity values ​​at multiple measuring points inside and outside the power distribution cabinet using sensors. The operating status of the anti-condensation equipment includes heater power, dehumidifier duty cycle, and fan speed. The data verification module is used to perform cross-prediction of the measured temperature and humidity values ​​of each measuring point in the distribution cabinet by utilizing the spatial correlation of the measuring points, and to determine the data status. At the same time, based on the physical constraint virtual sensor, it calculates the virtual predicted temperature and humidity values ​​of each measuring point according to the measured temperature, measured humidity values ​​and the operating status of the condensation equipment, and determines the sensor status. Based on the data status and sensor status, it determines the actual temperature and humidity values ​​of each measuring point in the distribution cabinet. The dew point temperature estimation module is used to divide the internal space of the distribution cabinet into multiple three-dimensional grids, and calculate the estimated dew point temperature and estimation uncertainty of each grid based on the actual temperature and humidity values ​​of each measuring point inside the distribution cabinet, the measured temperature values ​​of the measuring points outside the distribution cabinet, and the operating status of the anti-condensation equipment; at the same time, it calculates the average absolute humidity change rate inside the cabinet. The safety margin calculation module is used to calculate the safety margin of the mesh based on the estimated dew point temperature, the estimation uncertainty, and the mesh surface temperature. The condensation risk index calculation module is used to determine the risk grid and uncertainty grid based on the safety margin and estimated uncertainty; and calculates the condensation risk index by combining the average absolute humidity change rate inside the cabinet. The control module generates control commands for the anti-condensation equipment based on the condensation risk index, and controls the operation of the anti-condensation equipment.

Images