10kV ring main unit self-adaptive condensation eliminating device and humidity control method

Abstract

The invention belongs to the technical field of temperature and humidity control of power equipment, and particularly relates to a 10kV ring main unit self-adaptive condensation eliminating device and a humidity control method. The device comprises five modules: a power supply module uses solar energy and a control PT to control dual power supply to guarantee continuous power supply, a state detection module collects temperature, humidity and safety signals in a cabinet, a dehumidification execution module processes and discharges condensed water, a control module overall plans dehumidification start and stop and power supply switching, and a remote monitoring module supports data viewing and abnormity alarm. The control method comprises the following five steps: firstly, collecting data according to a dynamic sampling period, setting control for 5 minutes in rainfall or strong wind, calculating according to information entropy in other conditions, and filtering the data; combining historical data of federal learning processing, and using a control LSTM to control and an attention model to judge the condensation risk; then switching power supply, starting and stopping dehumidification and combining the aging state of the element; then carrying out transfer learning to judge result transmission data; and finally, according to a cabinet body structure coefficient adjustment strategy, condensation elimination and monitoring are realized, and energy consumption and standard reaching time are optimized.

Description

Technical Field

[0001] This invention belongs to the field of temperature and humidity control technology for power equipment, and particularly relates to an adaptive anti-condensation device and humidity control method for 10kV ring main units. Background Technology

[0002] 10kV ring main units are key core equipment in power distribution networks, widely used in outdoor power distribution lines, underground substations, industrial parks, and other scenarios. Their operational stability directly affects the reliability of power supply. Because ring main units operate in a closed environment, the internal equipment continuously generates heat during operation, creating a significant temperature difference with the outside environment. This easily leads to moisture accumulation inside the unit, resulting in condensation and even water accumulation. This not only severely degrades the insulation performance and corrodes the cabinet structure, but also makes it prone to partial discharge during current conduction, potentially causing the ring main unit to burn out and explode, seriously threatening personal and equipment safety. Currently, most ring main unit dehumidification devices are single-bay heaters, with each bay equipped with a separate unit. Furthermore, there is no low-voltage power supply inside the cabinet; the heaters rely on a PT (potential transformer) for power. If the PT fails, the heaters will directly cease functioning, failing to dehumidify in time and making it difficult to address the safety hazards posed by condensation.

[0003] Existing technologies for eliminating condensation in ring main units have several shortcomings, further limiting dehumidification efficiency and equipment reliability: data acquisition often uses fixed sampling periods, which cannot be dynamically adjusted according to external weather, and sensor drift is not effectively compensated, leading to distorted temperature and humidity data; condensation risk assessment models do not fully consider environmental interference, seasonal differences, and uneven temperature and humidity within the cabinet, resulting in a high rate of false positives and false negatives; power supply modes lack flexibility, with some devices relying on grid power having limited application in outdoor areas without a stable power grid, and power switching lacks an adaptive voltage regulation mechanism; dehumidification control does not consider component aging and ventilation conditions, leading to excessive energy consumption or incomplete dehumidification; historical data processing lacks an effective distributed learning mechanism, remote control has not formed a closed-loop optimization, and it is not adapted to differences in cabinet structure. Furthermore, some devices lack comprehensive safety monitoring and abnormal linkage alarm functions. Therefore, there is an urgent need for an adaptive condensation elimination device and control method that is adaptable to complex environments, eliminates reliance on a single power supply, and is precise and efficient. Summary of the Invention

[0004] To thoroughly address the prominent condensation hazards, reliance on a single power source, data distortion, and inefficient control issues in existing 10kV ring main units, this invention proposes an adaptive condensation elimination device and humidity control method for 10kV ring main units. Powered by solar energy and batteries, it dynamically samples and compensates for data to precisely control condensation. The specific technical solution is as follows:

[0005] An adaptive condensation elimination device for a 10kV ring main unit includes a control box and a solar charging controller. The controller and the solar charging controller are both installed inside the ring main unit and include solar panels, batteries, solar generators, dehumidifiers, and smoke detectors.

[0006] The solar panels are installed outside the ring main unit, while the battery, solar generator, dehumidifier, and smoke detector are installed inside the ring main unit.

[0007] The solar panel is electrically connected to the solar generator and signal-connected to the solar charge controller. The solar generator is electrically connected to the storage battery, which can power the ring main unit environmental monitoring and control box, dehumidifier, and smoke detector.

[0008] The dehumidifier and smoke detector are both connected to the control box via signal, and the control box can be controlled by a remote platform.

[0009] A humidity control method for an adaptive anti-condensation device in a 10kV ring main unit includes the following steps:

[0010] S1: Collect temperature and humidity data and safety status signals inside the cabinet according to the dynamic sampling period, and transmit the standardized data to the control module after filtering with sensor drift compensation.

[0011] S2: Combining historical data, the model, which includes environmental interference, seasonal factors, and differences in the cabinet area, determines the risk of condensation and the safety status, and generates power supply switching instructions and dehumidification demand signals.

[0012] S3: Switch power supply mode and stabilize voltage according to instructions; Dehumidification execution module starts and stops operation according to demand signals based on the aging status of dehumidification element, and feeds back the processing results to control module.

[0013] S4: Combine ventilation conditions to determine the dehumidification effect, and transmit environmental data and device status to the remote monitoring module; the monitoring module uploads information to the remote platform and alarms when abnormalities occur;

[0014] S5: After receiving information, the remote platform can issue feedback instructions; the control module retrieves historical data, adjusts the control strategy through an algorithm that includes cabinet structure coefficients, updates parameters, and pushes them to each module as the basis for the next round of control.

[0015] Preferably, in step S1, the state sampling period is changed according to external weather conditions, and the mechanism is as follows: when the external rainfall intensity... Or wind speed At that time, forced Set to 5 minutes and continue for 1 hour; otherwise, calculate according to the information entropy formula:

[0016] ;

[0017] In the formula, E is the information entropy of the humidity sequence over the past 1 hour. .

[0018] Preferably, in step S1, the filtering process with sensor drift compensation employs an improved Kalman filter, the relevant expression of which is as follows:

[0019] ;

[0020] In the formula, Let represent the noise covariance matrix at time t; Represents the fundamental noise covariance matrix; This represents the adaptive coefficient, with a value range of 0.1-0.3; Represents a diagonal matrix function; This represents the humidity drift coefficient, with a value of 0.02%RH / day. Indicates the number of days since the last sensor calibration; This indicates the humidity prediction error; This represents the temperature drift coefficient, with a value of 0.01℃ / day. This indicates the error in temperature prediction.

[0021] Preferably, in step S2, the model containing environmental disturbances, seasonal factors, and differences in the area within the cabinet is an LSTM model, and the relevant expression is as follows:

[0022] ;

[0023] In the formula, This represents the probability of condensation, with a value ranging from 0 to 1; This represents the Sigmoid activation function; This represents the output layer weight matrix; Represents a long short-term memory network function; This represents the temperature and humidity sequence over the past 24 hours, sampled at 5-minute intervals, with a total of 288 sampling points, where t is the current time and t-288 is the time 24 hours ago; Represents the real-time interference factor at time t; Indicates the output layer bias term; This indicates the cabinet door status, taking discrete values ​​from {0, 1}, where 1 indicates the cabinet door is open and 0 indicates the cabinet door is closed. Indicates solar radiation intensity; when hour, The weighting coefficient is automatically increased by 30%.

[0024] Preferably, in step S2, the historical data is processed using a federated learning model, incorporating a seasonal factor, as shown in the following expression:

[0025] ;

[0026] In the formula, This represents the training loss value of the i-th local node; Indicates the local node number; Indicates the time step; This represents the total number of time steps in the historical data. This represents the historical condensation label of the i-th local node at time t, with a value of {0,1}, where 1 indicates that condensation occurred at time t and 0 indicates that condensation did not occur. This represents the predicted condensation value of the i-th local node at time t; This represents the seasonal correction factor, with a value of 1.2 for summer, 0.8 for winter, and 1.0 for spring and autumn. Represents the regularization coefficient; This represents the model parameters of the i-th local node; Indicates global aggregation parameters; Let L2 norm be the square of the model parameters of the i-th local node and the global aggregation parameters.

[0027] Preferably, in step S2, the determination of condensation risk and safety status uses an attention model, which takes into account the differences in the area inside the cabinet as input. The relevant expression is as follows:

[0028]

[0029] In the formula, The self-attention weights for the i-th type of signal range from 0 to 1. Represents an exponential function; This represents the similarity between the i-th type of signal and the global average feature; This represents the eigenvector of the i-th type of signal; Represents the global average eigenvector; This indicates the area weight, with no unit. The value is 1.5 for the busbar compartment, 1.2 for the cable compartment, and 1.0 for the operating mechanism compartment. Indicates the signal category number; Indicates the total number of signal categories; This represents the feature vector of the j-th type of signal, which contains the feature information of the j-th type of signal; It represents the sum of the exponential transform values ​​corresponding to all signal categories.

[0030] Preferably, in step S3, the start and stop of the dehumidification execution module is controlled by the following expression:

[0031] ;

[0032] In the formula, Indicates the reward value; This indicates the weighting coefficient for the humidity compliance reward. This represents a positive reward item. When the current humidity is lower than the target humidity, the value is 0. When the current humidity is higher than the target humidity, the value is the difference between the target humidity and the current humidity. Indicates the target humidity; Indicates the current humidity; This represents the energy consumption penalty weighting coefficient; Indicates the current dehumidification power; This represents the aging coefficient of the dehumidification element, with a value ranging from 0 to 1. The calculation formula is as follows: ;when Timely maintenance reminders are triggered; This indicates the cumulative operating hours of the dehumidification module; This represents the power fluctuation penalty weighting coefficient; Indicates the absolute value of power fluctuation; This indicates the dehumidification power at the previous moment.

[0033] Preferably, in step S4, the method of determining the dehumidification effect by combining ventilation condition compensation is based on a transfer learning model, and the relevant expression is as follows:

[0034] ;

[0035] In the formula, This represents the humidity prediction value after correction at time t, which is the humidity prediction result inside the cabinet after adding ventilation compensation; Indicates the current time; This represents the humidity prediction value of the pre-trained model at time t; This represents the difference compensation function at time t, with inputs being the difference feature D between the current environment and the standard environment, and time t. This indicates the differences between the current environment and the standard environment; This represents the ventilation velocity inside the cabinet at time t; Indicates the ventilation impact factor; This represents the migration coefficient.

[0036] Preferably, in step S5, the algorithm adjustment control strategy including the cabinet structure coefficient has the following relevant expression:

[0037] ;

[0038] In the formula, This represents the first objective function, namely total energy consumption; This represents a definite integral operation, with the integration interval from 0 to time t; This indicates the time when integration ends, i.e., the time cutoff point for calculating the total energy consumption; The operating power at time t represents the total operating power of the device at time t. The cabinet structure coefficient is represented by the following formula: ; This indicates the ventilation opening area, which is the total area of ​​the openings used for ventilation on the ring main unit. This indicates the cabinet volume, that is, the total internal volume of the ring main unit; This represents the second objective function, namely, the time required to achieve the humidity target. Describes the minimum value function; The humidity inside the cabinet at time t is the actual humidity value collected by the status detection module at time t. This indicates the target humidity, which is the ideal humidity value that needs to be maintained inside the ring main unit; The humidity requirement is that the actual humidity at time t must be less than or equal to the product of the target humidity and the cabinet structure coefficient.

[0039] The beneficial effects of this invention are as follows:

[0040] 1. This invention adopts a hybrid power supply mode of solar energy + battery storage, combined with power supply switching and voltage stabilization design, which is independent of the power grid and green and energy-saving. The power supply strategy is dynamically adjusted to adapt to different operating conditions, ensuring the continuous and stable operation of core components such as dehumidifiers and alarms, and avoiding condensation protection failure caused by power outages.

[0041] 2. This invention adapts to weather changes through dynamic sampling periods and combines improved Kalman filtering to compensate for sensor drift, ensuring the accuracy of temperature and humidity data. By integrating LSTM, attention models, and federated learning, and incorporating features such as environmental interference, seasonal factors, and differences in areas within the cabinet, the condensation risk assessment more closely reflects actual working conditions, reducing false or missed dehumidification.

[0042] 3. The dehumidification execution module of this invention combines the start-stop function with the aging status of components, and balances humidity target achievement, energy consumption control, and power stability through a reward mechanism to avoid ineffective energy consumption. Ventilation condition compensation and cabinet structure coefficients are introduced to specifically adjust the dehumidification strategy, adapting to different ventilation and cabinet specifications, and shortening the time to achieve humidity targets.

[0043] 4. This invention integrates a smoke detector with multi-dimensional anomaly monitoring, and is equipped with a remote platform to upload environmental data and device status in real time, providing rapid alarm in case of anomalies. It provides closed-loop monitoring of condensation risks and equipment safety throughout the entire process, proactively preventing faults such as insulation degradation and equipment corrosion caused by condensation, thus ensuring the safe operation of the 10kV ring main unit.

[0044] 5. This invention supports remote control and remote policy distribution, eliminating the need for on-site monitoring and improving management efficiency. The control policy learns and iterates based on historical data, adapting to changes in environment and equipment status. Simultaneously, it triggers aging maintenance reminders for dehumidification components, extending the device's lifespan and reducing the frequency and cost of manual maintenance.

[0045] 6. This invention, through seasonal correction, ventilation compensation, and cabinet structure coefficient adjustment, adapts to ring main units with different climates, ventilation conditions, and cabinet specifications. The federated learning mode eliminates the need for centralized transmission of raw data, balancing data privacy protection with global model optimization, making it suitable for multi-site batch applications. Attached Figure Description

[0046] Figure 1 Schematic diagram of an adaptive condensation elimination device for a 10kV ring main unit;

[0047] Figure 2 A schematic diagram illustrating the steps of the humidity control method for the adaptive anti-condensation device of a 10kV ring main unit. Detailed Implementation

[0048] An adaptive condensation elimination device for a 10kV ring main unit includes a control box and a solar charging controller. Both the controller and the solar charging controller are installed inside the ring main unit. The device includes a solar panel, a battery, a solar generator, a dehumidifier, and a smoke detector.

[0049] The solar panels are installed outside the ring main unit, while the batteries, solar generator, dehumidifier, and smoke detector are installed inside the ring main unit.

[0050] The solar panel is electrically connected to the solar generator and signal-connected to the solar charge controller. The solar generator is electrically connected to the storage battery, which can power the ring main unit's environmental monitoring and control box, dehumidifier, and smoke detector.

[0051] The dehumidifier and smoke detector are both connected to the control box via signal, and the control box can be controlled by a remote platform.

[0052] A humidity control method for a 10kV ring main unit adaptive anti-condensation device, based on the aforementioned 10kV ring main unit adaptive anti-condensation device, includes the following steps:

[0053] S1: The status detection module collects temperature and humidity data and safety status signals inside the cabinet according to the dynamic sampling period. After filtering with sensor drift compensation, the standardized data is transmitted to the control module.

[0054] S2: The control module combines historical data and uses a model that includes environmental interference, seasonal factors and differences in the cabinet area to determine the risk of condensation and the safety status, and generates power supply switching commands and dehumidification demand signals.

[0055] S3: The power supply module switches the power supply mode and stabilizes the voltage according to the command; the dehumidification execution module starts and stops the operation according to the demand signal based on the aging status of the dehumidification element, and feeds back the processing results to the control module.

[0056] S4: The control module combines ventilation conditions to determine the dehumidification effect and transmits environmental data and device status to the remote monitoring module; the monitoring module uploads information to the remote platform and alarms when abnormalities occur;

[0057] S5: After receiving information, the remote platform can issue feedback instructions; the control module retrieves historical data, adjusts the control strategy through an algorithm that includes cabinet structure coefficients, updates parameters, and pushes them to each module as the basis for the next round of control.

[0058] In step S1, the sampling period is changed according to external weather conditions. The mechanism is as follows: when the external rainfall intensity changes... Or wind speed At that time, forced Set to 5 minutes and continue for 1 hour; otherwise, calculate according to the information entropy formula:

[0059] ;

[0060] In the formula, E is the information entropy of the humidity sequence over the past 1 hour. .

[0061] In step S1, the filtering process with sensor drift compensation uses an improved Kalman filter, and the relevant expression is as follows:

[0062] ;

[0063] In the formula, Let represent the noise covariance matrix at time t; Represents the fundamental noise covariance matrix; This represents the adaptive coefficient, with a value range of 0.1-0.3; Represents a diagonal matrix function; This represents the humidity drift coefficient, with a value of 0.02%RH / day. Indicates the number of days since the last sensor calibration; This indicates the humidity prediction error; This represents the temperature drift coefficient, with a value of 0.01℃ / day. This indicates the error in temperature prediction.

[0064] In step S2, the model incorporating environmental disturbances, seasonal factors, and differences in the cabinet's internal areas uses an LSTM model, with the relevant expressions as follows:

[0065] ;

[0066] In the formula, This represents the probability of condensation, with a value ranging from 0 to 1; This represents the Sigmoid activation function; This represents the output layer weight matrix; Represents a long short-term memory network function; This represents the temperature and humidity sequence over the past 24 hours, sampled at 5-minute intervals, with a total of 288 sampling points, where t is the current time and t-288 is the time 24 hours ago; Represents the real-time interference factor at time t; Indicates the output layer bias term; This indicates the cabinet door status, taking discrete values ​​from {0, 1}, where 1 indicates the cabinet door is open and 0 indicates the cabinet door is closed. Indicates solar radiation intensity; when hour, The weighting coefficient is automatically increased by 30%.

[0067] In step S2, historical data is processed using a federated learning model, incorporating a seasonal factor. The relevant expression is as follows:

[0068] ;

[0069] In the formula, This represents the training loss value of the i-th local node; Indicates the local node number; Indicates the time step; This represents the total number of time steps in the historical data. This represents the historical condensation label of the i-th local node at time t, with a value of {0,1}, where 1 indicates that condensation occurred at time t and 0 indicates that condensation did not occur. This represents the predicted condensation value of the i-th local node at time t; This represents the seasonal correction factor, with a value of 1.2 for summer, 0.8 for winter, and 1.0 for spring and autumn. Represents the regularization coefficient; This represents the model parameters of the i-th local node; Indicates global aggregation parameters; Let L2 norm be the square of the model parameters of the i-th local node and the global aggregation parameters.

[0070] In step S2, the risk and safety status of condensation are determined using an attention model. This model's input includes the differential features of the areas inside the cabinet. The relevant expressions are as follows:

[0071]

[0072] In the formula, The self-attention weights for the i-th type of signal range from 0 to 1. Represents an exponential function; This represents the similarity between the i-th type of signal and the global average feature; This represents the eigenvector of the i-th type of signal; Represents the global average eigenvector; This indicates the area weight, with no unit. The value is 1.5 for the busbar compartment, 1.2 for the cable compartment, and 1.0 for the operating mechanism compartment. Indicates the signal category number; Indicates the total number of signal categories; This represents the feature vector of the j-th type of signal, which contains the feature information of the j-th type of signal; It represents the sum of the exponential transform values ​​corresponding to all signal categories.

[0073] In step S3, the dehumidification execution module is started and stopped. The relevant expressions are as follows:

[0074] ;

[0075] In the formula, Indicates the reward value; This indicates the weighting coefficient for the humidity compliance reward. This represents a positive reward item. When the current humidity is lower than the target humidity, the value is 0. When the current humidity is higher than the target humidity, the value is the difference between the target humidity and the current humidity. Indicates the target humidity; Indicates the current humidity; This represents the energy consumption penalty weighting coefficient; Indicates the current dehumidification power; This represents the aging coefficient of the dehumidification element, with a value ranging from 0 to 1. The calculation formula is as follows: ;when Timely maintenance reminders are triggered; This indicates the cumulative operating hours of the dehumidification module; This represents the power fluctuation penalty weighting coefficient; Indicates the absolute value of power fluctuation; This indicates the dehumidification power at the previous moment.

[0076] In step S4, the dehumidification effect is judged by combining ventilation condition compensation. A transfer learning model is selected, and the relevant expression is as follows:

[0077] ;

[0078] In the formula, This represents the humidity prediction value after correction at time t, which is the humidity prediction result inside the cabinet after adding ventilation compensation; Indicates the current time; This represents the humidity prediction value of the pre-trained model at time t; This represents the difference compensation function at time t, with inputs being the difference feature D between the current environment and the standard environment, and time t. This indicates the differences between the current environment and the standard environment; This represents the ventilation velocity inside the cabinet at time t; Indicates the ventilation impact factor; This represents the migration coefficient.

[0079] In step S5, the algorithm adjusts the control strategy, which includes the cabinet structure coefficient. The relevant expression is as follows:

[0080] ;

[0081] In the formula, This represents the first objective function, namely total energy consumption; This represents a definite integral operation, with the integration interval from 0 to time t; This indicates the time when integration ends, i.e., the time cutoff point for calculating the total energy consumption; The operating power at time t represents the total operating power of the device at time t. The cabinet structure coefficient is represented by the following formula: ; This indicates the ventilation opening area, which is the total area of ​​the openings used for ventilation on the ring main unit. This indicates the cabinet volume, that is, the total internal volume of the ring main unit; This represents the second objective function, namely, the time required to achieve the humidity target. Describes the minimum value function; The humidity inside the cabinet at time t is the actual humidity value collected by the status detection module at time t. This indicates the target humidity, which is the ideal humidity value that needs to be maintained inside the ring main unit; The humidity requirement is that the actual humidity at time t must be less than or equal to the product of the target humidity and the cabinet structure coefficient.

[0082] The embodiments of the present invention are as follows:

[0083] This embodiment describes the application of an adaptive anti-condensation device for humidity control in an outdoor 10kV ring main unit (model: XGN15-12). The total area of ​​the ventilation openings of this ring main unit is 0.2m². 2 The cabinet has a volume of 1.5m³. 3The cabinet structure coefficient is calculated to be approximately 0.827. The application scenario is mid-to-late June in summer, with an outdoor temperature of 32℃, an initial cabinet temperature of 30℃, humidity of 78%RH, no rainfall, wind speed of 3m / s, a sensor 30 days since the last calibration, and a cumulative operating time of 1200 hours for the dehumidification element. The aging coefficient is 0.8, which does not reach the maintenance trigger value of 0.7, so no maintenance is required. The target humidity of the ring network cabinet is set to 60%RH. The basic parameters of the device are set as follows: basic sampling period of 30min, λ=2.0, basic noise covariance matrix [[0.01,0],[0,0.01]], adaptive coefficient α=0.2, humidity compliance reward weight coefficient of 5.0, energy consumption penalty weight coefficient of 0.3, power fluctuation penalty weight coefficient of 0.2, ventilation influence coefficient of 0.15, and migration coefficient of 0.9.

[0084] First, the status detection module performs data acquisition and processing. Since there is no rainfall and the wind speed is less than 6 m / s, the dynamic sampling period is calculated using the information entropy method. Humidity data collected at 10-minute intervals over the past hour are 78%RH, 77%RH, 79%RH, 78%RH, 80%RH, and 79%RH. The calculated information entropy value is approximately 1.79. Combined with the basic sampling period, the final sampling period is determined to be 1 minute. Subsequently, the acquired temperature and humidity data are processed using an improved Kalman filter with sensor drift compensation. Combining parameters such as temperature and humidity drift coefficients, prediction errors, and calibration intervals, the humidity-related drift term is calculated to be 1.1%RH, and the temperature-related drift term is calculated to be 0.6℃. The noise covariance matrix at time t is determined to be [[0.252,0],[0,0.082]]. After filtering, the standardized temperature and humidity data are obtained as 29.8℃ and 77.6%RH. At the same time, the acquired safety status signal shows that the cabinet door is closed and there is no fault alarm. The relevant standardized data is then transmitted to the control module. The control module first processes historical data through federated learning, selecting three identical ring main units as local nodes. It uses historical data covering 30 days, combined with a summer seasonal correction coefficient of 1.2 and a regularization coefficient of 0.01 for training. Taking one node as an example, at a certain time step, the historical condensation label was "no condensation," with a predicted value of 0.05. The local parameters and the globally aggregated parameters showed only minor differences, resulting in a training loss of approximately 0.003005. The model converged well. Subsequently, the control module processed temperature and humidity sequences at 5-minute intervals over nearly 24 hours, cabinet door closure, and solar radiation intensity of 800W / ㎡. The real-time interference factor is input into the LSTM model, and the condensation probability is calculated to be 0.82 by combining the output layer parameters with the Sigmoid activation function, which is judged as a high condensation risk. At the same time, the signal analysis is carried out by adding the attention model of the regional difference characteristics in the cabinet. The three types of signals, namely the bus room, cable room and operating mechanism room, are assigned regional weights of 1.5, 1.2 and 1.0 respectively. The calculation shows that the attention weight of the bus room signal is the highest, which is 0.45. Then, the humidity monitoring is focused on this key area. Combined with the judgment that there is no abnormality in the safety status, the control module generates a power supply switching command and a dehumidification demand signal.

[0085] The power supply module switches to stable AC mains power supply mode as instructed, and outputs regulated 220V AC power with fluctuation error controlled within ±2%. The dehumidification execution module, based on the aging status of the dehumidification element (0.8), performs start-stop judgment according to the reinforcement learning reward mechanism. Combining the current cabinet humidity of 77.6%RH, the target humidity of 60%RH, the current dehumidification power of 300W, the previous power of 0W, and various weight coefficients, it calculates a negative reward of -132. Then, it starts the dehumidification module and runs at 300W power, and feeds back the dehumidification start and current power of 300W processing results to the control module. The control module then uses a transfer learning model combined with ventilation condition compensation to determine the dehumidification effect. The humidity prediction value of 75%RH from the pre-trained model is corrected by taking into account the slight temperature and humidity deviation between the current environment and the standard environment, the ventilation velocity in the cabinet of 0.5m / s, and the correlation coefficient. The corrected humidity prediction value is approximately 75.0054%RH. Combined with the actual humidity of 70%RH collected on site after 1 hour of dehumidification, the dehumidification effect is determined to be effective. The dehumidification module is instructed to continue operating. At the same time, the environmental data of 28.5℃ and 70%RH at that moment and the normal operation status information of the device are transmitted to the remote monitoring module. The remote monitoring module then uploads the relevant information to the remote platform. Since there are no abnormalities, no alarm is triggered.

[0086] After receiving the information, the remote platform sends a feedback command to optimize energy consumption. The control module then retrieves historical data and adjusts the control strategy using an algorithm that includes the cabinet structure coefficient. First, it calculates the relevant objective function and finds that the total energy consumption for 1 hour of dehumidification is 248.1Wh. At the same time, it determines that the humidity standard is ≤49.6%RH inside the cabinet. It is predicted that the humidity inside the cabinet can drop to 49%RH after 4 hours of continuous dehumidification, that is, the humidity standard is achieved in 4 hours. Subsequently, the control module adjusts the dehumidification power to 280W through a multi-objective optimization algorithm. At the same time, because the humidity inside the cabinet decreases, the humidity sequence entropy value decreases, so the dynamic sampling period is adjusted to 2 minutes. The relevant adjusted parameters are then pushed to each module as the basis for the next round of control. After implementing the humidity control method of the adaptive condensation elimination device, the humidity inside the 10kV ring main unit was successfully reduced to 49%RH after 4 hours, meeting the humidity standard. The total energy consumption for 1 hour of dehumidification operation was 248.1Wh, which is 15% lower than the energy consumption of the traditional fixed parameter control method. No condensation was generated inside the ring main unit throughout the process, the device operated stably, and the humidity monitoring of key areas such as the busbar room was accurate, realizing the adaptive and efficient elimination control of condensation in the ring main unit.

Claims

1. A 10kV ring main unit adaptive condensation elimination device, comprising a control box and a solar charging controller, wherein the controller and the solar charging controller are installed inside the ring main unit, characterized in that, This includes solar panels, batteries, solar generators, dehumidifiers, and smoke detectors; The solar panels are installed outside the ring main unit, while the battery, solar generator, dehumidifier, and smoke detector are installed inside the ring main unit. The solar panel is electrically connected to the solar generator and signal-connected to the solar charge controller. The solar generator is electrically connected to the storage battery, which can power the ring main unit environmental monitoring and control box, dehumidifier, and smoke detector. The dehumidifier and smoke detector are both connected to the control box via signal, and the control box can be controlled by a remote platform.

2. A humidity control method for an adaptive anti-condensation device in a 10kV ring main unit, characterized in that, Includes the following steps: S1: Collect temperature and humidity data and safety status signals inside the cabinet according to the dynamic sampling period, and transmit the standardized data to the control module after filtering with sensor drift compensation. S2: Combining historical data, the model, which includes environmental interference, seasonal factors, and differences in the cabinet area, determines the risk of condensation and the safety status, and generates power supply switching instructions and dehumidification demand signals. S3: Switch power supply mode and stabilize voltage according to instructions; Dehumidification execution module starts and stops operation according to demand signals based on the aging status of dehumidification element, and feeds back the processing results to control module. S4: Combine ventilation conditions to determine the dehumidification effect, and transmit environmental data and device status to the remote monitoring module; the monitoring module uploads information to the remote platform and alarms when abnormalities occur; S5: After receiving information, the remote platform can issue feedback instructions; the control module retrieves historical data, adjusts the control strategy through an algorithm that includes cabinet structure coefficients, updates parameters, and pushes them to each module as the basis for the next round of control.

3. The humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S1, the state sampling period is changed according to external weather conditions. The mechanism is as follows: when the external rainfall intensity... Or wind speed At that time, forced Set to 5 minutes and continue for 1 hour; otherwise, calculate according to the information entropy formula: ； In the formula, E is the information entropy of the humidity sequence over the past 1 hour. .

4. The humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S1, the filtering process with sensor drift compensation uses an improved Kalman filter, and the relevant expression is as follows: ； In the formula, Let represent the noise covariance matrix at time t; Represents the fundamental noise covariance matrix; This represents the adaptive coefficient, with a value range of 0.1-0.3; Represents a diagonal matrix function; This represents the humidity drift coefficient, with a value of 0.02%RH / day. Indicates the number of days since the last sensor calibration; This indicates the humidity prediction error; This represents the temperature drift coefficient, with a value of 0.01℃ / day. This indicates the error in temperature prediction.

5. A humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S2, the model incorporating environmental disturbances, seasonal factors, and regional differences within the cabinet adopts an LSTM model, and the relevant expressions are as follows: ； In the formula, This represents the probability of condensation, with a value ranging from 0 to 1. This represents the Sigmoid activation function; This represents the output layer weight matrix; Represents a long short-term memory network function; This represents the temperature and humidity sequence over the past 24 hours, sampled at 5-minute intervals, with a total of 288 sampling points, where t is the current time and t-288 is the time 24 hours ago; This represents the real-time interference factor at time t; Indicates the output layer bias term; This indicates the cabinet door status, taking discrete values ​​from {0, 1}, where 1 indicates the cabinet door is open and 0 indicates the cabinet door is closed. Indicates solar radiation intensity; when hour, The weighting coefficient is automatically increased by 30%.

6. The humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S2, the historical data is processed using a federated learning model, incorporating a seasonal factor, as shown in the following expression: ； In the formula, This represents the training loss value of the i-th local node; Indicates the local node number; Indicates the time step; This represents the total number of time steps in the historical data. This represents the historical condensation label of the i-th local node at time t, with a value of {0,1}, where 1 indicates that condensation occurred at time t and 0 indicates that condensation did not occur. This represents the predicted condensation value of the i-th local node at time t; This represents the seasonal correction factor, with a value of 1.2 for summer, 0.8 for winter, and 1.0 for spring and autumn. Represents the regularization coefficient; This represents the model parameters of the i-th local node; Indicates global aggregation parameters; Let L2 norm be the square of the model parameters of the i-th local node and the global aggregation parameters.

7. A humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S2, the determination of condensation risk and safety status uses an attention model, which takes into account the differences in the area inside the cabinet as input. The relevant expression is as follows: In the formula, The self-attention weights for the i-th type of signal range from 0 to 1. Represents an exponential function; This represents the similarity between the i-th type of signal and the global average feature; This represents the eigenvector of the i-th type of signal; Represents the global average eigenvector; This indicates the area weight, with no unit. The value is 1.5 for the busbar compartment, 1.2 for the cable compartment, and 1.0 for the operating mechanism compartment. Indicates the signal category number; Indicates the total number of signal categories; This represents the feature vector of the j-th type of signal, which contains the feature information of the j-th type of signal; It represents the sum of the exponential transform values ​​corresponding to all signal categories.

8. A humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S3, the dehumidification control module is started and stopped. The relevant expression is as follows: ； In the formula, Indicates the reward value; This indicates the weighting coefficient for the humidity compliance reward. This represents a positive reward item. When the current humidity is lower than the target humidity, the value is 0. When the current humidity is higher than the target humidity, the value is the difference between the target humidity and the current humidity. Indicates the target humidity; Indicates the current humidity; This represents the energy consumption penalty weighting coefficient; Indicates the current dehumidification power; This represents the aging coefficient of the dehumidification element, with a value ranging from 0 to 1. The calculation formula is as follows: ;when Timely maintenance reminders are triggered; This indicates the cumulative operating hours of the dehumidification module; This represents the power fluctuation penalty weighting coefficient; Indicates the absolute value of power fluctuation; This indicates the dehumidification power at the previous moment.

9. A humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S4, the dehumidification effect is determined by combining ventilation condition compensation. A transfer learning model is used, and the relevant expression is as follows: ； In the formula, This represents the humidity prediction value after correction at time t, which is the humidity prediction result inside the cabinet after adding ventilation compensation; Indicates the current time; This represents the humidity prediction value of the pre-trained model at time t; This represents the difference compensation function at time t, with inputs being the difference feature D between the current environment and the standard environment, and time t. This indicates the differences between the current environment and the standard environment; This represents the ventilation velocity inside the cabinet at time t; Indicates the ventilation impact factor; This represents the migration coefficient.

10. A humidity control method for an adaptive anti-condensation device in a 10kV ring main unit according to claim 2, characterized in that, In step S5, the algorithm adjustment control strategy including the cabinet structure coefficient is expressed as follows: ； In the formula, This represents the first objective function, namely total energy consumption; This represents a definite integral operation, with the integration interval from 0 to time t; This indicates the time when integration ends, i.e., the time cutoff point for calculating the total energy consumption; The operating power at time t represents the total operating power of the device at time t. The cabinet structure coefficient is represented by the following formula: ; This indicates the ventilation opening area, which is the total area of ​​the openings used for ventilation on the ring main unit. This indicates the cabinet volume, that is, the total internal volume of the ring main unit; This represents the second objective function, namely, the time required to achieve the humidity target. Describes the minimum value function; The humidity inside the cabinet at time t is the actual humidity value collected by the status detection module at time t. This indicates the target humidity, which is the ideal humidity value that needs to be maintained inside the ring main unit; The humidity requirement is that the actual humidity at time t must be less than or equal to the product of the target humidity and the cabinet structure coefficient.

Images