An on-line monitoring and intelligent air supplement closed-loop control system for gas leakage of a ring main unit

Abstract

The application discloses a kind of ring network cabinet gas leakage on-line monitoring and intelligent air supplement closed-loop control system, it is related to electric power equipment state monitoring and intelligent operation and maintenance technical field, including following module: data acquisition pre-processing module, acquisition generates standardized input dataset;Leakage characteristic generation module, calculate theoretical pressure benchmark value, output leakage rate characteristic vector;Air supplement decision module, combine differential pressure value and residual gas input improved iTransformer model, generate variable frequency pulsating air supplement instruction using global state vector broadcast injection mechanism;Air supplement execution monitoring module, response instruction executes segmented air supplement and collects flow;Abnormal diagnosis protection module, calculate predicted pressure recovery value and measured value comparison output protection instruction;Air supplement verification management module, verify pressure retention rate after pressure meets the standard, output air supplement completion signal.The application realizes leakage accurate perception, air supplement adaptive optimization and closed-loop safety verification.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and intelligent operation and maintenance technology, and in particular to an online monitoring and intelligent gas replenishment closed-loop control system for gas leakage in ring main units. Background Technology

[0002] With the rapid development and increasing intelligence of power systems, ring main units (RMMs), as crucial power distribution equipment in urban power grids, are of paramount importance for their safe and stable operation. Currently, RMMs widely employ insulating gases such as sulfur hexafluoride (SF6) for arc extinguishing and insulation. The gas pressure directly affects the equipment's insulation performance and operational safety. To ensure reliable operation, existing technologies typically install pressure and temperature sensors within the RMM to monitor the gas state in the chambers in real time. Traditional monitoring methods rely primarily on simple threshold comparisons; when the monitored real-time pressure falls below a preset alarm threshold, the system issues a leak alarm, prompting maintenance personnel to manually replenish gas or perform repairs. While this method achieves a certain degree of automated monitoring, it is limited by environmental factors, especially during drastic temperature changes. Gas pressure fluctuates significantly with thermal expansion and contraction, easily leading to false alarms or missed alarms, and it cannot accurately distinguish between physical leaks and normal thermodynamic pressure changes.

[0003] Furthermore, existing gas replenishment control strategies mostly employ simple open-loop control or single PID control. Once insufficient pressure is detected, the valve is opened to replenish gas until the pressure returns to the set value. This crude gas replenishment method often has low control precision, making it difficult to cope with gas replenishment needs under different leakage rates. It is prone to gas replenishment overshoot leading to excessively high pressure, or gas replenishment response lag failing to suppress leakage in time. More seriously, existing monitoring methods do not fully consider the impact of the mechanical deformation of the ring main unit cabinet under pressure changes on the gas chamber volume, and lack accurate monitoring of the flow rate during the gas replenishment process and an automatic verification mechanism for airtightness after gas replenishment. As a result, the system cannot take timely protective measures when gas replenishment is abnormal, nor can it confirm the pressure maintenance effect after gas replenishment. The overall level of intelligence and reliability needs to be improved.

[0004] Therefore, how to provide an online monitoring and intelligent gas replenishment closed-loop control system for gas leakage in ring main units is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an online monitoring and intelligent gas replenishment closed-loop control system for gas leaks in ring main units. This invention fully integrates key steps such as multi-source data acquisition and preprocessing, leak feature generation, improved iTransformer model decision-making, variable frequency pulsed gas replenishment execution, and gas replenishment verification management. It constructs an intelligent gas replenishment process with data standardization, thermodynamic state decoupling analysis, heterogeneous feature embedding and alignment, and segmented gas replenishment timing planning, achieving accurate perception of leak status and adaptive optimization of gas replenishment volume under complex operating conditions. By introducing an improved iTransformer model and utilizing a heterogeneous feature embedding and alignment module with a global state vector broadcast injection mechanism, this invention effectively solves the problem of insufficient static resource constraint capture in traditional timing models, achieving deep coordination of leak rate, pressure difference, and remaining gas volume. This invention possesses advantages such as strong anti-interference capability for leak identification, high accuracy of gas replenishment strategy, fast pressure balance response, and verifiable gas replenishment results. It can significantly improve the stability of gas pressure in ring main units and the safety of the gas replenishment process, thereby effectively solving the problems of false alarms caused by changes in ambient temperature and overshoot or under-replenishment caused by coarse gas replenishment control in existing technologies.

[0006] According to an embodiment of the present invention, a closed-loop control system for online monitoring and intelligent gas replenishment of a ring main unit for gas leakage includes the following modules: The data acquisition and preprocessing module is used to simultaneously acquire multi-source data from the ring main unit and preprocess it to obtain a standardized input dataset; The leakage feature generation module is used to calculate the theoretical gas pressure benchmark value based on the thermodynamic equation of state and the standardized input dataset, and to compare and decouple the calculation with the real-time gas pressure in the standardized input dataset and output the leakage rate feature vector. The gas replenishment decision module is used to read the leakage rate feature vector, calculate the pressure difference between the current real-time gas pressure and the preset target gas replenishment pressure, and combine it with the current remaining gas volume of the gas replenishment storage tank in the standardized input dataset and input it into the improved iTransformer model to output a variable frequency pulsed gas replenishment command. The improved iTransformer model includes a heterogeneous feature embedding and alignment module, a temporal projection module, an inverted variable attention module, a feedforward network module and an output generation module. The heterogeneous feature embedding and alignment module constructs a global state vector broadcast injection mechanism. The gas replenishment execution monitoring module is used to open the solenoid valve in response to the frequency conversion pulsed gas replenishment command, to achieve pre-filling balance in the initial stage of gas replenishment, to achieve micro-precise gas replenishment in the subsequent stage of gas replenishment, and to collect the instantaneous gas flow rate value in the gas replenishment pipeline. The abnormal diagnosis and protection module is used to calculate the cumulative gas supply flow rate from the instantaneous gas flow rate, then convert it into the predicted pressure rise value by combining the gas chamber volume parameter, and calculate the difference between the predicted pressure rise value and the actual monitored pressure rise value to determine whether the gas supply is abnormal and output the fault protection interrupt command. The gas replenishment verification management module is used to control the solenoid valve to close when the real-time gas pressure reaches the target gas replenishment pressure range, and continuously read the real-time gas pressure data to calculate the pressure holding rate. If it meets the stability standard, it outputs a gas replenishment completion status signal and generates a gas replenishment log.

[0007] Optionally, the data acquisition and preprocessing module's operation includes: Real-time gas pressure, ambient temperature, strain values ​​on the cabinet surface, and the current remaining gas volume in the replenishment gas storage tank are collected as multi-source data. Outlier removal is performed on the multi-source data. The mean and standard deviation of each data point within the sliding time window are calculated. Data points whose absolute value differs from the mean by more than 3 times the standard deviation are marked as invalid and removed. Data cleaning is performed on the multi-source data after outlier removal. The moving average algorithm is used to calculate the average value of the data within the moving window to filter out high-frequency noise and normalize the data separately. The combined output is a standardized input dataset.

[0008] Optionally, the operation of the leakage feature generation module includes: Read the ambient temperature and cabinet surface strain values ​​from the standardized input dataset, multiply the cabinet surface strain values ​​by the inner surface area of ​​the ring network cabinet air chamber to calculate the air chamber volume deformation, and add it to the initial volume of the ring network cabinet air chamber to obtain the current actual air chamber volume. Based on the thermodynamic equation of state, the theoretical pressure reference value of the gas at the current moment is calculated using the current actual gas chamber volume and ambient temperature. The pressure deviation residual is obtained by subtracting the real-time gas pressure in the standardized input dataset from the theoretical gas pressure benchmark and taking the absolute value. If the pressure deviation residual is greater than the preset first judgment threshold, the current time is marked as a potential leak time. Within a preset sliding time window, the first slope of the real-time gas pressure change with time and the second slope of the ambient temperature change with time are calculated respectively. The measured pressure and temperature change rate is obtained by dividing the first slope by the second slope. If the absolute value of the difference between the measured pressure and temperature change rate and the preset thermodynamic correlation threshold exceeds the preset allowable deviation range, it is determined that the pressure decrease trend does not match the temperature change. When it is determined that the pressure drop trend does not match the temperature change, the difference between the endpoint value and the starting value of the real-time gas pressure within the preset sliding time window is calculated. The difference is divided by the time length of the preset sliding time window to calculate the average leakage rate. The average leakage rate is marked and output as a leakage rate feature vector.

[0009] Optionally, the working process of the gas replenishment decision module includes: Read the real-time gas pressure and the preset target gas replenishment pressure from the standardized input dataset, perform a subtraction operation to subtract the real-time gas pressure from the preset target gas replenishment pressure, and calculate the pressure difference value. Read the leakage rate feature vector and the current remaining gas volume of the gas replenishment tank in the standardized input dataset. Concatenate the leakage rate feature vector, the pressure difference value, and the current remaining gas volume of the gas replenishment tank in the feature dimension in order to construct a heterogeneous input vector. The heterogeneous input vector is input into the heterogeneous feature embedding and alignment module of the improved iTransformer model. The current remaining gas volume of the gas replenishment tank is extracted as the global state vector using the global state vector broadcast injection mechanism. The global state vector is copied to match the number of time steps of the leakage rate feature vector. The copied global state vector is added to the corresponding position values ​​of the leakage rate feature vector and the pressure difference value in the heterogeneous input vector. The embedded and aligned time series feature vector is output. The embedded and aligned temporal feature vector is input into the temporal projection module. A convolutional kernel of a preset size slides on the time axis of the embedded and aligned temporal feature vector. The values ​​within the coverage area of ​​the convolutional kernel are multiplied with the weight values ​​within the convolutional kernel and then summed to output the projected temporal features. The projected temporal features are input into the inverted variable attention module. The row and column indices of the projected temporal features are interchanged. The dot product between each row vector after transpose is calculated to obtain the original attention score. The score is then normalized and converted into a probability distribution as attention weights. The attention weights are multiplied with the transpose matrix. The row and column interchange is performed again on the multiplication result to restore the dimension. The inverted attention features are then output. The inverted attention features are input into the feedforward network module to perform a linear transformation, and an activation function is used to perform a nonlinear transformation on each value of the linear transformation result, outputting a feedforward feature vector. The weight matrix of the fully connected layer in the feedforward feature vector input-output generation module is multiplied with the feedforward feature vector. The result is mapped to the preset air replenishment command dimension. The numerical sequence obtained after mapping is used as the valve opening percentage at each moment. The results are combined in time sequence to form the frequency-controlled pulsed air replenishment command.

[0010] Optionally, the operation process of the gas replenishment execution monitoring module includes: Read the variable frequency pulsed gas replenishment command, send an opening signal to control the solenoid valve to be energized and opened, read the voltage signal output by the flow sensor installed in the gas replenishment pipeline, and convert it into a value according to a preset ratio to obtain the instantaneous gas flow rate value. Divide the preset gas replenishment start phase time by the short interval time to obtain the high frequency on / off count. Control the solenoid valve to alternately open and close according to the high frequency on / off count, so that the gas in the gas replenishment storage tank enters the ring network cabinet gas chamber in batches. Calculate the duration of the subsequent gas replenishment stage, divide the duration of the subsequent gas replenishment stage by the long interval duration to obtain the number of low-frequency on / off cycles, and control the solenoid valve to alternately open and close according to the number of low-frequency on / off cycles, thereby adjusting the gas flow rate from the gas replenishment storage tank into the ring network cabinet gas chamber to a trace range.

[0011] Optionally, the operation of the anomaly diagnosis and protection module includes: Read the instantaneous gas flow rate in the gas supply pipeline, add up the instantaneous gas flow rate values ​​at all times to calculate the total value, and mark the total value as the cumulative gas supply flow rate; Read the air chamber volume parameters of the ring main unit, divide the cumulative replenishment flow rate by the air chamber volume parameters to obtain the unit volume replenishment volume, and multiply the unit volume replenishment volume by the preset gas state correction coefficient to obtain the predicted pressure recovery value. Read the real-time gas pressure from the standardized input dataset, calculate the difference between the real-time gas pressure value at the start of gas replenishment and the real-time gas pressure value at the current time, and mark the difference as the actual monitored pressure increase. Calculate the absolute value of the difference between the predicted pressure rise and the actual monitored pressure rise. If the absolute value of the difference is greater than the preset abnormal judgment threshold, the gas supply is judged to be abnormal, and a fault protection interruption command is sent to control the solenoid valve to close.

[0012] Optionally, the working process of the gas replenishment verification management module includes: Read the real-time gas pressure from the standardized input dataset. If the real-time gas pressure is within the target replenishment pressure range, send a shutdown signal to control the solenoid valve to shut off. Calculate the average real-time gas pressure within the preset pressure monitoring time after the valve is closed, mark the average value as the stable pressure after valve closure, calculate the absolute value of the difference between the stable pressure after valve closure and the preset target gas replenishment pressure to obtain the pressure stability deviation, divide the pressure stability deviation by the preset target gas replenishment pressure to obtain the pressure relative deviation rate, and subtract the pressure relative deviation rate from the value 1 to obtain the pressure retention rate. The pressure holding rate is compared with the preset stability threshold. If the pressure holding rate is less than the preset stability threshold, it is determined that the stability standard is not met and an air replenishment failure alarm signal is output. If the pressure holding rate is greater than or equal to the preset stability threshold, it is determined that the stability standard is met.

[0013] The beneficial effects of this invention are: This invention addresses the issues of high false alarm rates in gas pressure monitoring and coarse gas replenishment control strategies in ring main units due to ambient temperature fluctuations by deploying a multi-source data acquisition module and an improved iTransformer model. It employs a thermodynamic equation of state combined with strain data from the cabinet surface to decouple leakage characteristics and calculate leakage rates, generating a leakage rate feature vector. A heterogeneous feature embedding and alignment module with a global state vector broadcast injection mechanism is constructed to copy and inject the current remaining gas volume in the replenishment tank as a global state vector into the time-series features. This is combined with the leakage rate feature vector and differential pressure value input into the improved iTransformer model to execute variable-frequency pulsed gas replenishment command planning. During the gas replenishment execution phase, short-interval high-frequency switching and long-interval low-frequency switching are used to achieve pre-charging balance and fine-tuned micro-gas replenishment. Anomaly diagnosis and protection are performed based on the integral of instantaneous gas flow rate and the comparison of pressure recovery. After gas replenishment, a gas replenishment completion status signal is output by calculating the pressure retention rate and comparing it with a stability threshold. Ultimately, it achieves accurate anti-interference identification of gas leaks in ring main units, adaptive optimization of gas replenishment strategies, and closed-loop safety verification of the gas replenishment process, effectively improving the accuracy of leak monitoring, the precision of gas replenishment control, and the safety of system operation. Attached Figure Description

[0014] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a structural diagram of a closed-loop control system for online monitoring and intelligent gas replenishment of a ring main unit, as proposed in this invention. Figure 2 This is a flowchart of the leakage characteristic generation and leakage rate calculation based on the thermodynamic equation of state and strain correction proposed in this invention. Figure 3 This is a flowchart of the gas replenishment decision and variable frequency pulsed gas replenishment command generation based on the improved iTransformer model proposed in this invention. Detailed Implementation

[0015] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0016] refer to Figures 1-3 A closed-loop control system for online monitoring and intelligent gas replenishment of gas in a ring main unit includes the following modules: The data acquisition and preprocessing module is used to synchronously acquire multi-source data of the ring main unit, including real-time gas pressure, ambient temperature, strain values ​​of the cabinet surface, and the current remaining gas volume of the gas replenishment tank, and preprocess it to obtain a standardized input dataset. The leakage feature generation module is used to calculate the theoretical gas pressure benchmark value at the current moment based on the thermodynamic equation of state, using the ambient temperature and cabinet surface strain values ​​in the standardized input dataset, and compare and decouple the calculation with the real-time gas pressure in the standardized input dataset. If the pressure drop trend is detected to be mismatched with the temperature change, the leakage rate feature vector is output. The gas replenishment decision module is used to read the leakage rate feature vector, calculate the pressure difference between the current real-time gas pressure and the preset target gas replenishment pressure, and input the leakage rate feature vector, pressure difference, and the current remaining gas volume of the gas replenishment storage tank in the standardized input dataset into the improved iTransformer model. The output includes a variable frequency pulsed gas replenishment command containing a segmented gas replenishment time sequence. The improved iTransformer model includes a heterogeneous feature embedding and alignment module, a temporal projection module, an inverted variable attention module, a feedforward network module, and an output generation module. The heterogeneous feature embedding and alignment module constructs a global state vector broadcast injection mechanism. The gas replenishment execution monitoring module is used to open the solenoid valve in response to the variable frequency pulsed gas replenishment command. In the initial stage of gas replenishment, it controls the valve to perform short-interval high-frequency on-off switching to achieve pre-filling balance. In the subsequent stage of gas replenishment, it controls the valve to perform long-interval low-frequency on-off switching to achieve micro-precise gas replenishment and collects the instantaneous gas flow rate value in the gas replenishment pipeline. The abnormal diagnosis and protection module is used to integrate the instantaneous gas flow rate to obtain the cumulative gas supply flow rate, convert the cumulative gas supply flow rate into the predicted pressure rise value based on the gas chamber volume parameter, calculate the difference between the predicted pressure rise value and the actual monitored pressure rise value, determine whether the gas supply is abnormal, and output a fault protection interrupt command. The gas replenishment verification management module is used to control the solenoid valve to close when the real-time gas pressure reaches the target gas replenishment pressure range, and continuously read the real-time gas pressure data for a set time after the valve closes to calculate the pressure holding rate. If the pressure holding rate meets the stability standard, it outputs a gas replenishment completion status signal and generates a gas replenishment log.

[0017] This implementation significantly improves the accuracy of gas leak monitoring and the precision of gas replenishment control in ring main units. By simultaneously collecting real-time gas pressure, ambient temperature, and cabinet surface strain values, and performing correlation decoupling calculations based on the thermodynamic equation of state, it effectively eliminates interference from temperature fluctuations and mechanical deformation, accurately identifies the actual leakage trend, and outputs a leakage rate feature vector. The gas replenishment decision module utilizes an improved iTransformer model and integrates remaining gas volume constraints through a global state vector broadcast injection mechanism to generate variable frequency pulsed gas replenishment commands containing segmented gas replenishment timing. Combined with high-frequency pre-charging and low-frequency fine-tuning gas replenishment strategies, it greatly optimizes the gas replenishment response speed and pressure control accuracy. Simultaneously, the anomaly diagnosis and protection module compares the predicted pressure recovery value with the actual monitoring data in real time, dynamically determines gas replenishment anomalies, and triggers interrupt commands. The gas replenishment verification and management module automatically evaluates the pressure retention rate after the operation is completed. This closed-loop control mechanism ensures the safety and reliability of the gas replenishment process while comprehensively reducing the false alarm rate and the risk of gas replenishment overshoot, achieving intelligent perception and efficient adaptive operation and maintenance of the ring main unit's insulation status.

[0018] In this embodiment, the data acquisition and preprocessing module's operation includes: Real-time gas pressure, ambient temperature, cabinet surface strain values, and the current remaining gas volume in the replenishment gas storage tank are collected as multi-source data. Outlier removal processing is performed on the multi-source data. The mean and standard deviation of each data point within the sliding time window are calculated. Data points whose absolute value differs from the mean by more than 3 times the standard deviation are marked as invalid and removed. The cabinet surface strain value is collected by resistance strain gauges deployed on the surface of the ring network cabinet. The resistance strain gauges are connected to one arm of a Wheatstone bridge circuit. When the cabinet undergoes mechanical deformation, the Wheatstone bridge loses balance and outputs a millivolt-level voltage signal. The millivolt-level voltage signal is differentially amplified by an instrumentation amplifier and converted into an analog voltage signal. The analog voltage signal is then sampled by an analog-to-digital converter chip and converted into a digital quantity. Data cleaning is performed on the multi-source data after outlier removal. The moving average algorithm is used to calculate the average value of the data within the moving window to filter out high-frequency noise and normalize the data separately. The combined output is a standardized input dataset.

[0019] In this embodiment, the working process of the leakage feature generation module includes: Read the ambient temperature and cabinet surface strain values ​​from the standardized input dataset, multiply the cabinet surface strain values ​​by the inner surface area of ​​the ring network cabinet air chamber to calculate the air chamber volume deformation, and add it to the initial volume of the ring network cabinet air chamber to obtain the current actual air chamber volume. According to the thermodynamic equation of state, the theoretical pressure benchmark value of the gas at the current moment is calculated using the current actual gas chamber volume and ambient temperature. The thermodynamic equation of state is to convert the ambient temperature value in the standardized input dataset into a thermodynamic temperature value by adding 273.15. The thermodynamic temperature value is then multiplied by the number of gas moles and the gas molar constant based on the initial state to obtain the numerator product. The current actual gas chamber volume is used as the denominator, and the numerator product is divided by the current actual gas chamber volume to obtain the theoretical pressure benchmark value of the gas at the current moment. The pressure deviation residual is obtained by subtracting the real-time gas pressure from the theoretical gas pressure benchmark value in the standardized input dataset and taking the absolute value. If the pressure deviation residual is greater than a preset first judgment threshold, the current moment is marked as a potential leak moment. The first judgment threshold is obtained by calculating the product of the standard deviation and the confidence coefficient. The standard deviation is the root mean square of the pressure deviation residual relative to the mean residual within a preset historical time period. The confidence coefficient ranges from 2.5 to 3.5. The preset historical time period is calculated by multiplying the system sampling frequency and the sampling period duration. The system sampling frequency ranges from 1Hz to 10Hz, and the sampling period duration ranges from 1 hour to 24 hours. Within a preset sliding time window, the first slope of the real-time gas pressure change over time and the second slope of the ambient temperature change over time are calculated respectively. The measured pressure-temperature change rate is obtained by dividing the first slope by the second slope. The preset sliding time window is obtained by calculating the maximum value of the leakage feature display duration and the ambient temperature sampling period. The leakage feature display duration is calculated based on the minimum leakage orifice diameter and the gas chamber volume parameter. The calculation formula is the gas chamber volume divided by the unit orifice diameter leakage rate, and the value range is 10 minutes to 30 minutes. The ambient temperature sampling period ranges from 5 minutes to 15 minutes. The measured pressure-temperature change rate is compared with a preset thermodynamic correlation threshold. If the absolute value of the difference between the measured pressure-temperature change rate and the preset thermodynamic correlation threshold exceeds the preset allowable deviation range, it is determined that the pressure decrease trend does not match the temperature change. The thermodynamic correlation threshold is obtained by calculating the partial derivative of the theoretical pressure with temperature change. The formula for calculating the partial derivative is the product of the ideal gas constant and the preset number of gas moles divided by the current actual gas chamber volume. Its value range is set to 0.003 MPa / ℃ to 0.006 MPa / ℃ according to the gas chamber volume parameter. The preset allowable deviation range is 0.0005 MPa / ℃ to 0.001 MPa / ℃. When it is determined that the pressure drop trend does not match the temperature change, the difference between the endpoint value and the starting value of the real-time gas pressure within the preset sliding time window is calculated. The difference is divided by the time length of the preset sliding time window to calculate the average leakage rate. The average leakage rate is marked and output as a leakage rate feature vector.

[0020] In this embodiment, the working process of the gas replenishment decision module includes: Read the real-time gas pressure and the preset target gas replenishment pressure from the standardized input dataset, perform a subtraction operation to subtract the real-time gas pressure from the preset target gas replenishment pressure, and calculate the pressure difference value. The preset target gas replenishment pressure is set according to the rated operating pressure parameter of the ring network cabinet gas chamber. The calculation formula is the rated operating pressure value multiplied by the safety factor. The safety factor ranges from 0.95 to 0.98. Read the leakage rate feature vector and the current remaining gas volume of the gas replenishment tank in the standardized input dataset. Concatenate the leakage rate feature vector, the pressure difference value, and the current remaining gas volume of the gas replenishment tank in the feature dimension in order to construct a heterogeneous input vector. The heterogeneous input vector is input into the heterogeneous feature embedding and alignment module of the improved iTransformer model. The current remaining gas volume of the gas replenishment tank is extracted as the global state vector using the global state vector broadcast injection mechanism. The global state vector is copied to match the number of time steps of the leakage rate feature vector. The copied global state vector is added to the corresponding position values ​​of the leakage rate feature vector and the pressure difference value in the heterogeneous input vector. The embedded and aligned time series feature vector is output. The embedded and aligned temporal feature vector is input into the temporal projection module. A convolutional kernel of a preset size slides on the time axis of the embedded and aligned temporal feature vector. The values ​​within the coverage area of ​​the convolutional kernel are multiplied with the weight values ​​within the convolutional kernel and then summed to output the projected temporal features. The size of the preset convolutional kernel is obtained by calculating the time series length of the model input layer and the downsampling factor. The time series length of the model input layer ranges from 60 to 120, and the downsampling factor ranges from 2 to 4. The projected temporal features are input into the inverted variable attention module. The row and column indices of the projected temporal features are interchanged to transpose the variable dimension and the time dimension. The dot product between each row vector after transposition is calculated to obtain the original attention score, which is then normalized to a probability distribution as attention weights. The attention weights are multiplied with the transposed matrix. The row and column interchange operation is performed again on the multiplication result to restore the dimension, and the inverted attention features are output. The inverted attention features are input into the feedforward network module to perform a linear transformation, and an activation function is used to perform a non-linear transformation on each value of the linear transformation result, outputting a feedforward feature vector. The dimension of the weight matrix of a preset size in the linear transformation is determined according to the number of hidden layer neurons in the feedforward network. The number of hidden layer neurons ranges from 64 to 256. The activation function used is the ReLU function. The weight matrix of the fully connected layer in the feedforward feature vector input-output generation module is multiplied with the feedforward feature vector. The result is mapped to a preset air replenishment command dimension. The numerical sequence obtained after mapping is used as the valve opening percentage at each moment. The numerical sequence is combined in time order to form a variable frequency pulsed air replenishment command containing a segmented air replenishment time sequence. The value of the air replenishment command dimension is calculated by dividing the preset total air replenishment duration by the air replenishment control time interval. The preset total air replenishment duration ranges from 10 minutes to 60 minutes, and the air replenishment control time interval ranges from 1 second to 10 seconds.

[0021] This implementation calculates the pressure difference by reading the real-time gas pressure and the preset target replenishment pressure, and constructs an improved iTransformer model by combining the leakage rate feature vector and the remaining gas volume in the storage tank. Utilizing the global state vector broadcast injection mechanism in the heterogeneous feature embedding alignment module, the remaining gas volume constraint is integrated into the temporal features. Through temporal projection, inverted variable attention, and deep feature extraction by the feedforward network module, a variable-frequency pulsed replenishment command containing segmented replenishment timing is finally generated. This invention can adaptively match different leakage rates and gas reserves, and achieves dynamic optimization and precise control of the replenishment process by accurately calculating the valve opening percentage at each moment. This effectively solves the problems of response lag and coarse control in traditional replenishment methods, significantly improving replenishment efficiency and system stability.

[0022] The improved iTransformer model of this invention is similar to the original iTransformer model in that both retain the deep learning architecture for time series data. That is, by independently embedding the input variables, using the attention mechanism to capture the long-range dependencies between variables, and using a multi-layer feedforward network for feature extraction and transformation, the final output is a predicted sequence. Both also retain standard operations such as residual connections and layer normalization to maintain training stability.

[0023] The difference lies in that this invention breaks through the limitation of the original iTransformer model, which only performs feature prediction based on single-modal time-series data. It introduces a heterogeneous feature embedding and alignment module and constructs a global state vector broadcast injection mechanism. Building upon the original model's direct processing of time-series input, this invention pre-processes feature concatenation and embedding alignment steps. This step, as a special constraint fusion module, extracts the current remaining gas volume of the gas replenishment tank as a global state vector, copies it to match the time step size of the leakage rate feature vector, and then performs addition operations at the corresponding positions, dynamically injecting static resource constraints into the time dimension. Subsequently, after processing by time-series projection, inverted variable attention, and feedforward network modules, the heterogeneous input vector is finally mapped to specific gas replenishment commands through the output generation module.

[0024] Based on the above improvements, the beneficial effects of this invention are that, through the global state vector broadcast injection mechanism, the improved iTransformer model can strongly correlate the key physical constraint of the remaining gas volume of the gas replenishment tank to the full-time decision-making process of leak gas replenishment. This breaks the limitation of the original model, which only focuses on historical pressure changes and ignores resource availability constraints, and realizes the resource adaptability and physical consistency of the gas replenishment strategy. This design significantly enhances the model's dynamic control capability over the gas replenishment process, and can automatically adjust the gas replenishment intensity when the gas volume is insufficient, avoiding execution conflicts after command generation. The deep fusion of multi-dimensional features combined with a refined command mapping mechanism not only improves the accuracy of gas replenishment decisions, but also effectively solves the problem of poor adaptability of traditional methods under complex working conditions, and enhances the operational safety and reliability of the system in variable environments.

[0025] In this embodiment, the working process of the gas replenishment execution monitoring module includes: The system reads the variable frequency pulsed gas replenishment command, sends an opening signal to control the solenoid valve to be energized and opened, reads the voltage signal output by the flow sensor installed in the gas replenishment pipeline, and converts it into a value according to a preset ratio to obtain the instantaneous gas flow rate value. The preset ratio is obtained by calculating the ratio of the upper limit of the flow sensor's range to the maximum output value of the analog-to-digital converter chip. The upper limit of the flow sensor's range is 0 to 100 liters per minute to cover the peak gas replenishment flow rate, and the maximum output value of the analog-to-digital converter chip's range is 4096 to 65535 to ensure the measurement accuracy during the micro-gas replenishment stage. The number of high-frequency on / off cycles is obtained by dividing the preset gas replenishment start phase time by the short interval time. The solenoid valve is controlled to alternately open and close according to the number of high-frequency on / off cycles, so that the gas in the gas replenishment storage tank enters the gas chamber of the ring network cabinet in batches. The preset gas replenishment start phase time ranges from 0 seconds to 60 seconds to limit the pre-filling time and prevent pressure fluctuations. The short interval time ranges from 0.1 seconds to 0.5 seconds to generate high-frequency pulsating airflow within the mechanical response limit of the solenoid valve. The duration of the subsequent gas replenishment stage is calculated, and the duration of the subsequent gas replenishment stage is divided by the long interval duration to obtain the number of low-frequency on / off cycles. The solenoid valve is controlled to alternately open and close according to the number of low-frequency on / off cycles, and the gas flow rate from the gas replenishment storage tank into the ring network cabinet gas chamber is adjusted to a trace range. The duration of the subsequent gas replenishment stage is obtained by subtracting the duration of the preset gas replenishment start stage from the preset total gas replenishment duration. The preset total gas replenishment duration ranges from 10 minutes to 60 minutes to match the gas replenishment requirements under different leakage levels. The long interval duration ranges from 2 seconds to 5 seconds to allow for pressure balance buffer time to prevent gas replenishment overshoot.

[0026] In this embodiment, the operation process of the anomaly diagnosis and protection module includes: Read the instantaneous gas flow rate in the gas supply pipeline, add up the instantaneous gas flow rate values ​​at all times to calculate the total value, and mark the total value as the cumulative gas supply flow rate; The gas chamber volume parameters of the ring main unit are read, the cumulative replenishment flow rate is divided by the gas chamber volume parameters to obtain the unit volume replenishment volume, and the unit volume replenishment volume is multiplied by the preset gas state correction coefficient to obtain the predicted pressure recovery value. The preset gas state correction coefficient is obtained by calculating the product of the preset gas molar number and the ideal gas constant and dividing it by the current ambient temperature value. The current ambient temperature value ranges from 273.15 Kelvin to 313.15 Kelvin to cover the temperature range under normal operating conditions. Read the real-time gas pressure from the standardized input dataset, calculate the difference between the real-time gas pressure value at the start of gas replenishment and the real-time gas pressure value at the current time, and mark the difference as the actual monitored pressure increase. The absolute value of the difference between the predicted pressure rise and the actual monitored pressure rise is calculated. If the absolute value of the difference is greater than a preset anomaly judgment threshold, the gas supply is determined to be abnormal, and a fault protection interruption command is sent to control the solenoid valve to close. The preset anomaly judgment threshold is obtained by calculating the product of the pressure sensor's measurement accuracy and a preset safety margin. The measurement accuracy of the pressure sensor ranges from 0.1% to 0.5% of the full scale, and the preset safety margin ranges from 2 to 5 times to cover measurement errors and slight fluctuations.

[0027] In this embodiment, the working process of the gas replenishment verification management module includes: The system reads the real-time gas pressure from the standardized input dataset. If the real-time gas pressure is within the target replenishment pressure range, a shutdown signal is sent to control the solenoid valve to de-energize and close. The lower limit of the target replenishment pressure range is obtained by calculating the difference between a preset target replenishment pressure and a negative deviation threshold. The negative deviation threshold ranges from 0.005 MPa to 0.01 MPa to allow for slight pressure deficiency. The upper limit of the target replenishment pressure range is obtained by calculating the sum of a preset target replenishment pressure and a positive deviation threshold. The positive deviation threshold ranges from 0.005 MPa to 0.01 MPa to prevent over-replenishment. The average real-time gas pressure within a preset pressure monitoring period after the valve is closed is calculated. The average value is marked as the stable pressure after the valve is closed. The absolute value of the difference between the stable pressure after the valve is closed and the preset target gas replenishment pressure is calculated to obtain the pressure stability deviation. The pressure stability deviation is divided by the preset target gas replenishment pressure to calculate the pressure relative deviation rate. The pressure retention rate is obtained by subtracting the pressure relative deviation rate from the value 1. The preset pressure monitoring period ranges from 10 minutes to 30 minutes to verify the airtightness stability. The pressure retention rate is compared with a preset stability threshold. If the pressure retention rate is less than the preset stability threshold, it is determined that the stability standard is not met, and an air replenishment failure alarm signal is output. If the pressure retention rate is greater than or equal to the preset stability threshold, it is determined that the stability standard is met, and an air replenishment completion status signal is output and an air replenishment log is generated. The preset stability threshold ranges from 0.95 to 0.99 to ensure that the air chamber pressure is maintained at an effective level.

[0028] Example 1: To verify the feasibility of this invention in the field of online monitoring and intelligent gas replenishment closed-loop control of ring main unit gas leaks, a high-fidelity hardware-in-the-loop simulation verification platform based on actual field conditions was constructed. The simulation environment was built primarily based on the actual operation records of power distribution sites within the jurisdiction of the core power supply branch of this coastal first-tier city: First, using the Matlab / Simulink thermodynamics toolbox, a gas pressure-temperature coupled dynamic model of the gas chamber was built based on the actual gas state equation of SF6 / N2 mixture; Second, real environmental temperature, humidity, and salt spray concentration data for one year were introduced from the bureau's historical monitoring database, and combined with the aging model of the ring main unit's sealing materials, the slight influence of mechanical deformation caused by external corrosion and alternating internal pressure on the gas chamber volume was simulated; Finally, through hardware-in-the-loop (HIL) technology, the actual control board and solenoid valve drive circuit were connected to simulate the gas flow resistance of the gas replenishment pipeline and the action delay characteristics of the actuator.

[0029] In this simulation application scenario, the system of this invention connects to a simulated ring network cabinet sensor network and cabinet surface strain monitoring nodes through a simulation platform, constructing a full-dimensional perception layer. The system collects simulated real-time gas pressure, ambient temperature, cabinet surface strain values, and the status of the replenishment gas storage tank in real-time at a frequency of 5Hz, utilizing multiple heterogeneous data sources. Traditional monitoring methods often suffer from pressure readings that falsely report low pressure in the morning due to large diurnal temperature variations, or mask actual minor leaks during midday high temperatures. To address this issue, this invention utilizes a leak feature generation module to correct the gas chamber volume in real-time based on the thermodynamic equation of state and the cabinet surface strain values, decoupling the influence of temperature changes and mechanical deformation on pressure, and accurately identifying the actual gas leak trend. Once a leak is detected and the pressure drop trend does not match the temperature change, the replenishment decision module is activated, reading the leak rate feature vector, pressure difference value, and the current remaining gas volume in the replenishment gas storage tank, and inputting these into the replenishment decision module with its built-in improved iTransformer model. This model, through a heterogeneous feature embedding alignment module and a global state vector broadcast injection mechanism, dynamically integrates the key constraint of remaining gas volume into the timing prediction, outputting a variable-frequency pulsed gas replenishment command containing segmented gas replenishment timing. The gas replenishment execution module responds immediately, controlling the solenoid valve to perform short-interval high-frequency on / off switching at the beginning of the replenishment phase to achieve pre-filling balance. Subsequently, in the later stages of replenishment, it switches to long-interval low-frequency on / off switching for fine-tuned micro-gas replenishment. The entire process requires no manual intervention, preventing over-replenishment from impacting the cabinet and avoiding insulation hazards caused by insufficient gas replenishment. In addition, the anomaly diagnosis and protection module monitors the instantaneous gas flow rate in the gas replenishment pipeline in real time, calculates the predicted pressure recovery value through integration, and compares it with the measured value. Once an abnormality in gas replenishment efficiency is detected, the system is immediately interrupted, ensuring absolute system safety.

[0030] To comprehensively verify the robustness and reliability of this invention under different power distribution network operating conditions, this hardware-in-the-loop simulation platform specifically selected three types of regions with typical generational characteristics to construct differentiated test environments. The core business district, as the city's load center, focuses on simulating alternating thermal stress conditions with large diurnal temperature differences and drastic load fluctuations, rigorously testing the system's ability to decouple pressure fluctuations and resist false alarms under severe temperature fluctuations. The simulation environment for older industrial areas is based on severe aging of seals and high incidence of salt spray corrosion. By introducing material creep effects and erosion parameters from high-humidity, high-salt-spray media into the model, it simulates minute leaks and mechanical deformations under extreme conditions, focusing on assessing the system's early identification accuracy for low-rate, long-cycle real leaks. The power distribution equipment in newly built residential areas is relatively new, and the ambient temperature and humidity are relatively stable. Its simulation construction emphasizes baseline comparison to verify the system's high sensitivity and low-energy consumption performance under normal ideal operating conditions. By providing gradient coverage across the three scenarios—from extreme disturbances to stable operation—the comprehensiveness and engineering reference value of the simulation test data are ensured. During simulation tests based on the aforementioned typical operating conditions, the system of this invention successfully handled multiple sudden leakage events, accumulating detailed operational data. Table 1 below shows the performance comparison between the method of this invention and the traditional fixed-threshold gas replenishment method in simulation operation, covering multiple key dimensions such as leakage identification accuracy, gas replenishment time, pressure stability, and operation and maintenance costs. Table 1. Performance Comparison Data Between the Invention and Traditional Methods: Based on the comparative data shown in Table 1, it can be seen that the closed-loop control system based on the improved iTransformer model proposed in this invention has significant performance advantages over traditional methods in gas leakage monitoring and gas replenishment control of ring network cabinets. In particular, it has achieved comprehensive improvement in key indicators such as leakage identification accuracy, gas replenishment response speed, pressure control accuracy, and false alarm rate control.

[0031] Regarding leak detection accuracy, this invention maintains a high level of over 97% in all three typical application scenarios, far exceeding the average accuracy of approximately 79% for traditional systems. For example, in harsh, old industrial areas, traditional methods struggle to distinguish between temperature fluctuations and actual leaks, resulting in an accuracy of only 72.6%. However, this invention, by introducing a thermodynamic equation of state and a strain correction mechanism, improves the accuracy to 97.1%, effectively solving the problem of missed detections in complex environments.

[0032] Regarding response timeliness, this invention improves the global state vector broadcast injection mechanism of the iTransformer model, achieving real-time optimization of gas replenishment decisions and significantly reducing response latency. In old industrial areas, the average gas replenishment response latency is reduced from 210 seconds using traditional methods to 28 seconds, a response speedup of nearly 7.5 times, ensuring rapid intervention in sudden leaks.

[0033] Regarding the accuracy of air replenishment control, this invention utilizes a variable frequency pulsed air replenishment strategy to strictly control the air replenishment pressure overshoot rate from the traditional maximum of 15.8% to within 1.5%, while simultaneously reducing the average time for a single air replenishment cycle by approximately 30%, significantly improving operational efficiency. In terms of false alarm control, this invention employs multi-dimensional feature decoupling to reduce false alarms in older industrial areas from 28 to 2, greatly reducing maintenance costs. Overall, this invention also demonstrates excellent performance in pressure retention rate and energy consumption control after air replenishment, validating its practical application value in terms of high efficiency, accuracy, and energy saving.

[0034] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An on-line monitoring and intelligent air supplement closed-loop control system for gas leakage of a ring main unit, characterized in that, Includes the following modules: The data acquisition and preprocessing module is used to simultaneously acquire multi-source data from the ring main unit and preprocess it to obtain a standardized input dataset; The leakage feature generation module is used to calculate the theoretical gas pressure benchmark value based on the thermodynamic equation of state and the standardized input dataset, and to compare and decouple the calculation with the real-time gas pressure in the standardized input dataset and output the leakage rate feature vector. The gas replenishment decision module is used to read the leakage rate feature vector, calculate the pressure difference between the current real-time gas pressure and the preset target gas replenishment pressure, and combine it with the current remaining gas volume of the gas replenishment storage tank in the standardized input dataset and input it into the improved iTransformer model to output a variable frequency pulsed gas replenishment command. The improved iTransformer model includes a heterogeneous feature embedding and alignment module, a temporal projection module, an inverted variable attention module, a feedforward network module and an output generation module. The heterogeneous feature embedding and alignment module constructs a global state vector broadcast injection mechanism. The gas replenishment execution monitoring module is used to open the solenoid valve in response to the frequency conversion pulsed gas replenishment command, to achieve pre-filling balance in the initial stage of gas replenishment, to achieve micro-precise gas replenishment in the subsequent stage of gas replenishment, and to collect the instantaneous gas flow rate value in the gas replenishment pipeline. The abnormal diagnosis and protection module is used to calculate the cumulative gas supply flow rate from the instantaneous gas flow rate, then convert it into the predicted pressure rise value by combining the gas chamber volume parameter, and calculate the difference between the predicted pressure rise value and the actual monitored pressure rise value to determine whether the gas supply is abnormal and output the fault protection interrupt command. The gas replenishment verification management module is used to control the solenoid valve to close when the real-time gas pressure reaches the target gas replenishment pressure range, and continuously read the real-time gas pressure data to calculate the pressure holding rate. If it meets the stability standard, it outputs a gas replenishment completion status signal and generates a gas replenishment log.

2. The gas leakage online monitoring and intelligent air supplement closed-loop control system of a ring main unit according to claim 1, characterized in that, The working process of the data acquisition and preprocessing module includes: Real-time gas pressure, ambient temperature, strain values ​​on the cabinet surface, and the current remaining gas volume in the replenishment gas storage tank are collected as multi-source data. Outlier removal is performed on the multi-source data. The mean and standard deviation of each data point within the sliding time window are calculated. Data points whose absolute value differs from the mean by more than 3 times the standard deviation are marked as invalid and removed. Data cleaning is performed on the multi-source data after outlier removal. The moving average algorithm is used to calculate the average value of the data within the moving window to filter out high-frequency noise and normalize the data separately. The combined output is a standardized input dataset.

3. The online monitoring and intelligent gas replenishment closed-loop control system for gas leakage in a ring main unit according to claim 1, characterized in that, The working process of the leakage feature generation module includes: Read the ambient temperature and cabinet surface strain values ​​from the standardized input dataset, multiply the cabinet surface strain values ​​by the inner surface area of ​​the ring network cabinet air chamber to calculate the air chamber volume deformation, and add it to the initial volume of the ring network cabinet air chamber to obtain the current actual air chamber volume. Based on the thermodynamic equation of state, the theoretical pressure reference value of the gas at the current moment is calculated using the current actual gas chamber volume and ambient temperature. The pressure deviation residual is obtained by subtracting the real-time gas pressure in the standardized input dataset from the theoretical gas pressure benchmark and taking the absolute value. If the pressure deviation residual is greater than the preset first judgment threshold, the current time is marked as a potential leak time. Within a preset sliding time window, the first slope of the real-time gas pressure change with time and the second slope of the ambient temperature change with time are calculated respectively. The measured pressure and temperature change rate is obtained by dividing the first slope by the second slope. If the absolute value of the difference between the measured pressure and temperature change rate and the preset thermodynamic correlation threshold exceeds the preset allowable deviation range, it is determined that the pressure decrease trend does not match the temperature change. When it is determined that the pressure drop trend does not match the temperature change, the difference between the endpoint value and the starting value of the real-time gas pressure within the preset sliding time window is calculated. The difference is divided by the time length of the preset sliding time window to calculate the average leakage rate. The average leakage rate is marked and output as a leakage rate feature vector.

4. The online monitoring and intelligent gas replenishment closed-loop control system for gas leakage in a ring main unit according to claim 1, characterized in that, The working process of the gas replenishment decision module includes: Read the real-time gas pressure and the preset target gas replenishment pressure from the standardized input dataset, perform a subtraction operation to subtract the real-time gas pressure from the preset target gas replenishment pressure, and calculate the pressure difference value; Read the leakage rate feature vector and the current remaining gas volume of the gas replenishment tank in the standardized input dataset. Concatenate the leakage rate feature vector, the pressure difference value, and the current remaining gas volume of the gas replenishment tank in the feature dimension in order to construct a heterogeneous input vector. The heterogeneous input vector is input into the heterogeneous feature embedding and alignment module of the improved iTransformer model. The current remaining gas volume of the gas replenishment tank is extracted as the global state vector using the global state vector broadcast injection mechanism. The global state vector is copied to match the number of time steps of the leakage rate feature vector. The copied global state vector is added to the corresponding position values ​​of the leakage rate feature vector and the pressure difference value in the heterogeneous input vector. The embedded and aligned time series feature vector is output. The embedded and aligned temporal feature vector is input into the temporal projection module. A convolutional kernel of a preset size slides on the time axis of the embedded and aligned temporal feature vector. The values ​​within the coverage area of ​​the convolutional kernel are multiplied with the weight values ​​within the convolutional kernel and then summed to output the projected temporal features. The projected temporal features are input into the inverted variable attention module. The row and column indices of the projected temporal features are interchanged. The dot product between each row vector after transpose is calculated to obtain the original attention score. The score is then normalized and converted into a probability distribution as attention weights. The attention weights are multiplied with the transpose matrix. The row and column interchange is performed again on the multiplication result to restore the dimension. The inverted attention features are then output. The inverted attention features are input into the feedforward network module to perform a linear transformation, and an activation function is used to perform a nonlinear transformation on each value of the linear transformation result, outputting a feedforward feature vector; The weight matrix of the fully connected layer in the feedforward feature vector input-output generation module is multiplied with the feedforward feature vector. The result is mapped to the preset air replenishment command dimension. The numerical sequence obtained after mapping is used as the valve opening percentage at each moment. The results are combined in time sequence to form the frequency-controlled pulsed air replenishment command.

5. The online monitoring and intelligent gas replenishment closed-loop control system for gas leakage in a ring main unit according to claim 1, characterized in that, The working process of the gas replenishment execution monitoring module includes: Read the variable frequency pulsed gas replenishment command, send an opening signal to control the solenoid valve to be energized and opened, read the voltage signal output by the flow sensor installed in the gas replenishment pipeline, and convert it into a value according to a preset ratio to obtain the instantaneous gas flow rate value. Divide the preset gas replenishment start phase time by the short interval time to obtain the high frequency on / off count. Control the solenoid valve to alternately open and close according to the high frequency on / off count, so that the gas in the gas replenishment storage tank enters the ring network cabinet gas chamber in batches. Calculate the duration of the subsequent gas replenishment stage, divide the duration of the subsequent gas replenishment stage by the long interval duration to obtain the number of low-frequency on / off cycles, and control the solenoid valve to alternately open and close according to the number of low-frequency on / off cycles, thereby adjusting the gas flow rate from the gas replenishment storage tank into the ring network cabinet gas chamber to a trace range.

6. The online monitoring and intelligent gas replenishment closed-loop control system for gas leakage in a ring main unit according to claim 1, characterized in that, The operation process of the anomaly diagnosis and protection module includes: Read the instantaneous gas flow rate in the gas supply pipeline, and integrate the instantaneous gas flow rate at all times over time to obtain the cumulative gas supply flow rate; Read the air chamber volume parameters of the ring main unit, divide the cumulative replenishment flow rate by the air chamber volume parameters to obtain the unit volume replenishment volume, and multiply the unit volume replenishment volume by the preset gas state correction coefficient to obtain the predicted pressure recovery value. Read the real-time gas pressure from the standardized input dataset, calculate the difference between the real-time gas pressure value at the start of gas replenishment and the real-time gas pressure value at the current time, and mark the difference as the actual monitored pressure increase. Calculate the absolute value of the difference between the predicted pressure rise and the actual monitored pressure rise. If the absolute value of the difference is greater than the preset abnormal judgment threshold, the gas supply is judged to be abnormal, and a fault protection interruption command is sent to control the solenoid valve to close.

7. The online monitoring and intelligent gas replenishment closed-loop control system for gas leakage in a ring main unit according to claim 1, characterized in that, The working process of the gas replenishment verification management module includes: Read the real-time gas pressure from the standardized input dataset. If the real-time gas pressure is within the target replenishment pressure range, send a shutdown signal to control the solenoid valve to shut off. Calculate the average real-time gas pressure within the preset pressure monitoring time after the valve is closed, mark the average value as the stable pressure after valve closure, calculate the absolute value of the difference between the stable pressure after valve closure and the preset target gas replenishment pressure to obtain the pressure stability deviation, divide the pressure stability deviation by the preset target gas replenishment pressure to obtain the pressure relative deviation rate, and subtract the pressure relative deviation rate from the value 1 to obtain the pressure retention rate. The pressure holding rate is compared with the preset stability threshold. If the pressure holding rate is less than the preset stability threshold, it is determined that the stability standard is not met and an air replenishment failure alarm signal is output. If the pressure holding rate is greater than or equal to the preset stability threshold, it is determined that the stability standard is met.

Images