A low-voltage switch cabinet temperature and humidity monitoring and early warning method and system

Abstract

The embodiment of the application provides a low-voltage switch cabinet temperature and humidity monitoring and early warning method and system, relates to the technical field of power equipment state monitoring and early warning, and comprises the following steps: obtaining internal data of a switch cabinet, identifying and correcting abnormal values by using heat transfer physics constraints, constructing a temperature and humidity dynamic reference model based on heat conduction and diffusion equations, calculating theoretical temperature and humidity values and comparing them with measured data, calculating a dynamic deviation risk index, quantifying model parameter uncertainty by using Bayesian inference, obtaining an uncertainty interval of the risk index, predicting future risk indexes by using an improved long short-term memory network, updating cumulative risk quantities by using an adaptive forgetting factor, formulating a three-level early warning strategy by comprehensively considering current risks, future predictions and cumulative damages, and performing matched linkage control operations.The application realizes three-level threshold linkage control based on current risks, future predictions and cumulative damages, and significantly improves the accuracy, foresight and disposal accuracy of switch cabinet temperature and humidity monitoring.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and early warning technology, specifically to a method and system for monitoring and warning the temperature and humidity of low-voltage switchgear. Background Technology

[0002] In recent years, with the development of the Internet of Things, some studies have applied multi-sensor data fusion methods to switch cabinet monitoring, using data such as temperature, humidity, and load current to make judgments; at the same time, some machine learning methods have also been adopted in switch cabinet fault prediction, using historical data to train models for judgment.

[0003] Existing technologies have significant limitations in practical applications: Fixed threshold methods cannot adapt to load and environmental changes. When equipment operates in high-temperature seasons or under high-load conditions, fixed thresholds are prone to false alarms, while excessively high thresholds can lead to missed alarms. Furthermore, temperature and humidity exist as independent parameters, failing to consider the interconnected effects of insulation degradation. While partial discharge detection technology can detect insulation status, it requires dedicated sensors, which are expensive and susceptible to electromagnetic interference, exhibiting poor identification capabilities for low-frequency discharges and latent faults. Existing multi-parameter fusion methods merely superimpose data at the data level, lacking modeling of the physical coupling relationship between temperature, humidity fields, and electrical parameters, resulting in models that are difficult to interpret and have low generalization. Machine learning methods suffer from imbalanced samples and black-box models, making it difficult to combine equipment physical mechanisms with data-driven approaches. Moreover, current multi-parameter fusion methods often use fixed time windows for prediction, failing to dynamically adjust predictions based on equipment uncertainties. More importantly, current multi-parameter fusion methods do not consider the long-term cumulative damage to insulation caused by temperature and humidity, relying solely on instantaneous values, thus failing to provide early warnings. Summary of the Invention

[0004] The present invention aims to at least solve one of the technical problems existing in the prior art, and provides a method and system for monitoring and early warning of temperature and humidity in low-voltage switchgear.

[0005] In a first aspect, embodiments of the present invention provide a method for monitoring and early warning of temperature and humidity in a low-voltage switchgear, comprising the following specific steps: Obtain the original monitoring data inside the low-voltage switchgear, identify and correct outliers in the original monitoring data according to physical constraints, and obtain the monitoring data after cleaning. A dynamic temperature and humidity benchmark model is constructed based on the physical laws of heat and mass transfer. The theoretical temperature and relative humidity values ​​are obtained based on the monitoring data after cleaning and the dynamic temperature and humidity benchmark model. The dynamic deviation risk index is calculated based on the monitoring data after cleaning, the theoretical temperature and relative humidity values. Based on the dynamic deviation risk index and the cleaned monitoring data, the uncertainty of the model parameters is quantified by Bayesian inference method to obtain the uncertainty range of the risk index, and an improved long short-term memory network model is constructed to predict the dynamic deviation risk index. Based on the dynamic deviation risk index, the cumulative risk is updated through an adaptive forgetting factor. Based on the dynamic deviation risk index, the cumulative risk, and the cumulative risk, a graded early warning strategy is formulated, and linkage control operations matching the early warning level are executed.

[0006] Optionally, the step of identifying and correcting outliers in the original monitoring data based on physical constraints to obtain cleaned monitoring data includes: Using the physical constraint that the rate of change of temperature gradient between adjacent locations within the same switch cabinet does not exceed the maximum allowable value of the material's thermal diffusivity, the difference in the rate of change of temperature gradient between each monitoring point and adjacent monitoring points is calculated, and monitoring point data with a difference in the rate of change of temperature gradient exceeding the maximum value are identified as outliers. Temperature data, relative humidity data, and partial discharge signal data identified as outliers are replaced with the average of the data from the previous moment and the data from the next moment. Monitoring points that are identified as outliers multiple times in a row are judged as sensor faults and fault alarm information is generated. The data after outlier identification and correction is used as the cleaned monitoring data.

[0007] Optionally, the step of constructing a dynamic temperature and humidity benchmark model based on the physical laws of heat and mass transfer, and obtaining theoretical temperature and relative humidity values ​​based on the monitoring data after cleaning and the dynamic temperature and humidity benchmark model, includes: The control equations for the internal temperature and humidity fields of the switchgear are established. The control equations for the temperature field include internal heat source terms related to real-time load current and contact resistance. The switchgear is then subjected to three-dimensional geometric modeling and mesh generation, and boundary conditions for convective heat transfer with the external environment are set. The initial temperature and humidity are set to ambient temperature and ambient humidity, respectively. The governing equations are solved to obtain the theoretical temperature distribution field and the theoretical relative humidity distribution field. The theoretical temperature and theoretical relative humidity values ​​that match the location of each monitoring point are extracted from the theoretical temperature distribution field and the theoretical relative humidity field and used as the theoretical temperature and theoretical relative humidity values.

[0008] Optionally, the step of calculating the dynamic deviation risk index based on the monitoring data after cleaning, the theoretical temperature value, and the theoretical relative humidity value includes: Calculate the ratio of the absolute value of the deviation between the temperature data and the theoretical temperature value at each monitoring point to the rated operating temperature to obtain the temperature normalized deviation; and calculate the ratio of the absolute value of the deviation between the relative humidity data and the theoretical relative humidity value at each monitoring point to the critical relative humidity to obtain the humidity normalized deviation. The partial discharge signal intensity is compared with a preset saturation threshold to obtain the normalized partial discharge signal intensity. The normalized temperature deviation, the normalized humidity deviation and the preset weighting coefficient are multiplied together, and the normalized partial discharge signal intensity and the normalized temperature deviation and the preset weighting coefficient are multiplied together to obtain the dynamic deviation risk index.

[0009] Optionally, the step of quantifying the uncertainty of model parameters and obtaining the uncertainty range of the risk index by using Bayesian inference methods based on the dynamic deviation risk index and the cleaned monitoring data includes: The physical parameters in the control equation of the temperature field and the weighting coefficients in the calculation process of the dynamic deviation risk index are defined as random variables, and the prior distribution of the random variables is determined according to the equipment factory parameters and material properties. Collect normal operating condition data for a preset duration, establish the posterior distribution of random variables based on the normal operating condition data, calculate the posterior prediction distribution of the dynamic deviation risk index based on the posterior distribution for the current monitoring data, and calculate the preset confidence interval width of the posterior prediction distribution. Use the preset confidence interval width as the uncertainty interval of the risk index. Among them, normal operating condition data refers to historical monitoring data in which the switchgear has no fault records and the dynamic deviation risk index is lower than the preset value.

[0010] Optionally, the construction of the improved long short-term memory network model to predict the dynamic deviation risk index includes: The cleaned monitoring data of each monitoring point, the dynamic deviation risk index, the rate of change of the dynamic deviation risk index, and the uncertainty range of the risk index are constructed into a multi-dimensional input feature vector. An attention mechanism layer is set up to perform weighted processing on the input sequence with the uncertainty range of the risk index as the attention weight. The system sets up a long short-term memory network layer to receive the weighted input sequence and extract temporal features. A fully connected output layer is set up to map the temporal features to predicted values ​​for multiple future time steps. The dynamic deviation risk index of future time steps is used as the dynamic deviation risk index of the predicted future time steps.

[0011] Optionally, updating the cumulative risk amount based on the dynamic deviation risk index using an adaptive forgetting factor includes: setting the current cumulative risk amount as the adaptive forgetting factor, adding the product of the current cumulative risk amount and the previous cumulative risk amount to the current dynamic deviation risk index, setting the adaptive forgetting factor as the product of the basic forgetting factor and a hyperbolic tangent function value minus the current dynamic deviation risk index, and setting the initial cumulative risk amount to zero, and recursively calculating the cumulative risk amount. The step of formulating a tiered early warning strategy based on the dynamic deviation risk index, the cumulative risk level, and executing coordinated control operations matching the early warning level includes: Set the first, second, and third thresholds for the dynamic deviation risk index, and set the first, second, and third thresholds for the cumulative risk amount; When the dynamic deviation risk index of the monitoring point reaches the third threshold, or the cumulative risk reaches the third threshold, or the dynamic deviation risk index reaches the second threshold, the first warning level is triggered; when the first warning level is triggered, the linkage control operation of cutting off the power supply of the fault circuit and starting the fire extinguishing device spraying is executed. When the dynamic deviation risk index of the monitoring point reaches the second threshold but not the third threshold, or when the cumulative risk reaches the second threshold but not the third threshold, or when the dynamic deviation risk index reaches the first threshold, the second warning level is triggered. When the second warning level is triggered, the linkage control operation of starting the dehumidification device and cooling fan and sending alarm information to the operation and maintenance terminal is executed. When the dynamic deviation risk index of a monitoring point reaches the first threshold but not the second threshold, or when the cumulative risk reaches the first threshold but not the second threshold, the third warning level is triggered. When the third warning level is triggered, the linkage control operation of recording the warning event and generating a prompt mark on the monitoring platform is executed.

[0012] Secondly, embodiments of the present invention provide a low-voltage switchgear temperature and humidity monitoring and early warning system, which applies a low-voltage switchgear temperature and humidity monitoring and early warning method as described above, including a data acquisition module, a data cleaning module, a benchmark model and risk calculation module, an uncertainty quantification and prediction module, and a cumulative assessment and decision execution module; The data acquisition module is used to acquire raw monitoring data inside the low-voltage switchgear. The data cleaning module is used to identify and correct outliers in the original monitoring data according to physical constraints, so as to obtain cleaned monitoring data. The benchmark model and risk calculation module are used to construct a dynamic benchmark model of temperature and humidity based on the physical laws of heat and mass transfer, and to obtain theoretical temperature and theoretical relative humidity values ​​based on the monitoring data after cleaning and the dynamic benchmark model of temperature and humidity. Based on the monitoring data after cleaning, theoretical temperature and theoretical relative humidity values, the dynamic deviation risk index is calculated. The uncertainty quantification and prediction module is used to quantify the uncertainty of model parameters based on the dynamic deviation risk index and the cleaned monitoring data using Bayesian inference methods, obtain the uncertainty range of the risk index, and construct an improved long short-term memory network model to predict the dynamic deviation risk index. The cumulative assessment and decision execution module is used to update the cumulative risk amount through an adaptive forgetting factor based on the dynamic deviation risk index, and to formulate a graded early warning strategy based on the dynamic deviation risk index, the dynamic deviation risk index and the cumulative risk amount, and to execute linkage control operations that match the early warning level.

[0013] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the low-voltage switchgear temperature and humidity monitoring and early warning method described above.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the low-voltage switchgear temperature and humidity monitoring and early warning method described above.

[0015] The low-voltage switchgear temperature and humidity monitoring and early warning method and system of this invention introduces heat transfer physical constraints for outlier identification, using the temperature gradient change rate of adjacent nodes as the judgment criterion instead of conventional statistical methods, giving the data cleaning process a clear physical basis and eliminating physically impossible outliers; it constructs a dynamic temperature and humidity benchmark model based on the heat conduction equation and diffusion equation, substituting the real-time load current as a heat source term into the solution, making the theoretical temperature and humidity distribution field conform to the actual operating conditions of the equipment, and providing a dynamic reference system for deviation assessment; when calculating the dynamic deviation risk index, it introduces a critical humidity threshold and a partial discharge weighting term, coupling the three parameters of temperature, humidity, and partial discharge together, solving the problem that the synergistic effect caused by the independent treatment of the three in traditional methods is ignored, and quantifying the causal chain between temperature and humidity anomalies and insulation degradation into a calculable index; it uses Bayesian inference to quantify the uncertainty of model parameters, treating the heat conduction parameters and weighting coefficients as random variables, and calculating... The confidence interval width of the risk index improves the credibility of the system's judgment results. By constructing an improved long short-term memory network model with uncertainty intervals as attention weights, the prediction model automatically devotes more attention to time periods with ambiguous states and difficult-to-judge risks during training and prediction, realizing the proactive guidance of the prediction process by the uncertainty quantification results. An adaptive forgetting factor is designed to update the cumulative risk amount. When healthy, historical information has a high weight and slow damage accumulates over a long period of time. When the risk increases, history is quickly forgotten and the current state takes the lead, simulating the risk memory mechanism in the physical world and solving the problem that traditional accumulation methods cannot distinguish the information weights of different risk stages. By establishing a three-level threshold decision tree, the current risk index, future predicted value, and cumulative risk amount are integrated to make graded judgments. Based on the trigger conditions, differentiated linkage operations are executed, such as cutting off power, starting fire extinguishing, starting dehumidification, sending alarms, and recording events, so that the risk assessment results are translated into precise disposal measures. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 The above is a flowchart illustrating a method for monitoring and early warning of temperature and humidity in a low-voltage switchgear, as provided in one embodiment of the present invention. Figure 2 A schematic diagram of the linkage control operation process that matches the warning level to the execution of a low-voltage switchgear temperature and humidity monitoring and early warning method provided in one embodiment of the present invention; Figure 3 The present invention provides a system scheme flowchart for a low-voltage switchgear temperature and humidity monitoring and early warning system according to an embodiment of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0020] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0021] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0022] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0023] Example 1, referring to Figure 1 As one embodiment of the present invention, a method for monitoring and early warning of temperature and humidity in a low-voltage switchgear is provided, comprising: S100: Acquire the original monitoring data inside the low-voltage switchgear, identify and correct outliers in the original monitoring data according to physical constraints, and obtain the cleaned monitoring data.

[0024] S200: Based on the physical laws of heat and mass transfer, a dynamic temperature and humidity benchmark model is constructed. Based on the monitoring data after cleaning and the dynamic temperature and humidity benchmark model, theoretical temperature and theoretical relative humidity values ​​are obtained. Based on the monitoring data after cleaning, theoretical temperature and theoretical relative humidity values, the dynamic deviation risk index is calculated.

[0025] S300: Based on the dynamic deviation risk index and the cleaned monitoring data, the uncertainty of the model parameters is quantified by Bayesian inference to obtain the uncertainty range of the risk index, and an improved long short-term memory network model is constructed to predict the dynamic deviation risk index.

[0026] S400: Based on the dynamic deviation risk index, the cumulative risk amount is updated through an adaptive forgetting factor. Based on the dynamic deviation risk index, the cumulative risk amount, a graded early warning strategy is formulated, and linkage control operations matching the early warning level are executed.

[0027] It should be noted that this invention integrates three types of parameters—temperature, humidity, and partial discharge—into a dynamic deviation risk index, from physical sensing, data cleaning, benchmark modeling, coupled quantification, uncertainty analysis, intelligent prediction, cumulative assessment to hierarchical decision-making. It also introduces Bayesian uncertainty quantification and adaptive accumulation mechanisms to achieve three-level threshold linkage control based on current risk, future prediction, and cumulative damage, which significantly improves the accuracy, foresight, and precision of switchgear temperature and humidity monitoring.

[0028] Example 2, refer to Figure 1 and Figure 2 This is a second embodiment of the present invention, which provides a method for monitoring and early warning of temperature and humidity in a low-voltage switchgear, including: In step S100, acquiring the raw monitoring data inside the low-voltage switchgear includes steps S101 to S104: S101: Deploy a heterogeneous sensor network consisting of integrated temperature and humidity sensors and ultra-high frequency partial discharge sensors in the target low-voltage switchgear. Deploy sensor nodes at key locations such as bus connection points, cable terminals, circuit breaker moving and stationary contacts, and the surface of insulation support components in the switchgear. A total of eight sensor nodes are deployed. The number of sensors is determined according to the size and internal structural complexity of the switchgear, and can cover all key heat points and weak insulation locations.

[0029] S102: Synchronously collect data from each sensor node using a unified time reference, setting the collection frequency to once per second. The current frequency can capture rapid changes in the device status while avoiding data redundancy.

[0030] S103: For each sensor node, at the sampling time, the following three types of raw monitoring data are acquired through the sensor: temperature data is acquired through the thermistor element in the integrated temperature and humidity sensor and marked as temperature data in degrees Celsius; relative humidity data is acquired through the humidity-sensitive capacitor element in the integrated temperature and humidity sensor and marked as relative humidity data in percentage; electromagnetic wave signals generated by partial discharge are acquired through UHF sensor coupling, and the current signal is processed by the preamplifier circuit to obtain partial discharge signal intensity data, marked as partial discharge signal intensity data in millivolts.

[0031] S104: Organize the three types of data according to timestamps and node numbers to construct an original data matrix. The original data matrix contains the temperature data, relative humidity data, and partial discharge signal intensity data of all sensor nodes at the current moment.

[0032] Furthermore, in step S100, obtaining the cleaned monitoring data includes cleaning the obtained original data matrix to remove outliers caused by sensor malfunctions or strong external electromagnetic interference, specifically including steps S111~S114: S111: An outlier identification method based on physical constraints is adopted, and physical constraints are set according to the principle of heat transfer.

[0033] It should be noted that, based on the thermal diffusivity of the copper busbar and the natural convection heat transfer coefficient of the air inside the switchgear, the physical constraint threshold is set to 0.5 degrees Celsius per second. The thermal diffusivity of the copper busbar is approximately 1.1 x 10^-4 square meters per second. Combined with the natural convection heat transfer coefficient of the air inside the switchgear, theoretical calculations and experimental verification have determined that it is physically impossible for the temperature gradient change rate between adjacent nodes to exceed 0.5 degrees Celsius per second. Therefore, 0.5 can be used as a reliable basis for outlier detection.

[0034] S112: For each time step, calculate the difference in the rate of change of temperature gradient between all adjacent sensor node pairs.

[0035] The system calculates the temperature change rate of each node by the ratio of the temperature difference between two consecutive time points to the time interval. When the system detects that the difference between the temperature change rate of a node and the temperature change rates of all its neighboring nodes exceeds 0.5 degrees Celsius per second, and determines that the temperature value of the current node at the current time is significantly inconsistent with the historical trend of the current node, the system identifies the current data point as an outlier.

[0036] S113: When the system detects an outlier, it uses linear interpolation to correct it.

[0037] For temperature data that is identified as abnormal at the current moment, the temperature data of the current node at the previous moment and the temperature data at the next moment are obtained. Half of the sum of the temperature data at the previous moment and the temperature data at the next moment is used as the correction value. The correction method is based on the physical characteristic of continuous change of temperature field and can provide a reasonable estimate.

[0038] S114: For relative humidity data and partial discharge signal intensity data, the same physical constraint logic is used for outlier identification and correction. When three or more outliers are identified consecutively, it is determined that there is a hardware fault in the current sensor node, a sensor fault alarm message is generated and pushed to the maintenance personnel.

[0039] Through physical constraint-driven outlier identification and correction, reliable monitoring data after cleaning is generated. The generated data is recorded as temperature data, relative humidity data, and partial discharge signal intensity data after cleaning.

[0040] In step S200, constructing the dynamic temperature and humidity reference model includes, based on the physical equations of heat and mass transfer, calculating the temperature and humidity distribution field under theoretical standard operating conditions, and constructing a dynamic reference model that can reflect the temperature and humidity distribution law inside the switchgear, specifically including steps S201~S204: S201: Establish the control equations for the temperature field inside the switchgear. Based on the theory of heat conduction, set the temperature at any point inside the switchgear at any given time to satisfy the following formula:

[0041] in, The air density inside the switchgear is set to 1.2 kg per cubic meter, based on the air density under standard atmospheric pressure. The specific heat capacity of air is set to 1005 joules per kilogram, based on the values ​​of air's thermophysical parameters. The thermal conductivity of air is set to 0.026 watts per meter Kelvin, based on the values ​​of air's thermophysical properties. This is the partial derivative of temperature with respect to time, reflecting the rate of change of temperature over time. For the Laplace operator acting on the temperature field, it represents the sum of the spatial second-order partial derivatives of temperature. The internal heat source intensity represents the heat generated per unit volume per unit time.

[0042] It should be noted that the intensity of the internal heat source is determined based on the heating characteristics of the electrical equipment in the switchgear. The heat source is mainly set as the contact resistance heating from the bus connection point and the moving and stationary contacts of the circuit breaker. For the location of the node, the intensity of the heat source is expressed as the product of the load distribution factor, the square of the real-time load current, and the nominal value of the contact resistance at the node.

[0043] The real-time load current is acquired through a current transformer installed on the incoming side of the switchgear to measure the total current flowing through the switchgear in real time. The contact resistance at the node is determined according to the equipment's factory parameters, which are provided by the manufacturer when the equipment is put into operation. The load distribution factor is pre-calibrated through finite element simulation, and a three-dimensional model is established based on the internal structure of the switchgear. By applying different currents, simulation calculations are performed to obtain the current distribution coefficient of each node.

[0044] S202: Establish the control equation for the humidity field. Based on the mass transfer theory, set the relative humidity at any point inside the switch cabinet to satisfy the diffusion equation.

[0045] The diffusion equation can be expressed as the partial derivative of relative humidity with respect to time equal to the diffusion coefficient multiplied by the Laplace operator of relative humidity and the intensity of the humidification source.

[0046] The diffusion coefficient is set to 2.5 x 10^-5 square meters per second, based on the diffusion characteristics of water vapor in the air. The moisture source intensity takes into account the moisture released from the insulating material inside the switchgear and the moisture infiltrated from the external environment. It is set as a constant based on the sealing performance of the switchgear and the moisture absorption characteristics of the material. The moisture source intensity parameter is obtained by experimental measurement, placing a dry and clean switchgear in a constant temperature and humidity environment and measuring the change curve of internal humidity over time.

[0047] S203: Numerical solution based on the finite element method using the governing equations.

[0048] A three-dimensional geometric model of the switchgear was established, which includes the main components such as the cabinet, busbars, circuit breakers, and insulation supports. The model size is completely consistent with the actual switchgear. The geometric model was meshed to generate about 500,000 computational nodes. The mesh density was increased at key locations (such as contact connections) to ensure computational accuracy.

[0049] It should be noted that the system boundary conditions include: convective heat exchange between the outer surface of the cabinet and the external environment, with the heat transfer coefficient determined based on the wind speed on the cabinet surface; a wind speed sensor is installed on the outer surface of the cabinet to measure the wind speed on the cabinet surface in real time, and the current heat transfer coefficient is calculated based on the wind speed-heat transfer coefficient relationship curve; ventilation opening boundaries are set at the bottom and top of the cabinet to allow airflow, and pressure boundary conditions are set at the ventilation openings.

[0050] It should also be noted that the initial system settings include setting the initial temperature value to the ambient temperature, which is set to 25 degrees Celsius according to standard environmental conditions; and setting the initial relative humidity value to the ambient relative humidity, which is set to 50% according to standard environmental conditions.

[0051] At each moment, the heat source intensity is calculated based on the real-time load current, and the heat source intensity is substituted into the heat conduction equation to obtain the theoretical temperature distribution field; at the same time, the diffusion equation is solved to obtain the theoretical relative humidity distribution field; for the location of each sensor node, the corresponding theoretical temperature value and theoretical relative humidity value are extracted from the solution results.

[0052] S204: Complete the construction of the dynamic temperature and humidity benchmark model, obtain the temperature and humidity distribution field under theoretical standard operating conditions, and use the theoretical temperature value and theoretical relative humidity value output by the model as dynamic benchmarks to calculate the degree of deviation from the measured values.

[0053] Furthermore, in step S200, the calculation of the dynamic deviation risk index includes constructing a dynamic deviation risk index that comprehensively reflects the degree of deviation between the current state and the theoretical health state of the equipment, based on real-time monitoring data after cleaning and dynamic benchmark model values. This index serves as the core input for uncertainty quantification and cumulative risk assessment, specifically including steps S211~S214: S211: The ratio of the absolute value of the temperature deviation of the computing node at the current moment to the rated operating temperature.

[0054] The rated operating temperature of the switchgear is obtained through a current transformer and is set to 85 degrees Celsius according to the equipment insulation class.

[0055] The basis for setting the rated operating temperature includes the fact that the heat resistance class of the insulation material in the switchgear is usually Class B or Class F. The maximum allowable temperature of Class B insulation material is 130 degrees Celsius. However, considering the safety margin and long-term operational reliability, the rated operating temperature is set at 85 degrees Celsius, which is within the range specified by industry standards and can provide sufficient advance warning.

[0056] S212: The ratio of the absolute value of the relative humidity deviation of the calculation node at the current moment to the critical relative humidity.

[0057] Based on experimental data showing that the surface resistivity of various insulating materials begins to decrease exponentially under critical relative humidity conditions, the critical relative humidity is set at 65%.

[0058] The criteria for setting the critical value include the fact that surface resistivity tests were conducted on three commonly used insulating materials—polyester film, epoxy resin, and polyimide—under different humidity conditions. It was found that when the relative humidity exceeded 65%, the surface resistivity of all three materials decreased significantly. Among them, the decrease in epoxy resin was more than an order of magnitude. Therefore, 65% was determined as the critical relative humidity.

[0059] S213: Perform Min-Max normalization on the partial discharge signal intensity, and record the normalized partial discharge signal intensity as the normalized partial discharge signal intensity. Determine the value of the normalized partial discharge signal intensity by comparing the cleaned partial discharge signal intensity with the preset partial discharge signal saturation threshold.

[0060] Among them, by statistically analyzing a large amount of historical data on partial discharge of switchgear, it was found that the partial discharge signal intensity rarely exceeds 100 millivolts under normal operating conditions. When the signal intensity exceeds 100 millivolts, it indicates that there is significant insulation degradation and should be given high weight. Therefore, the partial discharge signal saturation threshold is set to 100 millivolts.

[0061] S214: The two normalized deviations are linearly combined, and the partial discharge signal is introduced as a weighting factor to form a dynamic deviation risk index, the calculation formula of which is expressed as:

[0062] in, The dynamic deviation risk index is dimensionless. This is the ratio of the absolute value of the temperature deviation to the rated operating temperature, dimensionless. The relative humidity deviation weighting coefficient was determined to be 0.3 using the analytic hierarchy process (AHP) combined with historical fault data. This is the ratio of the absolute value of the relative humidity deviation to the critical relative humidity, dimensionless. The partial discharge weighting coefficient is determined to be 0.2 by combining the analytic hierarchy process with historical fault data. The normalized partial discharge signal intensity ranges from 0 to 1 and is dimensionless.

[0063] It should be noted that the first term in the calculation of the dynamic deviation risk index reflects the degree of temperature deviation; the larger the value, the more the equipment temperature deviates from the theoretical healthy state. The second term reflects the degree of relative humidity deviation, and the value is multiplied by a weighting coefficient and then added to the first term. The third term reflects the weighted amplification of temperature deviation by the partial discharge signal. The weighting term reflects that when partial discharge occurs, a small temperature deviation may indicate a more serious insulation fault. The formula establishes a reference system based on the theoretical healthy state and uses the partial discharge signal to nonlinearly weight the temperature and humidity deviations, thereby achieving a quantitative expression of the insulation degradation caused by temperature and humidity anomalies and the resulting partial discharge.

[0064] In step S300, obtaining the uncertainty range of the risk index includes introducing a Bayesian inference method to quantify the uncertainty of model parameters existing in the calculation of the dynamic deviation risk index, specifically including steps S301 to S305: S301: Define the set of parameters to be quantified. Set the set of parameters as the key parameters of the heat conduction equation in step S201 and the weighting coefficients in step S214. Specifically, these include the relative humidity deviation weighting coefficient, the partial discharge weighting coefficient, the thermal conductivity, and the contact resistance of the contact point. Treat the above parameters as random variables and determine the prior distribution of each parameter based on the equipment's factory parameters and material properties.

[0065] In this embodiment, the prior distribution of each parameter is set to a normal distribution, the mean is taken as the nominal value, and the standard deviation is taken as 10% of the nominal value. Among them, the equipment factory parameters and material properties have manufacturing tolerances and are normally distributed. Taking 10% of the nominal value as the standard deviation can cover most of the manufacturing tolerance range.

[0066] S302: Collect normal operating condition data for a period of time. Normal operating condition is defined as a historical data period in which the switchgear has no fault records, the dynamic deviation risk index is continuously lower than 0.1, and the partial discharge signal strength is lower than 10 millivolts.

[0067] This embodiment collects 30 consecutive days of normal operating condition data, obtaining approximately 2.6 million data points. Bayesian inference requires sufficient samples to obtain a stable posterior distribution. The 30 days of data can cover the operating status of the equipment under different loads and environmental conditions, while not introducing parameter drift caused by equipment aging.

[0068] S303: Based on Bayes' theorem, establish the posterior distribution of the parameter set, and express the posterior distribution as the product of the likelihood function and the prior distribution, divided by the marginal likelihood function.

[0069] Here, the prior distribution is the prior distribution of the parameter set, and the likelihood function represents the probability of observing normal operating data under given parameter conditions. In this embodiment, the likelihood function is established based on the calculation error of the dynamic deviation risk index. It is assumed that the error follows a normal distribution with a mean of 0 and an unknown variance, and that this assumption is based on the central limit theorem, applicable to the superposition of a large number of independent random errors.

[0070] Due to the high dimensionality of the parameter space, this embodiment uses the Markov chain Monte Carlo method to sample the posterior distribution. The Metropolis-Hastings algorithm is used, with four parallel Markov chains set up, and each chain is set to sample 10,000 times. After 2,000 burns, the sample set of the parameter posterior distribution is obtained. The burn period is used to eliminate the influence of the initial value on the sampling result and ensure that the sampling result converges to the target distribution.

[0071] S304: For the new monitoring data, namely the temperature data, relative humidity data, and partial discharge signal intensity data after cleaning at the current moment, calculate the posterior prediction distribution of the dynamic deviation risk index, and express the posterior prediction distribution as the integral of the posterior distribution and the conditional distribution of the risk index under given parameters. The current integral is approximated by the posterior sample.

[0072] S305: For each posterior parameter sample, the system substitutes it into the formula to calculate the corresponding dynamic deviation risk index, obtains the sample set of dynamic deviation risk index, and calculates the statistic of the posterior prediction distribution based on the sample set, outputting the uncertainty interval.

[0073] It should be noted that the uncertainty interval is defined as the width of the 90 percentile confidence interval in the posterior prediction distribution, that is, the difference between the 95th percentile and the 5th percentile of the posterior prediction distribution. The uncertainty interval is used as a quantitative indicator to measure the unpredictability of the equipment state. The larger the value of the uncertainty interval, the more unpredictable the equipment state and the higher the risk. The uncertainty interval is also used as the attention weight of the improved long short-term memory network model, so that the model pays more attention to the time period when the state is ambiguous and the risk is difficult to judge.

[0074] Furthermore, in step S300, the prediction of the dynamic deviation risk index includes introducing an uncertainty interval as the weight of the attention mechanism into the conventional long short-term memory network model, constructing an improved long short-term memory network model, so that the model pays more attention to the time period when the state is ambiguous and the risk is difficult to judge during training and prediction, and predicts the dynamic deviation risk index, specifically including steps S311~S314: S311: Construct the input feature vector of the model. The input feature vector of the node at the current time contains six dimensions: the first dimension is the temperature data after cleaning, the second dimension is the relative humidity data after cleaning, the third dimension is the partial discharge signal intensity data after cleaning, the fourth dimension is the calculated dynamic deviation risk index, the fifth dimension is the rate of change of the dynamic deviation risk index, which is calculated by forward differencing and set as the difference between the dynamic deviation risk index at the current time and the dynamic deviation risk index at the previous time, and the sixth dimension is the calculated uncertainty interval width.

[0075] S312: Construct the structure of an improved long short-term memory network model.

[0076] The first layer is the input layer, which is set to receive input sequences within a time window of length sixty, that is, input historical data of the past 60 seconds; the selection criteria for the length of the time window is: the time constant of the temperature and humidity change of the switch cabinet is usually on the order of tens of seconds, and a 60-second window can capture the complete dynamic change process.

[0077] The second layer is the attention mechanism layer. It uses the width of the uncertainty interval as the attention weight. For each time point in the input sequence, the attention weight for that time point is set to the softmax function value of the uncertainty interval width at that time point. The softmax function then normalizes the uncertainty interval width into a probability distribution, ensuring that the sum of all weights is 1. This attention mechanism enables the model to give higher attention to time points with high uncertainty when calculating the hidden state, demonstrating the guiding role of quantization results in the prediction model.

[0078] The third layer is the first long short-term memory network layer, which contains 128 hidden units. It receives a weighted input sequence and outputs a hidden state sequence.

[0079] The fourth layer is the second long short-term memory network layer. It is set to contain 128 hidden units, receive the output of the first layer, and further extract temporal features.

[0080] The fifth layer is the Dropout layer, with a dropout rate of 0.3 to prevent the model from overfitting. The dropout rate is determined through cross-validation.

[0081] The sixth layer is a fully connected output layer that maps the output of the second long short-term memory network layer to a four-dimensional output space.

[0082] S313: The improved long short-term memory network model outputs predicted values ​​for multiple future time steps.

[0083] Based on the sampling frequency and early warning response time requirements, the prediction step size is set to 12, meaning the system predicts the state in the next 12 seconds.

[0084] The output vector is set to four dimensions: output predicted future temperature data, output predicted future relative humidity data, output predicted future partial discharge signal intensity data, and output predicted future dynamic deviation risk index.

[0085] S314: Train the improved long short-term memory network model using historical monitoring data from the past year as the training set, which contains approximately 31.5 million data points. Use data from 48 hours before the failure as positive samples and normal operating data as negative samples. Set the loss function to root mean square error and add a penalty term to punish inaccurate predictions for periods of high uncertainty.

[0086] Set the optimizer to Adam optimizer, set the initial learning rate to 0.001, decrease the learning rate by 10% every 10 training epochs, and train for 100 training epochs. Terminate training early when the validation set loss no longer decreases for 10 consecutive training epochs.

[0087] In step S400, updating the cumulative risk amount through an adaptive forgetting factor includes designing a cumulative risk function with an adaptive forgetting factor, dynamically adjusting the weights of historical risks through the adaptive forgetting factor, capturing the long-term cumulative damage to insulation caused by temperature, humidity, and partial discharge, and simulating the risk memory mechanism in the physical world, specifically including steps S401~S404: S401: Establish an update function for the cumulative risk amount, and update the cumulative risk amount recursively over time at the node.

[0088] The update function includes the following: the cumulative risk at the current moment is equal to the adaptive forgetting factor multiplied by the cumulative risk at the previous moment plus the dynamic deviation risk index at the current moment. It is similar to the traditional cumulative sum algorithm, but this embodiment introduces an adaptive forgetting factor.

[0089] S402: Define the calculation method for the adaptive forgetting factor. In this embodiment, the adaptive forgetting factor is set to be related to the dynamic deviation risk index at the current moment. The formula for calculating the adaptive forgetting factor is expressed as:

[0090] in, This is an adaptive forgetting factor, ranging from 0 to 1, and is dimensionless. The basic forgetting factor, calibrated to 0.99 based on the aging rate of insulating materials under normal operating conditions, is used as the basic forgetting factor. It is a hyperbolic tangent function, and its range is from -1 to 1.

[0091] By fitting the accelerated aging experimental data of insulating materials, the basic forgetting factor was found to be consistent with the exponential decay model of the cumulative damage to insulation performance. With a time constant of 100 sampling periods, the forgetting factor of 0.99 can fit the experimental data well.

[0092] S403: Analyze the characteristics of the adaptive forgetting factor. When the dynamic deviation risk index is small at the current moment, i.e. the equipment is in good condition, the hyperbolic tangent function is close to 0, the adaptive forgetting factor is close to the basic forgetting factor, i.e., 0.99, the historical information has a high contribution weight to the cumulative amount, the cumulative effect is significant, and it can effectively reflect the small damage that accumulates slowly over a long period of time.

[0093] When the dynamic deviation risk index is large at the current moment, i.e. the risk is very high, the hyperbolic tangent function is close to 1, the adaptive forgetting factor is close to 0, historical information is quickly forgotten, and the accumulated risk is mainly determined by the current high risk value. This simulates the risk memory mechanism in the physical world, i.e., the small damage accumulated over a long period of time in a healthy state cannot be ignored, while in a high-risk state, the current state plays a decisive role and the impact of historical small damage decays rapidly.

[0094] S404: Set the initial value of the cumulative risk to 0. Starting from the initial moment, the cumulative risk is recursively calculated according to the update function time by time, which comprehensively reflects the cumulative damage suffered by the equipment during long-term operation.

[0095] By recursively calculating, the cumulative risk of each sensor node at each time moment is obtained, and it is used together with the dynamic deviation risk index and future prediction value for graded early warning judgment.

[0096] Furthermore, in step S400, such as Figure 2 The execution of the linkage control operation matching the warning level includes, based on the dynamic deviation risk index, the future predicted value obtained in step S313, and the cumulative risk amount, constructing a multi-level threshold decision tree model to realize hierarchical warning and differentiated linkage control of the switch cabinet status, specifically including steps S411~S415: S411: Thresholds for early warning at all levels are set based on the statistical distribution of a large amount of historical failure case data and the requirements of operation and maintenance procedures.

[0097] The dynamic deviation risk index is set with three threshold levels: a first threshold, a second threshold, and a third threshold. The first threshold, the "concern level" threshold, is set to 0.15 based on the 95th percentile of the dynamic deviation risk index under normal operating conditions. The quantile is chosen because 95% of the data points under normal operating conditions are below the current value; exceeding the current value indicates a deviation from the normal range. The second threshold, the "alarm level" threshold, is set to 0.3 based on the 75th percentile of early fault data. The quantile is chosen because 75% of the dynamic deviation risk index values ​​exceed the current value in the early stages before a fault occurs, serving as a reference point for fault warning. The third threshold, the "emergency level" threshold, is set to 0.6 based on the 90th percentile of the data from 30 minutes before a fault occurs. The quantile is chosen because 90% of the dynamic deviation risk index values ​​exceed the current value within 30 minutes before a fault occurs, serving as a trigger condition for emergency response.

[0098] S412: Setting three threshold levels for cumulative risk includes setting a first threshold, a second threshold, and a third threshold for cumulative risk, wherein the third threshold is greater than the second threshold, and the second threshold is greater than the first threshold. The thresholds are calibrated through simulation experiments and cumulative risk curves of historical faults. The first threshold, i.e., the attention level threshold, is set to 3.0, the second threshold, i.e., the alarm level threshold, is set to 5.0, and the third threshold, i.e., the emergency level threshold, is set to 8.0.

[0099] The criteria for setting the thresholds include a retrospective analysis of fifty historical switchgear failures, calculation of the cumulative risk at different times before the failure, and finding that the average cumulative risk was 3.0 24 hours before the failure, 5.0 12 hours before the failure, and 8.0 1 hour before the failure.

[0100] S413: The judgment logic used for future prediction values ​​includes that if the dynamic deviation from the predicted risk index value exceeds the second-level threshold of 0.3 or the third-level threshold of 0.6 after 12 seconds of prediction, the corresponding level of warning will be triggered in advance; among them, the early warning mechanism can provide early warning before the risk actually occurs, giving maintenance personnel time to deal with it.

[0101] S414: Construct a hierarchical decision tree logic, perform judgments and branches sequentially, execute the first level judgment (emergency level judgment), traverse all sensor nodes, and directly trigger the first warning level (emergency warning) when any of the following conditions are met: Condition 1: Determine that the cumulative risk amount is greater than or equal to 8.0.

[0102] Condition 2: Determine if the dynamic deviation risk index is greater than or equal to 0.6.

[0103] Condition 3: Determine that the predicted value of the future dynamic deviation risk index is greater than or equal to 0.6.

[0104] When any node meets any of the above conditions, it enters the emergency response process and no longer executes subsequent level judgments.

[0105] If the first-level judgment is not triggered, the second-level judgment, i.e., the alarm-level judgment, is executed. All sensor nodes are traversed, and if any of the following conditions are met, the system triggers the second warning level, i.e., an alarm warning: Condition 1: Determine that the cumulative risk amount is greater than or equal to 5.0 and less than 8.0.

[0106] Condition 2: Determine that the dynamic deviation risk index is greater than or equal to 0.3 and less than 0.6.

[0107] Condition 3: Determine that the predicted value of the future dynamic deviation risk index is greater than or equal to 0.3 and less than 0.6.

[0108] When any node meets any of the above conditions, it enters the alarm-level response process.

[0109] If neither the first nor the second level of judgment is triggered, the third level of judgment, i.e., the attention level judgment, is executed. All sensor nodes are traversed, and if any of the following conditions are met, the third warning level, i.e., the attention warning, is triggered: Condition 1: Determine that the cumulative risk amount is greater than or equal to 3.0 and less than 5.0.

[0110] Condition 2: Determine that the dynamic deviation risk index is greater than or equal to 0.15 and less than 0.3.

[0111] When any node meets any of the above conditions, the system enters the attention-level response process.

[0112] If none of the above conditions are met, the system is considered to be in a normal state and no warning is triggered.

[0113] S415: Execute matching linkage control operations according to different warning levels.

[0114] When the first warning level, i.e., the emergency warning, is triggered, the following linkage control operations are executed: sending a trip command to the incoming circuit breaker of the switchgear via the programmable logic controller (PLC) to automatically cut off the power supply to the fault circuit within 100 milliseconds; sending a start command to the perfluorohexanone fire extinguishing device pre-installed in the switchgear via the PLC to start the fire extinguishing agent spraying; sending an emergency shutdown signal to the central control monitoring platform; and popping up the highest-level alarm window on the monitoring interface to display the sensor node number and location coordinates of the highest risk, so as to facilitate maintenance personnel to quickly locate the fault point.

[0115] The response time setting includes ensuring that the delay from detecting an emergency to the circuit breaker tripping is less than the thermal time constant of the electrical equipment; 100 milliseconds can meet the protection requirements of most switchgear. Perfluorohexanone fire extinguishing agent has the characteristics of rapid fire extinguishing, non-conductivity, and no residue, making it suitable for fire extinguishing of electrical equipment.

[0116] Furthermore, when the system triggers the second warning level, i.e., the alarm warning, the linkage control operations include activating the dehumidification device and cooling fan installed in the switch cabinet through the programmable logic controller to reduce the humidity and temperature inside the cabinet, actively intervening to mitigate the development of risks, and pushing alarm information to the mobile terminals of maintenance personnel. The information displays the fault type prompt, the number of the sensor node with the highest risk, the node location, and the current risk index value, and suggests arranging maintenance.

[0117] Furthermore, when the third warning level, i.e., the attention warning, is triggered, the system performs the following linked control operations: recording the warning event in the monitoring platform database, recording the warning time, warning level, triggering conditions, and risk index value to form a complete event traceability chain; generating a prompt mark on the equipment status visualization interface of the monitoring platform, marking the corresponding switch cabinet icon as a yellow attention status for maintenance personnel to focus on during daily inspections; and incorporating the current warning information into the equipment status evaluation system as a reference for the next maintenance plan, thereby realizing status-based maintenance decision optimization.

[0118] Furthermore, when the system is determined to be in a normal state, it does not perform any linkage control operations but continues to monitor.

[0119] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0120] Example 3, referring to Figure 3 The third embodiment of the present invention provides a low-voltage switchgear temperature and humidity monitoring and early warning system, including a data acquisition module, a data cleaning module, a benchmark model and risk calculation module, an uncertainty quantification and prediction module, and a cumulative assessment and decision execution module.

[0121] The data acquisition module is used to deploy a heterogeneous sensor network in the switch cabinet to simultaneously collect three types of raw data: temperature, relative humidity, and partial discharge signal intensity at each monitoring point.

[0122] The data cleaning module is used to identify and correct outliers in the original data based on the physical constraints of heat transfer. It uses the fact that the rate of change of temperature gradient between adjacent nodes does not exceed the maximum allowable value of thermal diffusivity of the material as the criterion, determines continuous outliers as sensor faults and generates alarms, and outputs the cleaned monitoring data.

[0123] The benchmark model and risk calculation module are used to construct a theoretical temperature and humidity distribution field as a dynamic benchmark by solving the heat conduction equation, which includes real-time load current and contact resistance as heat sources, and the humidity diffusion equation. Based on the deviation between the measured data and the benchmark value, and combined with the weighted amplification effect of the partial discharge signal, the module calculates a dynamic deviation risk index that reflects the degree of deviation of the equipment state.

[0124] The uncertainty quantification and prediction module is used to define the model parameters as random variables using the Bayesian inference method, establish the posterior distribution, calculate the posterior prediction distribution and confidence interval width of the dynamic deviation risk index as the uncertainty quantification result, and construct an improved long short-term memory network model with uncertainty interval as attention weight to predict the dynamic deviation risk index at future times.

[0125] The cumulative assessment and decision execution module is used to recursively update the cumulative risk amount time by time using an adaptive forgetting factor that is inversely proportional to the current risk index, to simulate the different weights of historical information and current state in the risk memory mechanism, and to establish a three-level threshold decision tree based on the dynamic deviation risk index, future prediction value and cumulative risk amount, and to perform hierarchical linkage control operations such as cutting off power, starting fire extinguishing, starting dehumidification, sending alarms and recording events according to the trigger conditions.

[0126] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0127] Example 4 is the fourth embodiment of the present invention, which differs from the previous three embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0128] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0129] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0130] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0131] It is understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for monitoring and early warning of temperature and humidity in a low-voltage switchgear, characterized in that, The specific steps include the following: Obtain the original monitoring data inside the low-voltage switchgear, identify and correct outliers in the original monitoring data according to physical constraints, and obtain the monitoring data after cleaning. A dynamic temperature and humidity benchmark model is constructed based on the physical laws of heat and mass transfer. The theoretical temperature and relative humidity values ​​are obtained based on the monitoring data after cleaning and the dynamic temperature and humidity benchmark model. The dynamic deviation risk index is calculated based on the monitoring data after cleaning, the theoretical temperature and relative humidity values. Based on the dynamic deviation risk index and the cleaned monitoring data, the uncertainty of the model parameters is quantified by Bayesian inference method to obtain the uncertainty range of the risk index, and an improved long short-term memory network model is constructed to predict the dynamic deviation risk index. Based on the dynamic deviation risk index, the cumulative risk is updated through an adaptive forgetting factor. Based on the dynamic deviation risk index, the cumulative risk, and the dynamic deviation risk index, a graded early warning strategy is formulated, and linkage control operations matching the early warning level are executed. The dynamic deviation risk index is calculated based on monitoring data after cleaning, theoretical temperature values, and theoretical relative humidity values, including: Calculate the ratio of the absolute value of the deviation between the temperature data and the theoretical temperature value at each monitoring point to the rated operating temperature to obtain the temperature normalized deviation; and calculate the ratio of the absolute value of the deviation between the relative humidity data and the theoretical relative humidity value at each monitoring point to the critical relative humidity to obtain the humidity normalized deviation. The partial discharge signal intensity is compared with a preset saturation threshold to obtain the normalized partial discharge signal intensity. The product of temperature normalization deviation, humidity normalization deviation and preset weighting coefficient, and the product of normalized partial discharge signal intensity and temperature normalization deviation and preset weighting coefficient are summed to obtain the dynamic deviation risk index. The step of updating the cumulative risk amount based on the dynamic deviation risk index and the adaptive forgetting factor includes: setting the current cumulative risk amount as the adaptive forgetting factor, multiplying it by the previous cumulative risk amount and adding it to the current dynamic deviation risk index, setting the adaptive forgetting factor as the product of the basic forgetting factor and a hyperbolic tangent function value minus the current dynamic deviation risk index, and setting the initial cumulative risk amount to zero, and recursively calculating the cumulative risk amount.

2. The method for monitoring and early warning of temperature and humidity in a low-voltage switchgear according to claim 1, characterized in that, The process of identifying and correcting outliers in the original monitoring data based on physical constraints to obtain cleaned monitoring data includes: Using the physical constraint that the rate of change of temperature gradient between adjacent locations within the same switch cabinet does not exceed the maximum allowable value of the material's thermal diffusivity, the difference in the rate of change of temperature gradient between each monitoring point and adjacent monitoring points is calculated, and monitoring point data with a difference in the rate of change of temperature gradient exceeding the maximum value are identified as outliers. Temperature data, relative humidity data, and partial discharge signal data identified as outliers are replaced with the average of the data from the previous moment and the data from the next moment. Monitoring points that are identified as outliers multiple times in a row are judged as sensor faults and fault alarm information is generated. The data after outlier identification and correction is used as the cleaned monitoring data.

3. The method for monitoring and early warning of temperature and humidity in a low-voltage switchgear according to claim 2, characterized in that, The aforementioned dynamic temperature and humidity benchmark model is constructed based on the physical laws of heat and mass transfer. Based on the monitoring data after cleaning and the dynamic temperature and humidity benchmark model, theoretical temperature and relative humidity values ​​are obtained, including... The control equations for the internal temperature and humidity fields of the switchgear are established. The control equations for the temperature field include internal heat source terms related to real-time load current and contact resistance. The switchgear is then subjected to three-dimensional geometric modeling and mesh generation, and boundary conditions for convective heat transfer with the external environment are set. The initial temperature and humidity are set to ambient temperature and ambient humidity, respectively. The governing equations are solved to obtain the theoretical temperature distribution field and the theoretical relative humidity distribution field. The theoretical temperature and theoretical relative humidity values ​​that match the location of each monitoring point are extracted from the theoretical temperature distribution field and the theoretical relative humidity field and used as the theoretical temperature and theoretical relative humidity values.

4. The method for monitoring and early warning of temperature and humidity in a low-voltage switchgear according to claim 3, characterized in that, The uncertainty range of the risk index is obtained by quantifying the uncertainty of the model parameters using Bayesian inference methods based on the dynamic deviation risk index and the cleaned monitoring data, including: The physical parameters in the control equation of the temperature field and the weighting coefficients in the calculation process of the dynamic deviation risk index are defined as random variables, and the prior distribution of the random variables is determined according to the equipment factory parameters and material properties. Collect normal operating condition data for a preset duration, establish the posterior distribution of random variables based on the normal operating condition data, calculate the posterior prediction distribution of the dynamic deviation risk index based on the posterior distribution for the current monitoring data, and calculate the preset confidence interval width of the posterior prediction distribution. Use the preset confidence interval width as the uncertainty interval of the risk index. Among them, normal operating condition data refers to historical monitoring data in which the switchgear has no fault records and the dynamic deviation risk index is lower than the preset value.

5. The method for monitoring and early warning of temperature and humidity in a low-voltage switchgear according to claim 4, characterized in that, The construction of the improved long short-term memory network model to predict the dynamic deviation risk index includes: The cleaned monitoring data of each monitoring point, the dynamic deviation risk index, the rate of change of the dynamic deviation risk index, and the uncertainty range of the risk index are constructed into a multi-dimensional input feature vector. An attention mechanism layer is set up to perform weighted processing on the input sequence with the uncertainty range of the risk index as the attention weight. The system sets up a long short-term memory network layer to receive the weighted input sequence and extract temporal features. A fully connected output layer is set up to map the temporal features to predicted values ​​for multiple future time steps. The dynamic deviation risk index of future time steps is used as the dynamic deviation risk index of the predicted future time steps.

6. The method for monitoring and early warning of temperature and humidity in a low-voltage switchgear according to claim 5, characterized in that, The step of formulating a tiered early warning strategy based on the dynamic deviation risk index, the cumulative risk level, and executing coordinated control operations matching the early warning level includes: Set the first, second, and third thresholds for the dynamic deviation risk index, and set the first, second, and third thresholds for the cumulative risk amount; When the dynamic deviation risk index of the monitoring point reaches the third threshold, or the cumulative risk reaches the third threshold, or the dynamic deviation risk index reaches the second threshold, the first warning level is triggered; when the first warning level is triggered, the linkage control operation of cutting off the power supply of the fault circuit and starting the fire extinguishing device spraying is executed. When the dynamic deviation risk index of the monitoring point reaches the second threshold but not the third threshold, or when the cumulative risk reaches the second threshold but not the third threshold, or when the dynamic deviation risk index reaches the first threshold, the second warning level is triggered. When the second warning level is triggered, the linkage control operation of starting the dehumidification device and cooling fan and sending alarm information to the operation and maintenance terminal is executed. When the dynamic deviation risk index of a monitoring point reaches the first threshold but not the second threshold, or when the cumulative risk reaches the first threshold but not the second threshold, the third warning level is triggered. When the third warning level is triggered, the linkage control operation of recording the warning event and generating a prompt mark on the monitoring platform is executed.

7. A low-voltage switchgear temperature and humidity monitoring and early warning system, employing the low-voltage switchgear temperature and humidity monitoring and early warning method according to any one of claims 1 to 6, characterized in that, It includes a data acquisition module, a data cleaning module, a benchmark model and risk calculation module, an uncertainty quantification and prediction module, and a cumulative assessment and decision execution module; The data acquisition module is used to acquire raw monitoring data inside the low-voltage switchgear. The data cleaning module is used to identify and correct outliers in the original monitoring data according to physical constraints, so as to obtain cleaned monitoring data. The benchmark model and risk calculation module are used to construct a dynamic temperature and humidity benchmark model based on the physical laws of heat and mass transfer. Based on the monitoring data after cleaning and the dynamic temperature and humidity benchmark model, theoretical temperature and relative humidity values ​​are obtained. The dynamic deviation risk index is calculated based on the monitoring data after cleaning, the theoretical temperature value, and the theoretical relative humidity value. Specifically, the ratio of the absolute value of the deviation between the temperature data and the theoretical temperature value at each monitoring point to the rated operating temperature is calculated to obtain the temperature normalized deviation. The ratio of the absolute value of the deviation between the relative humidity data and the theoretical relative humidity value at each monitoring point to the critical relative humidity is also calculated to obtain the humidity normalized deviation. The partial discharge signal intensity is compared with a preset saturation threshold to obtain the normalized partial discharge signal intensity. The product of the temperature normalized deviation, the humidity normalized deviation, and the preset weighting coefficient, as well as the product of the normalized partial discharge signal intensity and the temperature normalized deviation, are summed to obtain the dynamic deviation risk index. The uncertainty quantification and prediction module is used to quantify the uncertainty of model parameters based on the dynamic deviation risk index and the cleaned monitoring data using Bayesian inference methods, obtain the uncertainty range of the risk index, and construct an improved long short-term memory network model to predict the dynamic deviation risk index. The cumulative assessment and decision execution module is used to update the cumulative risk amount based on the dynamic deviation risk index using an adaptive forgetting factor. Specifically, the current cumulative risk amount is set as the adaptive forgetting factor, and the product of the current cumulative risk amount and the previous cumulative risk amount is added to the current dynamic deviation risk index. The adaptive forgetting factor is set as the product of the basic forgetting factor and the hyperbolic tangent function value minus the current dynamic deviation risk index. The initial cumulative risk amount is set to zero. The cumulative risk amount is calculated recursively. Based on the dynamic deviation risk index, the cumulative risk amount, and the cumulative risk amount, a graded early warning strategy is formulated, and linkage control operations matching the early warning level are executed.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the low-voltage switchgear temperature and humidity monitoring and early warning method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the low-voltage switchgear temperature and humidity monitoring and early warning method according to any one of claims 1 to 6.

Images