Switch cabinet solid insulating material performance detection method based on temperature detection

Abstract

The invention provides a switch cabinet solid insulating material performance detection method based on temperature detection, which comprises the following steps: constructing a gridding detection matrix on the surface of an insulating material, and collecting temperature data of each monitoring area through a passive radio frequency sensor to establish an initial temperature field data set; adopting a deep learning algorithm to calculate a temperature gradient value and thermal stress distribution, and establishing a temperature-stress coupling model; calculating a dielectric loss factor and a local ionization degree based on the model, constructing a performance degradation index and establishing a degradation prediction model; combining a recurrent neural network and an attention mechanism to predict a performance evolution trajectory, and setting a multi-stage dynamic early warning threshold value; when early warning is triggered, a sampling strategy is adaptively adjusted, model parameters are optimized through real-time data, and an intelligent diagnosis report containing state evaluation, life prediction and maintenance suggestions is generated. According to the invention, accurate prediction and intelligent diagnosis of performance degradation of the solid insulating material can be realized, and the real-time performance and reliability of monitoring the insulation state of the switch cabinet are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment status monitoring, and more specifically, to a method for detecting the performance of solid insulation materials of a switch cabinet based on temperature detection. Background Art

[0002] Switchgear plays a crucial role in controlling and protecting electrical equipment in power systems. The performance of its solid insulation materials is directly related to the safe operation and service life of the switchgear. Traditional methods for testing the performance of switchgear solid insulation materials rely primarily on periodic insulation resistance testing, partial discharge testing, and visual inspections. While these methods can provide a certain degree of insight into the macroscopic state of the insulation material, they have numerous limitations. For example, insulation resistance testing only reflects the overall insulation level of the insulation material and struggles to capture subtle changes in localized performance degradation. While partial discharge testing can detect internal discharges within the insulation material, it is not sensitive enough to detect early-stage performance degradation. Visual inspections rely on manual experience, are highly subjective, and difficult to quantify. Furthermore, most of these traditional methods are offline and cannot monitor the dynamic performance changes of the insulation material under actual operating conditions in real time.

[0003] With the continuous improvement of the power system's requirements for equipment reliability and operating efficiency, as well as the rapid development of intelligent monitoring technology, higher requirements have been placed on the performance testing of solid insulation materials in switchgear. In recent years, insulation material performance monitoring technology based on temperature detection has gradually attracted attention. Temperature is one of the key factors affecting the performance of solid insulation materials. During operation, insulation materials will deteriorate due to thermal stress, electrical stress, and other effects. By monitoring the temperature distribution and changes on the surface of the insulation material, its internal thermal field and stress state can be indirectly reflected, thus providing an important basis for performance evaluation. However, existing temperature detection-based methods mostly use simple temperature sensor arrangements and single temperature analysis models, which cannot accurately capture the complex temperature-stress coupling relationship inside the insulation material, nor can they accurately predict and dynamically warn of performance degradation.

[0004] In the process of implementing the embodiments of the present invention, there are at least the following problems or defects in the existing technology: First, traditional detection methods are unable to monitor the performance changes of solid insulating materials in real time and comprehensively, and it is difficult to meet the power system's demand for real-time monitoring of equipment operating status; second, existing temperature detection-based technologies fail to fully consider the coupling effect between temperature and stress, resulting in inaccurate assessment of insulation material performance degradation; third, there is a lack of effective performance degradation prediction models and dynamic early warning mechanisms, making it impossible to predict the performance evolution trend of insulation materials in advance and take maintenance measures in a timely manner; finally, the existing technology is relatively fixed in the layout of monitoring points and data sampling strategies, and cannot be adaptively adjusted according to the actual operating status of the insulation material, resulting in low monitoring efficiency and accuracy. Summary of the Invention

[0005] The present invention provides a method for detecting the performance of solid insulation materials of a switch cabinet based on temperature detection, comprising:

[0006] Constructing a grid detection matrix on the surface of the solid insulation material of the switch cabinet, the grid detection matrix includes multiple monitoring areas;

[0007] The real-time temperature data of each monitoring area is collected through a passive radio frequency identification temperature sensor to establish an initial temperature field data set;

[0008] Based on the calibrated initial temperature field data set, a deep learning algorithm is used to calculate the temperature gradient value and thermal stress distribution of each monitoring area of ​​the solid insulation material;

[0009] Combining the temperature gradient value with the thermal stress distribution, a temperature-stress coupling model is established;

[0010] The dielectric loss factor and local ionization degree of each monitoring area of ​​the solid insulation material are calculated based on the temperature-stress coupling model, and a performance degradation index that includes the microstructural evolution characteristics is constructed.

[0011] Based on the temporal variation of the performance degradation index, a performance degradation prediction model for solid insulation materials is established;

[0012] Using a performance degradation prediction model, combined with a recurrent neural network and an attention mechanism, the performance evolution trajectory of solid insulating materials is predicted;

[0013] Establish multi-level dynamic warning thresholds based on performance evolution trajectories;

[0014] When an early warning trigger is detected, the data sampling strategy of the monitoring points in the grid detection matrix is ​​adaptively adjusted based on the change trend of the calibrated initial temperature field data set;

[0015] The real-time monitoring data obtained after adjustment is compared with the performance degradation prediction model, and the parameters of the temperature-stress coupling model are continuously optimized to generate an intelligent diagnostic report including performance degradation status assessment, remaining life prediction and maintenance recommendations.

[0016] Furthermore, based on the calibrated initial temperature field dataset, a deep learning algorithm is used to calculate the temperature gradient values ​​and thermal stress distribution of each monitoring area of ​​the solid insulation material. The temperature gradient values ​​are combined with the thermal stress distribution to establish a temperature-stress coupling model, including:

[0017] The initial temperature field dataset is input into the convolutional neural network to extract the spatial temperature distribution characteristics of each monitoring area;

[0018] The temperature gradient value of each monitoring area is output through the gradient calculation layer. The temperature gradient value is expressed as:

[0019]

[0020] Where T is the temperature of the monitoring area, x and y are the coordinates of the grid detection matrix, is the temperature gradient value of the monitoring area, indicating the degree of temperature change in the x and y directions.

[0021] Solve the thermal stress distribution equation based on the finite element analysis method:

[0022] σ=E·α·ΔT

[0023] Where σ is the thermal stress value, E is the elastic modulus of the solid insulation material, which characterizes the material's ability to resist elastic deformation; α is the thermal expansion coefficient, which describes the material's dimensional change due to temperature changes; and ΔT is the temperature change in the monitoring area.

[0024] The temperature gradient value and thermal stress distribution are input into the coupling layer to generate a temperature-stress coupling model.

[0025] Furthermore, the dielectric loss factor and local ionization degree of each monitored area of ​​the solid insulation material are calculated based on the temperature-stress coupling model, and a performance degradation index that includes the microstructural evolution characteristics is constructed. Based on the temporal changes of the performance degradation index, a performance degradation prediction model for solid insulation materials is established, including:

[0026] The dielectric loss factor tanδ and local ionization degree η of each monitoring area are output through the temperature-stress coupling model;

[0027] The calculation formula for constructing the performance degradation index is:

[0028]

[0029] Among them, α, β, γ, and λ are weight coefficients used to adjust the importance of temperature gradient value, thermal stress value, dielectric loss factor, and local ionization degree in the calculation of performance degradation index; PI is the performance degradation index, which comprehensively reflects the performance degradation degree of solid insulating materials.

[0030] Collect the time series of performance degradation index changes in historical data, input it into the long short-term memory network, and train it to obtain a performance degradation prediction model;

[0031] The performance degradation prediction model outputs the predicted value of the performance degradation index in the future time step.

[0032] Furthermore, the performance degradation prediction model is used in conjunction with a recurrent neural network and an attention mechanism to predict the performance evolution trajectory of solid insulation materials. Based on the performance evolution trajectory, a multi-level dynamic warning threshold is established, including:

[0033] Input the performance degradation index prediction value into the recurrent neural network to generate the performance evolution trajectory function f(t);

[0034] The weight distribution of different time steps is calculated through the attention mechanism to optimize the performance evolution trajectory function f(t);

[0035] Based on the first-order derivative of the performance evolution trajectory function f(t) Determine the rate of performance degradation, Indicates the rate of change of performance degradation index over time;

[0036] According to the performance degradation rate and the preset probability distribution threshold, multi-level dynamic warning thresholds are divided, including warning levels L1, L2, and L3, where L3 corresponds to the highest degradation rate.

[0037] Furthermore, when an early warning trigger is detected, the data sampling strategy of the monitoring points in the grid detection matrix is ​​adaptively adjusted based on the change trend of the calibrated initial temperature field data set, including:

[0038] When the warning level L1 is triggered, the data sampling frequency of the low-risk monitoring area is reduced to the first frequency value;

[0039] When the warning level L2 is triggered, the data sampling frequency of the high-risk monitoring area is increased to the second frequency value, and the redundant sensors in the adjacent monitoring areas are activated;

[0040] When the warning level L3 is triggered, the highest sampling frequency of the full grid detection matrix is ​​started, and the power supply current of the passive RFID temperature sensor is increased synchronously.

[0041] Furthermore, the real-time monitoring data obtained after adjustment is compared with the performance degradation prediction model, and the parameters of the temperature-stress coupling model are continuously optimized to generate an intelligent diagnostic report containing performance degradation status assessment, remaining life prediction and maintenance recommendations, including:

[0042] Calculate the residual value between the real-time monitoring data and the output of the performance degradation prediction model;

[0043] When the residual value exceeds the preset error tolerance, the particle swarm optimization algorithm is used to update the weight parameters of the temperature-stress coupling model;

[0044] Based on the updated temperature-stress coupling model, the performance degradation status assessment results are output through the fuzzy comprehensive evaluation method;

[0045] The remaining life prediction value is calculated by combining the three-parameter Weibull distribution model:

[0046]

[0047] Among them, θ is the characteristic life parameter, which reflects the characteristic value of the average life of the material; κ is the shape parameter, which describes the shape of the material life distribution; R(t) is the predicted value of the remaining life at time t.

[0048] A maintenance recommendation strategy table is generated based on the remaining life prediction value, and the maintenance recommendation strategy table includes a maintenance time window and maintenance level.

[0049] Furthermore, after generating the intelligent diagnosis report, it also includes:

[0050] Associate and store intelligent diagnostic reports with switchgear operation logs;

[0051] When multiple switchgears with the same solid insulation material type are detected to trigger the same warning level, a batch defect analysis report is generated;

[0052] Based on the batch defect analysis report, the initial parameter settings of the temperature-stress coupling model are optimized.

[0053] Furthermore, the specific steps of constructing a grid detection matrix include:

[0054] The solid insulation surface is divided into m×n rectangular monitoring areas at equal intervals;

[0055] Deploy passive RFID temperature sensors at the center and vertices of each rectangular monitoring area;

[0056] The temperature data of each monitoring area is polled by radio frequency readers in a time-division multiplexing manner.

[0057] Furthermore, the deployment of passive RFID temperature sensors must meet the following requirements:

[0058] The accuracy of sensors deployed at the center is ±0.1°C, and the accuracy of sensors deployed at the vertex is ±0.5°C;

[0059] The interval between the operating frequency bands of sensors in adjacent monitoring areas shall not be less than 10 MHz.

[0060] Furthermore, the update mechanism of the multi-level dynamic warning threshold includes:

[0061] Extract the maximum performance degradation rate V from the historical warning data every month max ;

[0062] If three consecutive V max If the upper threshold of the current warning level L2 is exceeded, the warning thresholds of each level will be increased in proportion to ρ:

[0063]

[0064] Among them, V max is the maximum performance degradation rate in the historical warning data, V base is the baseline degradation rate, which is used to measure the baseline value of the performance degradation rate; ρ is the ratio of the warning threshold to increase, L i is the i-th level warning threshold before adjustment, L′ i is the adjusted level i warning threshold.

[0065] The above embodiments of the present invention have at least the following beneficial effects:

[0066] 1. Through the coordinated deployment of a grid detection matrix and passive radio frequency identification temperature sensors, high-precision distributed monitoring of the surface temperature field of solid insulating materials is achieved. This solves the problem that traditional point temperature measurement methods cannot fully reflect the thermal distribution characteristics of insulating materials, and provides a reliable data foundation for subsequent performance analysis.

[0067] 2. The temperature-stress coupling model is combined with a multi-parameter fusion algorithm of dielectric loss factor and local ionization degree, which breaks through the limitation of evaluating insulation performance with a single temperature indicator. It can accurately characterize the degradation process of the material microstructure and improve the comprehensiveness and accuracy of insulation status diagnosis.

[0068] 3. An adaptive monitoring strategy based on recursive neural networks and dynamic warning thresholds enables full-cycle performance evaluation from real-time status monitoring to long-term life prediction, solving the problem of delayed response in traditional periodic maintenance models and providing intelligent decision-making support for switchgear insulation maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which:

[0070] Figure 1 A schematic flow chart of a method for detecting the performance of solid insulation materials of a switch cabinet based on temperature detection provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0071] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided solely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0072] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0073] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0074] Reference below Figure 1 , Figure 1 This is a flow chart of a method for detecting the performance of solid insulation materials of a switch cabinet based on temperature detection according to an embodiment of the present invention. Figure 1 As shown, a method for detecting the performance of solid insulation materials of a switch cabinet based on temperature detection includes:

[0075] S1. Construct a grid detection matrix on the surface of the solid insulation material of the switch cabinet, and the grid detection matrix includes multiple monitoring areas;

[0076] S2. Collect real-time temperature data of each monitoring area through a passive radio frequency identification temperature sensor to establish an initial temperature field data set;

[0077] S3. Based on the calibrated initial temperature field dataset, a deep learning algorithm is used to calculate the temperature gradient value and thermal stress distribution of each monitoring area of ​​the solid insulation material;

[0078] S4. Combining the temperature gradient value with the thermal stress distribution to establish a temperature-stress coupling model;

[0079] S5. Calculate the dielectric loss factor and local ionization degree of each monitoring area of ​​the solid insulation material based on the temperature-stress coupling model, and construct a performance degradation index that includes the microstructural evolution characteristics;

[0080] S6. Establish a performance degradation prediction model for solid insulation materials based on the temporal changes in the performance degradation index;

[0081] S7. Use the performance degradation prediction model, combined with recurrent neural networks and attention mechanisms, to predict the performance evolution trajectory of solid insulation materials;

[0082] S8. Establish multi-level dynamic warning thresholds based on performance evolution trajectories;

[0083] S9, when an early warning trigger is detected, adaptively adjust the data sampling strategy of the monitoring points in the grid detection matrix based on the change trend of the calibrated initial temperature field data set;

[0084] S10. Compare the real-time monitoring data obtained after adjustment with the performance degradation prediction model, continuously optimize the parameters of the temperature-stress coupling model, and generate an intelligent diagnosis report including performance degradation status assessment, remaining life prediction and maintenance recommendations.

[0085] It should be noted that the grid detection matrix is ​​a detection layout method that divides the surface of a solid insulating material into multiple small areas. This method can achieve refined monitoring of the temperature distribution on the surface of the insulating material. A passive radio frequency identification temperature sensor is a sensor that does not require an external power supply and transmits temperature data via radio frequency signals. It can operate stably in complex electromagnetic environments. The initial temperature field data set refers to the set of temperature data collected by the sensor for each monitoring area. This data reflects the temperature distribution of the insulating material in its initial state. The deep learning algorithm is a machine learning method based on neural networks that can automatically learn features and patterns from large amounts of data. It is used to calculate the temperature gradient value and thermal stress distribution of each monitoring area of ​​the solid insulating material.

[0086] Specifically, the grid-based detection matrix is ​​constructed by dividing the surface of the solid insulating material into multiple rectangular monitoring areas at regular intervals. For example, the area can be divided into 10 cm x 10 cm intervals, forming an m x n grid layout. Passive RFID temperature sensors are deployed at the center and vertices of each monitoring area. The sensor accuracy at the center is ±0.1°C, and at the vertices is ±0.5°C, meeting different accuracy requirements. The initial temperature field dataset is obtained by polling the temperature data of each monitoring area using a radio frequency reader in a time-division multiplexing manner. This data includes the temperature values ​​of each monitoring area at different time points. The convolutional neural network in the deep learning algorithm is used to extract the spatial temperature distribution characteristics of each monitoring area and can identify the temperature variation patterns in different directions. The temperature gradient value represents the degree of temperature variation in space. It is obtained through the gradient calculation layer and reflects the temperature variation in the x and y directions. The thermal stress distribution is obtained by solving the thermal stress distribution equation using finite element analysis. The elastic modulus E represents the material's resistance to elastic deformation, the thermal expansion coefficient α describes the material's dimensional change due to temperature changes, and the temperature change ΔT is the difference between the monitoring area temperature and the initial temperature.

[0087] Preferably, when constructing a gridded detection matrix, the grid spacing and layout can be flexibly adjusted according to the size and shape of the solid insulating material. For example, for insulating materials with larger areas, the grid spacing can be appropriately increased to improve monitoring efficiency. The deployment of passive RFID temperature sensors needs to consider mutual interference between sensors. The operating frequency band interval of sensors in adjacent monitoring areas should be no less than 10 MHz to avoid signal conflicts. The convolutional neural network in the deep learning algorithm can extract the features of temperature data through multi-layer convolution and pooling operations. The input parameters include the matrix form of the temperature data and the corresponding monitoring area coordinates. The temperature gradient value can be calculated by performing a differential or differential operation on the temperature data. Specifically, the temperature gradient is approximated by calculating the difference between the temperatures of adjacent monitoring areas and dividing it by the spacing. The solution to the thermal stress distribution equation needs to be set according to the specific physical parameters of the solid insulating material. For example, for common epoxy resin insulating materials, its elastic modulus E and thermal expansion coefficient α can be obtained through experimental measurement, and the temperature change ΔT is calculated by the difference between the real-time monitoring data and the initial temperature data.

[0088] In some embodiments, based on the calibrated initial temperature field data set, a deep learning algorithm is used to calculate the temperature gradient value and thermal stress distribution of each monitoring area of ​​the solid insulation material; the temperature gradient value and thermal stress distribution are combined to establish a temperature-stress coupling model, including:

[0089] The initial temperature field dataset is input into the convolutional neural network to extract the spatial temperature distribution characteristics of each monitoring area;

[0090] The temperature gradient value of each monitoring area is output through the gradient calculation layer. The temperature gradient value is expressed as:

[0091]

[0092] Where T is the temperature of the monitoring area, x and y are the coordinates of the grid detection matrix, is the temperature gradient value of the monitoring area, indicating the degree of temperature change in the x and y directions.

[0093] Solve the thermal stress distribution equation based on the finite element analysis method:

[0094] σ=E·α·ΔT

[0095] Where σ is the thermal stress value, E is the elastic modulus of the solid insulation material, which characterizes the material's ability to resist elastic deformation; α is the thermal expansion coefficient, which describes the material's dimensional change due to temperature changes; and ΔT is the temperature change in the monitoring area.

[0096] The temperature gradient value and thermal stress distribution are input into the coupling layer to generate a temperature-stress coupling model.

[0097] It should be noted that using a deep learning algorithm to calculate the temperature gradient and thermal stress distribution of each monitoring area of ​​the solid insulation material based on a calibrated initial temperature field dataset is a key step in achieving performance testing. The initial temperature field dataset refers to the temperature data collected from each monitoring area using passive radio frequency identification temperature sensors. After calibration, this data accurately reflects the temperature distribution on the insulation material surface. Deep learning algorithms are powerful data processing tools that can extract useful information from complex temperature data. The temperature gradient value indicates the degree of temperature variation in space and reflects the direction and intensity of heat transfer within the insulation material. The thermal stress distribution is the stress distribution within the material caused by temperature changes and is closely related to the material's physical properties. By combining the temperature gradient value and the thermal stress distribution, a temperature-stress coupling model can be established. This model can more comprehensively reflect the stress and thermal state of the insulation material during operation, providing an important basis for subsequent performance evaluation.

[0098] Specifically, the initial temperature field dataset is collected using a gridded detection matrix constructed on the surface of the solid insulation material. Temperature data for each monitoring area is transmitted in real time via passive radio frequency identification temperature sensors. A convolutional neural network within a deep learning algorithm is used to process this temperature data. It extracts spatial features from the temperature data through convolutional and pooling layers, which reflect the temperature variations across different regions. The temperature gradient is obtained through the gradient calculation layer. It represents the degree of temperature variation in the x and y directions. The calculation method uses partial derivatives to measure the temperature change with spatial coordinates. The thermal stress distribution is calculated using finite element analysis, which obtains results by solving the thermal stress distribution equation. In this equation, the elastic modulus E is a physical parameter of the material, reflecting its ability to resist elastic deformation; the thermal expansion coefficient α describes the dimensional change of the material due to temperature changes; and the temperature change ΔT is the difference between the current and initial temperatures of the monitoring area. These parameters together determine the magnitude and distribution of thermal stress.

[0099] Preferably, when constructing a temperature-stress coupling model, the model construction steps can be further refined. First, the initial temperature field dataset is input into a convolutional neural network. The network's input layer receives a temperature data matrix, where each element corresponds to the temperature value of a monitoring area. The convolution layer extracts local temperature features through a sliding convolution kernel operation, while the pooling layer is used to reduce the feature dimension while retaining important information. In the gradient calculation layer, the temperature gradient value is obtained by calculating the partial derivatives of the temperature data in the x and y directions. Specifically, the partial derivatives are approximated using the finite difference method, that is, the gradient value is obtained by dividing the temperature difference between adjacent monitoring areas by the spatial separation. To calculate the thermal stress distribution, the finite element analysis method divides the solid insulation material into multiple small units. Based on the temperature change ΔT of each unit, combined with the material's elastic modulus E and thermal expansion coefficient α, the thermal stress distribution equation is numerically solved to obtain the thermal stress value of each unit. Finally, the temperature gradient value and thermal stress distribution are input into the coupling layer, and a weighted summation or other fusion method is used to generate a temperature-stress coupling model that can comprehensively reflect the effects of temperature and stress on the performance of the insulation material.

[0100] In some embodiments, the dielectric loss factor and local ionization degree of each monitored area of ​​the solid insulating material are calculated based on the temperature-stress coupling model, and a performance degradation index including microstructural evolution characteristics is constructed. Based on the temporal variation of the performance degradation index, a performance degradation prediction model for the solid insulating material is established, including:

[0101] The dielectric loss factor tanδ and local ionization degree η of each monitoring area are output through the temperature-stress coupling model;

[0102] The calculation formula for constructing the performance degradation index is:

[0103]

[0104] Among them, α, β, γ, and λ are weight coefficients used to adjust the importance of temperature gradient value, thermal stress value, dielectric loss factor, and local ionization degree in the calculation of performance degradation index; PI is the performance degradation index, which comprehensively reflects the performance degradation degree of solid insulating materials.

[0105] Collect the time series of performance degradation index changes in historical data, input it into the long short-term memory network, and train it to obtain a performance degradation prediction model;

[0106] The performance degradation prediction model outputs the predicted value of the performance degradation index in the future time step.

[0107] It should be noted that outputting the dielectric loss factor and local ionization level of each monitoring area through the temperature-stress coupling model and constructing a performance degradation index that incorporates microstructural evolution characteristics are key steps in achieving insulation material performance evaluation. The dielectric loss factor reflects the energy loss of the insulating material under the action of an electric field, while the local ionization level reveals the degree of change in the insulating material's internal microstructure. The performance degradation index is a comprehensive indicator, obtained by weighted summation of the temperature gradient value, thermal stress value, dielectric loss factor, and local ionization level, which can comprehensively reflect the performance degradation of the insulating material. The performance degradation prediction model uses a long short-term memory network to learn and predict the time-series changes of the performance degradation index, providing a basis for early warning and maintenance.

[0108] Specifically, the dielectric loss factor is an indicator that measures the energy loss of an insulating material under the action of an electric field, and is usually obtained by measuring the loss tangent of the insulating material in an AC electric field. The degree of local ionization reflects the changes in the internal microstructure of the insulating material, such as the local discharge phenomenon inside the material. In the calculation formula of the performance degradation index, α, β, γ, and λ are weight coefficients used to adjust the importance of the temperature gradient value, thermal stress value, dielectric loss factor, and local ionization degree in the calculation of the performance degradation index. These weight coefficients can be adjusted according to actual needs and experimental data to ensure that the performance degradation index can accurately reflect the performance degradation of the insulating material. The long short-term memory network is a neural network that can process time series data. By learning the time series change sequence of the performance degradation index, the performance degradation index of the future time step can be predicted.

[0109] Preferably, when constructing the performance degradation index, the weight coefficient can be adjusted according to the actual application scenario and importance of the insulating material. For example, if the temperature gradient has a greater impact on the performance of the insulating material, the weight of α can be appropriately increased; if the local ionization phenomenon has a more significant impact on the performance of the insulating material, the weight of λ can be increased. When calculating the dielectric loss factor, the loss tangent value can be obtained by applying an AC voltage to the surface of the insulating material and measuring the current phase lag angle. The degree of local ionization can be quantified by monitoring the local discharge phenomenon inside the insulating material, for example, by measuring the amplitude and frequency of the discharge pulse to evaluate the degree of ionization. The process of constructing the performance degradation prediction model includes: first, collecting the time series change sequence of the performance degradation index in the historical data as training data; then, inputting these data into the long short-term memory network for training, optimizing the weight parameters of the network so that it can accurately predict the changing trend of the future performance degradation index; finally, using the trained model to predict the real-time monitoring data, and output the predicted value of the performance degradation index for the future time step.

[0110] In some embodiments, a performance degradation prediction model is used in combination with a recurrent neural network and an attention mechanism to predict the performance evolution trajectory of solid insulating materials. Based on the performance evolution trajectory, a multi-level dynamic warning threshold is established, including:

[0111] Input the performance degradation index prediction value into the recurrent neural network to generate the performance evolution trajectory function f(t);

[0112] The weight distribution of different time steps is calculated through the attention mechanism to optimize the performance evolution trajectory function f(t);

[0113] Based on the first-order derivative of the performance evolution trajectory function f(t) Determine the rate of performance degradation, Indicates the rate of change of performance degradation index over time;

[0114] According to the performance degradation rate and the preset probability distribution threshold, multi-level dynamic warning thresholds are divided, including warning levels L1, L2, and L3, where L3 corresponds to the highest degradation rate.

[0115] It is important to note that using a performance degradation prediction model combined with a recurrent neural network and an attention mechanism to predict the performance evolution trajectory of solid insulation materials and establish multi-level dynamic warning thresholds based on this performance evolution trajectory is a key step in achieving real-time monitoring and early warning of insulation material performance. The performance degradation prediction model is based on the temporal changes in the performance degradation index. By learning from historical data using a long short-term memory (LSTM) network, it can predict future performance degradation trends. Recurrent neural networks (RNNs) are a neural network structure suitable for processing sequential data and can capture dynamic changes in time series. The attention mechanism enhances the model's focus on important time steps, improving prediction accuracy. The multi-level dynamic warning thresholds are set according to different intervals of performance degradation rates and are used to provide graded early warnings of the insulation material's performance status, enabling timely implementation of appropriate maintenance measures.

[0116] Specifically, the performance degradation prediction model is constructed using a long short-term memory (LSTM) network, which can process time series data of performance degradation indices and capture long-term dependencies. A recurrent neural network (RNN) is used to generate a performance evolution trajectory function. By modeling the time series of predicted performance degradation indices, it reflects the performance trends of insulation materials over time. The attention mechanism assigns weights to the performance degradation indices at different time steps, allowing the model to focus on time points that have a greater impact on performance evolution. The first-order derivative of the performance evolution trajectory function represents the performance degradation rate, that is, the rate of change of the performance degradation index over time. Multi-level dynamic warning thresholds are divided according to the performance degradation rate. For example, warning levels can be divided into L1, L2, and L3. L3 corresponds to the highest degradation rate, indicating that the insulation material performance is deteriorating most severely and requires immediate action.

[0117] Preferably, when constructing a performance degradation prediction model, the time series data of the performance degradation index can be input into a long short-term memory (LSTM) network, and a model capable of predicting future performance degradation indices can be obtained through training. The recursive neural network (RNN) can further optimize the generation process of the performance evolution trajectory function, for example by adjusting the number of network layers and neurons to improve the model's fitting ability. The specific implementation of the attention mechanism can be achieved by calculating the weight distribution for each time step, for example, by assigning weights based on the proximity of the time step to the current time or the magnitude of the change in the performance degradation index. The first-order derivative of the performance evolution trajectory function can be calculated using a numerical difference method, that is, the degradation rate is approximated by dividing the difference in the performance degradation index between adjacent time steps by the time interval. The division of multi-level dynamic warning thresholds can be set according to actual application requirements and historical data statistics. For example, by analyzing the distribution of performance degradation rates in historical data, the rate range corresponding to different warning levels can be determined.

[0118] In some embodiments, when a warning trigger is detected, the data sampling strategy of the monitoring points in the grid detection matrix is ​​adaptively adjusted based on the change trend of the calibrated initial temperature field data set, including:

[0119] When the warning level L1 is triggered, the data sampling frequency of the low-risk monitoring area is reduced to the first frequency value;

[0120] When the warning level L2 is triggered, the data sampling frequency of the high-risk monitoring area is increased to the second frequency value, and the redundant sensors in the adjacent monitoring areas are activated;

[0121] When the warning level L3 is triggered, the highest sampling frequency of the full grid detection matrix is ​​started, and the power supply current of the passive RFID temperature sensor is increased synchronously.

[0122] It should be noted that when a warning trigger is detected, adaptively adjusting the data sampling strategy at monitoring points in the gridded detection matrix based on the changing trends of the calibrated initial temperature field dataset is a key step in achieving accurate monitoring and early warning. The changing trends of the initial temperature field dataset reflect the dynamic changes in the temperature distribution of the insulation material during operation. Analysis of these changes can be used to determine the degree of performance degradation of the insulation material. Adaptive adjustment of the data sampling strategy means dynamically changing the sampling frequency of monitoring points according to different warning levels to optimize the allocation of monitoring resources and ensure more accurate data is obtained in key areas and at critical moments.

[0123] Specifically, the initial temperature field data set is a set of temperature data collected by a passive radio frequency identification temperature sensor. After calibration, these data can accurately reflect the temperature distribution on the surface of the insulating material. The change trend refers to the change pattern of temperature data over time, such as the rate of temperature increase or decrease. Adaptive adjustment refers to the automatic adjustment of the sampling frequency of the monitoring point according to the different warning levels. For example, when a lower warning level is triggered, the sampling frequency of the low-risk area can be appropriately reduced; when a higher warning level is triggered, the sampling frequency of the high-risk area needs to be increased, and redundant sensors need to be activated to enhance monitoring capabilities. The warning level is divided according to the performance degradation rate. For example, L1, L2, and L3 correspond to different degradation rate intervals, where L3 represents the most serious degradation situation.

[0124] Preferably, when adaptively adjusting the data sampling strategy, different measures can be taken based on the specific circumstances of the warning level. For example, when warning level L1 is triggered, the data sampling frequency in the low-risk monitoring area can be reduced to a first frequency value, such as once every 10 minutes, to reduce the waste of monitoring resources. When warning level L2 is triggered, the data sampling frequency in the high-risk monitoring area is increased to a second frequency value, such as once every minute, and redundant sensors in adjacent monitoring areas are activated to enhance monitoring capabilities in that area. When warning level L3 is triggered, the highest sampling frequency of the fully gridded detection matrix is ​​activated, such as once every 30 seconds, and the power supply current of the passive RFID temperature sensor is simultaneously increased to ensure the sensor can operate normally and obtain high-precision data. This adaptive adjustment strategy can dynamically optimize the allocation of monitoring resources based on the actual operating status of the insulation material, improving monitoring efficiency and warning accuracy.

[0125] In some embodiments, the real-time monitoring data acquired after adjustment is compared with the performance degradation prediction model, and the parameters of the temperature-stress coupling model are continuously optimized to generate an intelligent diagnostic report containing performance degradation status assessment, remaining life prediction, and maintenance recommendations, including:

[0126] Calculate the residual value between the real-time monitoring data and the output of the performance degradation prediction model;

[0127] When the residual value exceeds the preset error tolerance, the particle swarm optimization algorithm is used to update the weight parameters of the temperature-stress coupling model;

[0128] Based on the updated temperature-stress coupling model, the performance degradation status assessment results are output through the fuzzy comprehensive evaluation method;

[0129] The remaining life prediction value is calculated by combining the three-parameter Weibull distribution model:

[0130]

[0131] Among them, θ is the characteristic life parameter, which reflects the characteristic value of the average life of the material; κ is the shape parameter, which describes the shape of the material life distribution; R(t) is the predicted value of the remaining life at time t.

[0132] A maintenance recommendation strategy table is generated based on the remaining life prediction value, and the maintenance recommendation strategy table includes a maintenance time window and maintenance level.

[0133] It should be noted that comparing the real-time monitoring data obtained after adjustment with the performance degradation prediction model, continuously optimizing the parameters of the temperature-stress coupling model, and generating an intelligent diagnostic report that includes performance degradation status assessment, remaining life prediction, and maintenance recommendations are key steps in achieving accurate performance assessment and maintenance of insulation materials. Real-time monitoring data refers to the latest temperature data obtained after adaptively adjusting the sampling strategy. These data can reflect the current actual operating status of the insulation material. The performance degradation prediction model is a model trained based on historical data and is used to predict future performance degradation trends. By comparing the real-time monitoring data with the model prediction values, the deviation between the model and the actual status can be found, and the model parameters can be optimized accordingly. Performance degradation status assessment is a quantitative analysis of the current performance status of the insulation material, and the remaining life prediction is an estimate of the future service life of the insulation material based on the current status and degradation trend. Maintenance recommendations are targeted maintenance measures proposed based on the evaluation and prediction results.

[0134] Specifically, real-time monitoring data is collected by passive radio frequency identification temperature sensors using an adjusted sampling strategy. This data reflects the temperature changes of the insulation material at different warning levels. The performance degradation prediction model is constructed based on a long short-term memory (LSTM) network, and its output is a predicted performance degradation index for future time steps. The comparison process involves calculating the residual between the real-time monitoring data and the model prediction, which reflects the accuracy of the model prediction. When the residual exceeds the preset error tolerance, the model needs to be updated. The particle swarm optimization algorithm, an optimization algorithm based on swarm intelligence, is used to update the weight parameters of the temperature-stress coupling model. The fuzzy comprehensive evaluation method, a decision-making method based on fuzzy mathematics, is used to output performance degradation status assessment results. The three-parameter Weibull distribution model is a commonly used life prediction model for calculating the remaining life prediction value. The characteristic life parameter θ reflects the characteristic value of the material's average life, and the shape parameter κ describes the shape of the material's life distribution.

[0135] When comparing real-time monitoring data with the performance degradation prediction model, a reasonable error tolerance can be set, for example, ±5% based on the fluctuation range of historical data. When the residual value exceeds this error tolerance, a particle swarm optimization algorithm can be used to update the weight parameters of the temperature-stress coupling model. The specific steps include: initializing a particle swarm, with each particle representing a set of possible weight parameters; calculating the fitness value of each particle, which can be measured by the magnitude of the residual value; and updating the particle's velocity and position based on the fitness value, ultimately finding the optimal weight parameter combination. For performance degradation status assessment, a fuzzy comprehensive evaluation method can be used, taking the residual value, temperature gradient value, thermal stress value, and other parameters as input parameters. Fuzzy rules are used for comprehensive evaluation, outputting the performance degradation level. For remaining life prediction, a three-parameter Weibull distribution model can be combined to fit the model parameters θ and κ based on the current performance degradation index and historical data, thereby calculating the remaining life prediction value. A maintenance recommendation strategy table can be generated based on the remaining life prediction value. For example, when the remaining life falls below a certain threshold, maintenance or replacement of insulation materials is recommended.

[0136] In some embodiments, after generating the smart diagnosis report, the method further includes:

[0137] Associate and store intelligent diagnostic reports with switchgear operation logs;

[0138] When multiple switchgears with the same solid insulation material type are detected to trigger the same warning level, a batch defect analysis report is generated;

[0139] Based on the batch defect analysis report, the initial parameter settings of the temperature-stress coupling model are optimized.

[0140] It should be noted that after generating the intelligent diagnostic report, it is stored in association with the switchgear operation log to facilitate subsequent comprehensive analysis and traceability of the equipment's operating status. When multiple switchgears of the same solid insulation material type trigger the same warning level, a batch defect analysis report is generated to identify any systemic or batch-specific quality issues. The initial parameter settings for the temperature-stress coupling model are optimized based on the batch defect analysis report to improve the model's universality and accuracy, enabling it to better adapt to the characteristics of different batches of materials.

[0141] Specifically, the intelligent diagnostic report is generated based on real-time monitoring data and performance degradation prediction models, and includes performance degradation status assessment, remaining life prediction, and maintenance recommendations. The switchgear operation log records the operating parameters, operation records, and maintenance history of the switchgear. Storing it in association with the intelligent diagnostic report can provide data support for the full life cycle management of the equipment. The batch defect analysis report is generated when multiple switchgear trigger the same warning level. It compares and analyzes the monitoring data and diagnostic results of these switchgear to identify whether there are batch defects. The initial parameter settings of the temperature-stress coupling model refer to the parameter values ​​set at the initial stage of model construction. These parameters include the physical property parameters of the material, such as elastic modulus, thermal expansion coefficient, etc., as well as the weight parameters of the model. By optimizing these initial parameter settings, the model parameters can be adjusted to better reflect the characteristics of different batches of materials by analyzing the common problems found in the batch defect reports.

[0142] Preferably, when the intelligent diagnostic report is stored in association with the switch cabinet operation log, a unified data management system can be used to classify and store information according to the equipment number, timestamp, etc. for quick query and analysis. When generating a batch defect analysis report, the performance degradation index, temperature gradient, thermal stress and other data of multiple switch cabinets that trigger the same warning level can be statistically analyzed, such as calculating the average value, standard deviation, etc., to identify whether there are batch problems. When optimizing the initial parameter settings of the temperature-stress coupling model, the parameters related to the material properties in the model can be adjusted according to the problems found in the batch defect analysis report. For example, if it is found that there is a deviation in the thermal expansion coefficient of a batch of materials, the initial value of the thermal expansion coefficient in the model can be updated after experimental verification. In addition, the model can be retrained and the data in the batch defect analysis report can be used as a supplementary training set to further optimize the weight parameters of the model and improve the adaptability and prediction accuracy of the model to different batches of materials.

[0143] In some embodiments, the specific steps of constructing a gridded detection matrix include:

[0144] The solid insulation surface is divided into m×n rectangular monitoring areas at equal intervals;

[0145] Deploy passive RFID temperature sensors at the center and vertices of each rectangular monitoring area;

[0146] The temperature data of each monitoring area is polled by radio frequency readers in a time-division multiplexing manner.

[0147] It's important to note that constructing a grid-based detection matrix is ​​a fundamental step in testing the performance of switchgear solid insulation materials. By dividing the solid insulation surface into multiple rectangular monitoring zones at equal intervals and deploying passive RFID temperature sensors at the center and vertices of each zone, refined monitoring of the insulation surface temperature distribution is achieved. This layout ensures comprehensive temperature data collection. Furthermore, RFID readers poll each monitoring zone using time-division multiplexing, enabling efficient data collection and transmission.

[0148] Specifically, a grid-based detection matrix divides the surface of a solid insulating material into multiple small rectangular areas, forming a grid-like layout. Equally spaced partitioning involves evenly dividing the insulating material surface at regular intervals, such as 10 cm x 10 cm intervals, to ensure that each monitoring area is of uniform size. Rectangular monitoring areas are each unit in the grid and are used to determine the deployment location of sensors. Passive RFID temperature sensors, which do not require an external power supply, transmit temperature data via RF signals, offering advantages such as easy installation and strong resistance to electromagnetic interference. The center and vertices refer to the geometric center and four corners of the rectangular monitoring area, which serve as sensor deployment points. The center point requires higher sensor accuracy, such as ±0.1°C, while the vertex points can have slightly lower accuracy, such as ±0.5°C. An RFID reader is a device used to read data from passive RFID sensors. It polls sensors in each monitoring area using time-division multiplexing, reading data from each sensor in a timed sequence, thereby achieving comprehensive monitoring of the entire grid-based detection matrix.

[0149] Preferably, when constructing a grid detection matrix, the grid spacing can be flexibly adjusted according to the size and shape of the solid insulating material. For example, for insulating materials with a larger area, the grid spacing can be appropriately increased to improve monitoring efficiency; while for areas with complex shapes, the grid spacing can be appropriately reduced to ensure data accuracy. When deploying passive RFID temperature sensors, it is necessary to ensure that the sensor is installed in the correct position and that the communication frequency bands between sensors do not interfere with each other. For example, the operating frequency band interval of sensors in adjacent monitoring areas should be no less than 10MHz to avoid signal conflicts. In addition, in order to improve the efficiency of data acquisition, the RF reader can adopt a high-speed polling mechanism, such as polling once per second, to ensure that the temperature data of each monitoring area can be obtained in a timely manner. In actual applications, the grid detection matrix can also be expanded or adjusted as needed, such as increasing the density of monitoring points in certain key areas to achieve more accurate monitoring.

[0150] In some embodiments, the deployment of passive RFID temperature sensors must meet the following requirements:

[0151] The accuracy of sensors deployed at the center is ±0.1°C, and the accuracy of sensors deployed at the vertex is ±0.5°C;

[0152] The interval between the operating frequency bands of sensors in adjacent monitoring areas shall not be less than 10 MHz.

[0153] It's important to note that passive RFID temperature sensors must meet specific accuracy and frequency band spacing requirements to ensure accurate monitoring data and prevent interference between sensors. Sensors deployed at central locations require higher accuracy, as they typically better reflect the average temperature of the entire monitoring area. Sensors deployed at vertex locations can be less accurate, as vertex locations primarily serve as auxiliary monitoring and supplementary boundary temperature data. The frequency band spacing between sensors in adjacent monitoring areas must be at least 10 MHz to prevent signal interference between sensors and ensure that each sensor can transmit temperature data stably and accurately.

[0154] Specifically, a passive RFID temperature sensor is a sensor that uses radio frequency signals to transmit temperature data. Its accuracy refers to the accuracy of the sensor's temperature measurement. For example, a sensor deployed at a central point has an accuracy of ±0.1°C, which means that the error range between the temperature value measured by the sensor and the actual temperature is within ±0.1°C; while a sensor deployed at a vertex has an accuracy of ±0.5°C, and the error range is within ±0.5°C. The operating frequency band refers to the frequency range in which the sensor transmits and receives radio frequency signals. The operating frequency band interval of adjacent sensors refers to the minimum frequency difference set to avoid mutual signal interference. For example, if the operating frequency band of a sensor is 900MHz to 910MHz, the starting frequency of the operating frequency band of the adjacent sensor should be at least 920MHz to ensure that the interval between the two is no less than 10MHz.

[0155] When deploying passive RFID temperature sensors, a layered deployment strategy is recommended. For example, for solid insulation materials inside switchgear, sensor density can be increased near the power supply and load sides, as temperature fluctuations in these areas more significantly impact equipment operation. Furthermore, to further optimize sensor deployment, sensor accuracy requirements can be fine-tuned based on the actual application scenario. For example, in applications requiring extremely high temperature monitoring accuracy, the accuracy of the central point sensor can be increased to ±0.05°C.

[0156] Furthermore, the sensor's location and orientation must be considered during installation to minimize external interference and accurately capture temperature data. For example, the sensor should be installed on a flat surface of insulating material and avoid direct exposure to strong electromagnetic fields or bright light.

[0157] In some embodiments, the updating mechanism of the multi-level dynamic warning threshold includes:

[0158] Extract the maximum performance degradation rate V from the historical warning data every month max ;

[0159] If three consecutive V max If the upper threshold of the current warning level L2 is exceeded, the warning thresholds of each level will be increased in proportion to ρ:

[0160]

[0161] Among them, V max is the maximum performance degradation rate in the historical warning data, V base is the baseline degradation rate, which is used to measure the baseline value of the performance degradation rate; ρ is the ratio of the warning threshold to increase, L i is the i-th level warning threshold before adjustment, L′ i is the adjusted level i warning threshold.

[0162] It should be noted that the multi-level dynamic warning threshold update mechanism dynamically adjusts the warning threshold based on historical data, allowing the warning system to adapt to the dynamic characteristics of insulation material performance changes. This method determines whether the warning threshold needs to be adjusted by analyzing the maximum performance degradation rate in historical warning data. If the performance degradation rate exceeds the upper threshold of the current warning level multiple times in a row, it indicates that the current warning threshold may no longer accurately reflect the actual performance changes of the insulation material and needs to be adjusted according to a certain ratio. This dynamic adjustment mechanism can effectively avoid false alarms or missed alarms caused by fixed thresholds, thereby improving the accuracy and reliability of the warning system.

[0163] Specifically, the multi-level dynamic warning threshold refers to dividing the warning into multiple levels, each level corresponding to a different threshold, which is used to distinguish different degrees of performance degradation. The performance degradation rate refers to the rate of change of the performance degradation index over time, which reflects the speed of deterioration of the insulation material performance. Historical warning data refers to the relevant data on performance degradation and warning triggering accumulated by the system during operation. The maximum performance degradation rate refers to the highest value of the performance degradation rate obtained from the historical warning data within a certain period of time. The warning threshold improvement ratio refers to the ratio calculated based on the difference between the maximum performance degradation rate and the baseline degradation rate, which is used to adjust the warning thresholds at all levels. The baseline degradation rate is a pre-set reference value used to measure the average level of insulation material performance degradation under normal operating conditions. The adjusted warning threshold is calculated based on the original warning threshold and the warning threshold improvement ratio, and is used to update the threshold setting of the warning system.

[0164] Preferably, when implementing a multi-level dynamic early warning threshold update mechanism, a reasonable statistical period can be set, for example, extracting the maximum value of the performance degradation rate from the historical early warning data once a month. If the performance degradation rate is found to exceed the upper threshold limit of the current early warning level L2 for three consecutive times, it means that the current early warning threshold may be too low, and it is necessary to adjust the early warning thresholds at all levels according to the early warning threshold increase ratio. The early warning threshold increase ratio can be determined based on the difference between the maximum value of the performance degradation rate and the baseline degradation rate. For example, if the difference is large, it means that the performance degradation rate is accelerating, and the early warning threshold increase ratio can be increased accordingly; conversely, if the difference is small, the early warning threshold increase ratio can be appropriately reduced. The adjusted early warning threshold can be calculated by multiplying the original early warning threshold by (1 + early warning threshold increase ratio). In practical applications, parameters such as the statistical period, baseline degradation rate, and early warning threshold increase ratio can be flexibly adjusted according to different types of insulation materials and operating environments to ensure the accuracy and adaptability of the early warning system.

[0165] The above embodiments of the present invention have the following beneficial effects:

[0166] 1. Through the coordinated deployment of a grid detection matrix and passive radio frequency identification temperature sensors, high-precision distributed monitoring of the surface temperature field of solid insulating materials is achieved. This solves the problem that traditional point temperature measurement methods cannot fully reflect the thermal distribution characteristics of insulating materials, and provides a reliable data foundation for subsequent performance analysis.

[0167] 2. The temperature-stress coupling model is combined with a multi-parameter fusion algorithm of dielectric loss factor and local ionization degree, which breaks through the limitation of evaluating insulation performance with a single temperature indicator. It can accurately characterize the degradation process of the material microstructure and improve the comprehensiveness and accuracy of insulation status diagnosis.

[0168] 3. An adaptive monitoring strategy based on recursive neural networks and dynamic warning thresholds enables full-cycle performance evaluation from real-time status monitoring to long-term life prediction, solving the problem of delayed response in traditional periodic maintenance models and providing intelligent decision-making support for switchgear insulation maintenance.

[0169] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0170] The above descriptions are merely some preferred embodiments of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also encompass other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned inventive concept. For example, a technical solution formed by mutually replacing the above-mentioned features with (but not limited to) technical features having similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for detecting the performance of solid insulation materials of a switch cabinet based on temperature detection, characterized in that: The following steps are involved: Constructing a grid detection matrix on the surface of the solid insulation material of the switch cabinet, the grid detection matrix includes multiple monitoring areas; The real-time temperature data of each monitoring area is collected through a passive radio frequency identification temperature sensor to establish an initial temperature field data set; Based on the calibrated initial temperature field data set, a deep learning algorithm is used to calculate the temperature gradient value and thermal stress distribution of each monitoring area of ​​the solid insulation material; Combining the temperature gradient value with the thermal stress distribution, a temperature-stress coupling model is established; The dielectric loss factor and local ionization degree of each monitoring area of ​​the solid insulation material are calculated based on the temperature-stress coupling model, and a performance degradation index that includes the microstructural evolution characteristics is constructed. Based on the temporal variation of the performance degradation index, a performance degradation prediction model for solid insulation materials is established; Using a performance degradation prediction model, combined with a recurrent neural network and an attention mechanism, the performance evolution trajectory of solid insulating materials is predicted; Establish multi-level dynamic warning thresholds based on performance evolution trajectories; When an early warning trigger is detected, the data sampling strategy of the monitoring points in the grid detection matrix is ​​adaptively adjusted based on the change trend of the calibrated initial temperature field data set; The real-time monitoring data obtained after adjustment is compared with the performance degradation prediction model, and the parameters of the temperature-stress coupling model are continuously optimized to generate an intelligent diagnostic report including performance degradation status assessment, remaining life prediction and maintenance recommendations.

2. The method according to claim 1, characterized in that Based on the calibrated initial temperature field data set, a deep learning algorithm is used to calculate the temperature gradient value and thermal stress distribution of each monitoring area of ​​the solid insulation material; Combining the temperature gradient value with the thermal stress distribution, a temperature-stress coupling model is established, including: The initial temperature field dataset is input into the convolutional neural network to extract the spatial temperature distribution characteristics of each monitoring area; The temperature gradient value of each monitoring area is output through the gradient calculation layer. The temperature gradient value is expressed as: Where T is the temperature of the monitoring area, x and y are the coordinates of the grid detection matrix, is the temperature gradient value of the monitoring area, indicating the degree of temperature change in the x and y directions. Solve the thermal stress distribution equation based on the finite element analysis method: σ=E·α·ΔT Where σ is the thermal stress value, E is the elastic modulus of the solid insulation material, α is the thermal expansion coefficient, and ΔT is the temperature change in the monitoring area. The temperature gradient value and thermal stress distribution are input into the coupling layer to generate a temperature-stress coupling model.

3. The method according to claim 2, characterized in that The dielectric loss factor and local ionization degree of each monitoring area of ​​the solid insulation material are calculated based on the temperature-stress coupling model, and a performance degradation index that includes the microstructural evolution characteristics is constructed. Based on the temporal changes of the performance degradation index, a performance degradation prediction model for solid insulation materials is established, including: The dielectric loss factor tanδ and local ionization degree η of each monitoring area are output through the temperature-stress coupling model; The calculation formula for constructing the performance degradation index is: Among them, α, β, γ, and λ are weight coefficients used to adjust the importance of temperature gradient value, thermal stress value, dielectric loss factor, and local ionization degree in the calculation of performance degradation index; PI is the performance degradation index, which comprehensively reflects the performance degradation degree of solid insulating materials. Collect the time series of performance degradation index changes in historical data, input it into the long short-term memory network, and train it to obtain a performance degradation prediction model; The performance degradation prediction model outputs the predicted value of the performance degradation index in the future time step.

4. The method according to claim 3, characterized in that Using a performance degradation prediction model, combined with a recurrent neural network and an attention mechanism, the performance evolution trajectory of solid insulating materials is predicted; Based on the performance evolution trajectory, multi-level dynamic warning thresholds are established, including: Input the performance degradation index prediction value into the recurrent neural network to generate the performance evolution trajectory function f(t); The weight distribution of different time steps is calculated through the attention mechanism to optimize the performance evolution trajectory function f(t); Based on the first-order derivative of the performance evolution trajectory function f(t) Determine the rate of performance degradation, Indicates the rate of change of performance degradation index over time; According to the performance degradation rate and the preset probability distribution threshold, multi-level dynamic warning thresholds are divided, including warning levels L1, L2, and L3, where L3 corresponds to the highest degradation rate.

5. The method according to claim 4, characterized in that When an early warning trigger is detected, the data sampling strategy of the monitoring points in the grid detection matrix is ​​adaptively adjusted based on the change trend of the calibrated initial temperature field data set, including: When the warning level L1 is triggered, the data sampling frequency of the low-risk monitoring area is reduced to the first frequency value; When the warning level L2 is triggered, the data sampling frequency of the high-risk monitoring area is increased to the second frequency value, and the redundant sensors in the adjacent monitoring areas are activated; When the warning level L3 is triggered, the highest sampling frequency of the full grid detection matrix is ​​started, and the power supply current of the passive RFID temperature sensor is increased synchronously.

6. The method according to claim 5, characterized in that Compare the adjusted real-time monitoring data with the performance degradation prediction model, continuously optimize the parameters of the temperature-stress coupling model, and generate an intelligent diagnostic report that includes performance degradation status assessment, remaining life prediction, and maintenance recommendations, including: Calculate the residual value between the real-time monitoring data and the output of the performance degradation prediction model; When the residual value exceeds the preset error tolerance, the particle swarm optimization algorithm is used to update the weight parameters of the temperature-stress coupling model; Based on the updated temperature-stress coupling model, the performance degradation status assessment results are output through the fuzzy comprehensive evaluation method; The remaining life prediction value is calculated by combining the three-parameter Weibull distribution model: Among them, θ is the characteristic life parameter, which reflects the characteristic value of the average life of the material; κ is the shape parameter, which describes the shape of the material life distribution; R(t) is the predicted value of the remaining life at time t. A maintenance recommendation strategy table is generated based on the remaining life prediction value, and the maintenance recommendation strategy table includes a maintenance time window and maintenance level.

7. The method according to claim 1, characterized in that After generating the smart diagnostic report, it also includes: Associate and store intelligent diagnostic reports with switchgear operation logs; When multiple switchgears with the same solid insulation material type are detected to trigger the same warning level, a batch defect analysis report is generated; Based on the batch defect analysis report, the initial parameter settings of the temperature-stress coupling model are optimized.

8. The method according to claim 1, characterized in that The specific steps of constructing a gridded detection matrix include: The solid insulation surface is divided into m×n rectangular monitoring areas at equal intervals; Deploy passive RFID temperature sensors at the center and vertices of each rectangular monitoring area; The temperature data of each monitoring area is polled by radio frequency readers in a time-division multiplexing manner.

9. The method according to claim 8, characterized in that The deployment of passive RFID temperature sensors must meet the following requirements: The accuracy of sensors deployed at the center is ±0.1°C, and the accuracy of sensors deployed at the vertex is ±0.5°C; The interval between the operating frequency bands of sensors in adjacent monitoring areas shall not be less than 10 MHz.

10. The method according to claim 1, characterized in that The update mechanism for the multi-level dynamic warning threshold includes: Extract the maximum performance degradation rate V from the historical warning data every month max ; If three consecutive V max If the upper threshold of the current warning level L2 is exceeded, the warning thresholds of each level will be increased in proportion to ρ: Among them, V max is the maximum performance degradation rate in the historical warning data, V base is the baseline degradation rate; ρ is the ratio of the warning threshold to increase, L i is the i-th level warning threshold before adjustment, L′ i is the adjusted level i warning threshold.

Images