Hot spot temperature detection method of environmental protection gas switchgear based on multi-physics field simulation

Abstract

The present invention belongs to the field of simulation analysis technology, and specifically relates to a hotspot temperature detection method for an environmentally friendly gas switchgear based on multi-physics field simulation. The method comprises the following steps: establishing a geometric model for multi-physics field coupling analysis and importing it into COMSOL, adding corresponding material properties; adding various physical fields and performing multi-physics field coupling; performing meshing; adding a frequency domain transient study and using a transient solver for direct solution; post-processing the simulation solution results; measuring the temperature during actual operation and optimizing and adjusting the simulation model; extracting the hotspot temperature based on the optimized and adjusted simulation model and performing temperature inversion; constructing a deep neural network (DNN) model and training it; using the gas chamber shell temperature measured during actual operation as input, and using the DNN model for prediction to obtain the corresponding hotspot temperature. The present invention can improve the efficiency and accuracy of hotspot temperature detection without changing the existing switchgear structure.

Description

Technical Field

[0001] The present invention belongs to the technical field of simulation analysis, and in particular relates to a hot spot temperature detection method for an environmentally friendly gas switch cabinet based on multi-physical field simulation. Background Art

[0002] Environmentally friendly gas switchgear, as a crucial piece of power equipment, is widely used in high- and ultra-high-voltage power systems. While the environmentally friendly gases (such as SF6 alternatives) used in these switchgear offer significant environmental advantages, their specialized operating environment and structural design also complicate temperature distribution. During operation, as load current increases, hotspot temperatures within the switchgear can rise significantly, particularly in the circuit breaker's moving and stationary contact areas. Failure to monitor these hotspot temperatures in a timely and effective manner can lead to equipment failure, impacting both equipment stability and the safe operation of the power system.

[0003] According to existing research, the hotspot temperature distribution of environmentally friendly gas switchgear is closely related to multiple factors, including load current, gas chamber shell temperature, gas flow conditions, and heat generation of conductors. Currently, the hotspot temperature detection methods for environmentally friendly gas switchgear mainly include temperature sensor method and infrared imaging method. The temperature sensor method monitors temperature changes in real time by installing sensors at key locations. Although the technology is mature and the cost is low, due to the uneven temperature distribution inside the switchgear, the sensor cannot directly measure the hotspot area, and may be affected by environmental factors, resulting in reduced accuracy and stability. The infrared imaging method infers the temperature by measuring infrared radiation on the surface of the equipment. It has the advantages of non-contact measurement and rapid location of hotspots, but it also has the problem of only being able to measure surface temperature and cannot accurately reflect the internal hotspot temperature. It is also greatly affected by environmental factors, making long-term dynamic monitoring difficult.

[0004] In practical applications, rising temperatures are often a precursor to equipment failure. Failure to promptly monitor the hotspot temperature of environmentally friendly gas switchgear can lead to serious consequences such as overheating, insulation damage, and even fire. Existing hotspot temperature detection methods still have limitations in accuracy and comprehensiveness. There is an urgent need to optimize existing detection schemes by combining multiple technical means to improve the accuracy and real-time performance of temperature monitoring. Developing an accurate and efficient hotspot temperature detection method is particularly important, especially when direct measurement of the hotspot area is not possible. Summary of the Invention

[0005] In view of the above deficiencies in the prior art, the purpose of the present invention is to provide a hotspot temperature detection method for an environmentally friendly gas switchgear based on multi-physics field simulation. By measuring the shell temperature of the environmentally friendly gas switchgear during actual operation, combining it with a pre-established multi-physics field coupling simulation model, and using an algorithm model to solve the hotspot temperature, the efficiency and accuracy of hotspot temperature detection can be improved without changing the existing switchgear structure.

[0006] To achieve the above objectives, the present invention provides a hotspot temperature detection method for an environmentally friendly gas switchgear based on multi-physics field simulation, comprising the following steps:

[0007] S1. Use 3D drawing software to establish a geometric model for multi-physics coupling analysis of environmental gas switchgear;

[0008] S2. Import the geometric model into COMSOL finite element simulation software to obtain a simulation model, and add corresponding material properties to each component of the simulation model;

[0009] S3. Add the current physics field, solid and fluid heat transfer physics field, surface-to-surface radiation physics field, and turbulent k-ε physics field in COMSOL, set the current field calculation domain, temperature field calculation domain, radiation field calculation domain, and flow field calculation domain accordingly, and set the boundary conditions for electric field analysis, temperature analysis, radiation analysis, and fluid flow analysis accordingly;

[0010] S4. Couple multiple physical fields: the current physical field and the solid and fluid heat transfer physical field are coupled into an electromagnetic heat field; the surface-to-surface radiation physical field and the solid and fluid heat transfer physical field are coupled into a surface-to-surface radiation heat transfer field; the solid and fluid heat transfer physical field and the turbulent k-ε physical field are coupled into a non-isothermal flow field;

[0011] S5. Use the free tetrahedron meshing method to mesh the simulation model;

[0012] S6. Add a frequency-domain transient study in COMSOL and use the transient solver for direct solution.

[0013] S7. Post-process the simulation results to obtain the temperature distribution of the current-carrying conductor and the streamline distribution inside the air chamber;

[0014] S8. Select temperature measurement points to measure the current-carrying conductor temperature and the gas chamber shell temperature during actual operation of the environmentally friendly gas switchgear;

[0015] S9. Compare the measured data with the simulation results, evaluate the error of the simulation model in temperature prediction, and optimize and adjust the simulation model;

[0016] S10. In the optimized and adjusted simulation model, input the operating current and environmental conditions, perform simulation, extract the hotspot temperature, construct a data sample including multi-dimensional input features and target output, and construct a data set, wherein the target output is the hotspot temperature;

[0017] S11. Combining the thermal resistance characteristics of the internal structure of the switchgear, an adaptive weight distribution method based on the thermal resistance network is used to perform temperature inversion;

[0018] S12. Build a deep neural network (DNN) model, introduce physical constraints into the loss function, and optimize the loss function using the constructed dataset.

[0019] S13, allowing the DNN model to learn the mapping relationship between the air chamber shell temperature and the hot spot temperature, and obtaining a trained DNN model;

[0020] S14. The air chamber shell temperature measured during actual operation is used as input, and the trained DNN model is used for prediction to obtain the corresponding hotspot temperature.

[0021] As a preferred embodiment of the present invention, in S1, the three-dimensional drawing software adopts SOLIDWORKS, and the geometric model includes a current-carrying busbar, a three-position switch, a circuit breaker, a heat sink, an air chamber shell of an environmentally friendly gas switch cabinet, and a heat dissipation backpack, wherein the current-carrying busbar, the three-position switch, and the circuit breaker are current-carrying conductors. According to the actual structural parameters of the environmentally friendly gas switch cabinet, a high-precision three-dimensional geometric model of the original size is constructed.

[0022] As a preferred embodiment of the present invention, in S2, the material of the current-carrying busbar, three-position switch, and heat sink is set to copper, the material of the moving and static contacts of the circuit breaker is set to copper-chromium alloy, the material of the vacuum interrupter sleeve of the circuit breaker is set to alumina, and the material of the air chamber shell and the heat dissipation backpack is set to stainless steel, and the setting method is to select from the material library.

[0023] As a preferred solution of the present invention, in S3, the specific setting method of each physical field is:

[0024] S3.1. For the current physical field, the current field calculation domain is selected as the current-carrying busbar, three-position switch, heat sink, and the moving and static contacts of the circuit breaker. A surface terminal is added to provide a current input surface, and a ground plane is added to provide a current outflow surface. The three-phase AC current is set to I0, I0×exp(+120[deg]×j), and I0×exp(-120[deg]×j), respectively. Where I0 is the effective current value of the environmental gas switchgear in actual operation, exp is an exponential function used to express the power of the base e of the natural logarithm, 120[deg] represents 120 degrees, which is the phase difference between every two phases in three-phase AC, and j is an imaginary unit used to represent a 90-degree phase difference in complex number representation.

[0025] By adding contact impedance to the contact surface to simulate the contact resistance of the contacts and the bolt fastening surface, the contact heating during the operation of the environmental protection gas switch cabinet is simulated. The contacts are the moving and static contacts of the circuit breaker. The contact resistance calculation formula is:

[0026] ;

[0027] Where, K is a constant whose value is determined by the contact material; m is an index whose value is related to the contact form of the two contact surfaces; F is the contact pressure;

[0028] S3.2. For the solid and fluid heat transfer physical fields, the temperature field calculation domain selects all domains except the vacuum medium in the vacuum interrupter of the circuit breaker. The solid domain includes the current-carrying busbar, three-position switch, circuit breaker, heat sink and air chamber shell, and the fluid domain includes the air inside the air chamber. Heat flux is added to simulate the heat exchange between the air chamber shell and the outside air. The flux type selects convection heat flux-user-defined heat transfer coefficient. The heat transfer coefficient Calculated by the following formula:

[0029] ;

[0030] Where, is the outer wall surface temperature of the gas chamber shell; is the external air temperature of the air chamber shell;

[0031] S3.3. For the surface-to-surface radiation physics field, select the entire model domain as the radiation field calculation domain. Set the emitted radiation direction to be controlled by opacity. Set the air inside the gas chamber and the vacuum inside the vacuum interrupter of the circuit breaker as transparent domains, and set the rest as opaque domains.

[0032] S3.4. For the turbulent k-ε physics field, select the air domain inside the air chamber as the flow field calculation domain. Set the fluid domain physical model to compressible flow, include gravity, use reduced pressure, and have a Mach number Ma less than 0.3. Set the wall condition to no slip, add a pressure point constraint, and select a point far away from the calculation domain as the pressure point.

[0033] As a preferred embodiment of the present invention, in S5, when meshing the simulation model, the moving and static contacts of the circuit breaker are meshed using a refined mesh, and the remaining parts, namely the current-carrying busbar, the three-position switch, the gas chamber housing, and the air domain inside the gas chamber, are meshed using a conventional mesh.

[0034] As a preferred embodiment of the present invention, in the aforementioned S8, an experimental platform is built to perform actual measurements on the environmentally friendly gas switch cabinet. During the measurement, the temperature measurement points of the current-carrying conductor temperature are selected as follows: based on the temperature distribution of the current-carrying conductor, the temperature measurement points cover the hotspot temperature area identified in the simulation, and the temperature measurement points are arranged at positions where the temperature sensor can contact under the test conditions and at the joints of the current-carrying conductors, and the temperature of the measurement points is collected using a contact temperature measurement method;

[0035] The temperature measurement points of the air chamber shell are selected based on the analysis of the streamline distribution inside the air chamber. When the air reaches the vicinity of the air chamber shell, the projection on the shell is the path of heat transport, and the temperature measurement points are set along the path of heat transport.

[0036] As a preferred embodiment of the present invention, in S9, the simulation model is optimized and adjusted by adjusting the material parameters, boundary conditions, and mesh density of the mesh subdivision based on the comparison between the measurement data and the simulation results; in S10, the multidimensional input features include operating current parameters, environmental condition parameters, air chamber shell temperature, material parameters, geometric structure parameters, and boundary conditions of the simulation analysis; the hot spot temperature is the temperature of the moving and static contacts of the circuit breaker during the simulation process.

[0037] As a preferred solution of the present invention, in said S11, the adaptive weight distribution method based on the thermal resistance network is specifically to design a dynamic weight distribution formula to calculate the influence weight of each temperature measurement point on the hot spot temperature ,Will As the weighting coefficient of the input feature of the DNN model, the dynamic weight allocation formula is:

[0038] ;

[0039] Where, is the thermal resistance from the temperature measuring point i to the hot spot; is the spatial distance from the temperature measurement point i to the hotspot; is the attenuation coefficient; exp is the exponential function; the hot spots are the moving and static contacts of the circuit breaker.

[0040] As a preferred embodiment of the present invention, in the S12, the DNN model includes three hidden layers and one output layer, the first hidden layer is provided with 128 neurons and adopts the ReLU activation function; the second hidden layer is provided with 64 neurons; the third hidden layer is provided with 64 neurons; the output layer is provided with 1 neuron to output the predicted hotspot temperature value;

[0041] In the DNN model, the loss function L that introduces physical constraints is:

[0042] ;

[0043] Where, Indicates the predicted temperature and simulation temperature The mean square error of Representing the Laplace operator of the heat conduction equation The residual of 、 is a hyperparameter used to adjust data fitting and physical constraint weights.

[0044] As a preferred embodiment of the present invention, in the S13, during training, the weights of each layer of the DNN model are updated according to the gradient of the loss function through the back propagation algorithm, and a dynamic feedback mechanism is introduced in the training process. After each simulation model update, the parameters of the DNN model are calibrated using the latest actual measurement data.

[0045] The simulation and algorithm involved in the present invention can be executed by an electronic device, which includes a memory, a processor, and a computer program stored in the memory and run on the processor. The above-mentioned simulation and algorithm calculation are realized by executing the software through the processor.

[0046] The beneficial effects of the present invention are:

[0047] The present invention measures the shell temperature of the environmentally friendly gas switchgear during actual operation, combines it with a pre-established multi-physics field coupling simulation model, and uses an algorithm model to solve the hotspot temperature. This can improve the efficiency and accuracy of hotspot temperature detection without changing the existing switchgear structure.

[0048] Based on a finite element multi-physics field simulation model, this invention clearly demonstrates the temperature rise process of environmentally friendly gas switchgear, enabling visual simulation of the entire process and constantly monitoring hotspot temperature changes during the temperature rise. Simulation boundary conditions and temperature detection locations can be adjusted to suit different operating environments and conditions. In particular, it can detect the real-time temperature of current-carrying conductors, casings, and other parts of the environmentally friendly gas switchgear when it overheats, providing a reference for temperature detection methods for environmentally friendly gas switchgear.

[0049] By collecting real-time temperature data from current-carrying conductors and combining it with a multi-physics simulation model, this method accurately predicts hotspot temperatures in environmentally friendly gas switchgear, avoiding the complex sensor layout and error accumulation issues. Furthermore, this method provides a low-cost, high-precision temperature rise detection method, which is of great significance for improving the safety and reliability of switchgear operation.

[0050] The present invention enhances the adaptability and reliability of the DNN model by introducing physical constraints and dynamic weight distribution, and ultimately ensures that the model can continuously adapt to changes in operating conditions through a closed-loop optimization mechanism, thereby significantly improving the safety and reliability of power equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the process of the present invention;

[0052] Figure 2 is a schematic diagram of a geometric model of an environmentally friendly gas switchgear cabinet in an embodiment of the present invention;

[0053] Figure 3 1 is a schematic diagram of the interior of a simulation model of an environmentally friendly gas switchgear cabinet according to an embodiment of the present invention;

[0054] Figure 4 It is a flow diagram of Example 2. DETAILED DESCRIPTION

[0055] The embodiments of the present invention are further described below with reference to the accompanying drawings:

[0056] Example 1:

[0057] like Figure 1 As shown, the hot spot temperature detection method of the environmental protection gas switch cabinet based on multi-physics field simulation includes the following steps:

[0058] S1. Use 3D drawing software to establish a geometric model for multi-physics coupling analysis of environmental gas switchgear;

[0059] S2. Import the geometric model into COMSOL finite element simulation software to obtain a simulation model, and add corresponding material properties to each component of the simulation model;

[0060] S3. Add the electric current (ec) physics field, solid and fluid heat transfer (ht) physics field, surface-to-surface radiation (rad) physics field, and turbulent k-ε (spf) physics field in COMSOL, set the current field calculation domain, temperature field calculation domain, radiation field calculation domain, and flow field calculation domain accordingly, and set the boundary conditions for electric field analysis, temperature analysis, radiation analysis, and fluid flow analysis accordingly;

[0061] S4. Couple multiple physical fields: the current physics field, solid and fluid heat transfer physics field are coupled into electromagnetic heat field (emh); the surface-to-surface radiation physics field, solid and fluid heat transfer physics field are coupled into surface-to-surface radiation heat transfer field (htrad); the solid and fluid heat transfer physics field, turbulent k-ε physics field are coupled into non-isothermal flow field (nitf);

[0062] S5. Use the free tetrahedron meshing method to mesh the simulation model;

[0063] S6. Add a frequency-domain transient study in COMSOL and use the transient solver for direct solution.

[0064] S7. Post-process the simulation results to obtain the temperature distribution of the current-carrying conductor and the streamline distribution inside the air chamber;

[0065] S8. Measure the current-carrying conductor temperature and the gas chamber shell temperature during actual operation of the environmentally friendly gas switchgear;

[0066] S9. Compare the measured data with the simulation results, evaluate the error of the simulation model in temperature prediction, and optimize and adjust the simulation model;

[0067] S10. In the optimized and adjusted simulation model, input the operating current and environmental conditions, perform simulation, extract the hotspot temperature, construct a data sample including multi-dimensional input features and target output, and construct a data set (based on the data sample), wherein the target output is the hotspot temperature;

[0068] S11. Combining the thermal resistance characteristics of the internal structure of the switchgear, an adaptive weight distribution method based on the thermal resistance network is used to perform temperature inversion;

[0069] S12. Build a deep neural network (DNN) model, introduce physical constraints into the loss function, and optimize the loss function using the constructed dataset.

[0070] S13, allowing the DNN model to learn the mapping relationship between the air chamber shell temperature and the hot spot temperature, and obtaining a trained DNN model;

[0071] S14. The air chamber shell temperature measured during actual operation is used as input, and the trained DNN model is used for prediction to obtain the corresponding hotspot temperature.

[0072] The turbulent k-ε (k-Epsilon) physics field refers to the distribution of physical quantities related to turbulent kinetic energy (k) and turbulent dissipation rate (ε) in COMSOL's k-ε turbulence model. A frequency-domain-transient study is a simulation method used in COMSOL that combines two different solution types: frequency-domain and transient, each targeting different application modes.

[0073] In S1, the 3D drawing software used is SOLIDWORKS (other software with the same function can also be used). The geometric model includes the current-carrying busbar, three-position switch, circuit breaker, heat sink, gas chamber shell of the environmental gas switch cabinet, and heat dissipation backpack. The current-carrying busbar, three-position switch and circuit breaker are current-carrying conductors. According to the actual structural parameters of the environmental gas switch cabinet, a high-precision 3D geometric model of the original size is constructed. The complete geometric model is as follows Figure 2 As shown, in actual construction, the connectors can be removed to simplify the geometric model, and only the main components mentioned above are retained. After the simulation model is constructed, its internal structure is as follows Figure 3 shown.

[0074] A current-carrying busbar (including the top and bottom busbars) is a conductor that carries current and is primarily used to collect, distribute, and transmit electrical energy and connect primary equipment. A three-position switch has closed, open, and grounded positions. A circuit breaker is a switching device capable of closing, carrying, and interrupting current under normal circuit conditions, and closing, carrying, and interrupting current under abnormal circuit conditions within a specified timeframe. The gas chamber housing is part of the environmentally friendly gas switchgear and, together with the heat sink, constitutes the protection and heat dissipation system of the environmentally friendly gas switchgear.

[0075] In S2, the materials of the current-carrying busbar, three-position switch, and heat sink are set to copper, the materials of the moving and static contacts of the circuit breaker are set to copper-chromium alloy, the material of the vacuum interrupter sleeve of the circuit breaker is set to aluminum oxide, and the material of the gas chamber shell and heat dissipation backpack is set to stainless steel. The setting method is to select from the material library.

[0076] In S3, the specific settings of each physical field are as follows:

[0077] S3.1. For the current physical field, the current field calculation domain is selected as the current-carrying busbar, three-position switch, heat sink, and the moving and static contacts of the circuit breaker. A surface terminal is added to provide a current input surface, and a ground plane is added to provide a current outflow surface. The three-phase AC current is set to I0, I0×exp(+120[deg]×j), and I0×exp(-120[deg]×j), respectively. Where I0 is the effective current value of the environmental gas switchgear in actual operation, exp is an exponential function used to express the power of the base e of the natural logarithm, 120[deg] represents 120 degrees, which is the phase difference between every two phases in three-phase AC, and j is an imaginary unit used to represent a 90-degree phase difference in complex number representation.

[0078] By adding contact impedance to the contact surface to simulate the contact resistance of the contacts and the bolt fastening surface, the contact heating during the operation of the environmental protection gas switch cabinet is simulated. The contacts are the moving and static contacts of the circuit breaker. The contact resistance calculation formula is:

[0079] ;

[0080] Where R is the contact resistance between the contacts, measured in micro-ohms; K is a constant whose value is determined by the contact material. For example, for copper-chromium alloy, K is 800-1500; m is an index whose value is related to the contact form of the two contact surfaces; F is the contact pressure, measured in Newtons.

[0081] S3.2. For the solid and fluid heat transfer physical fields, the temperature field calculation domain selects all domains except the vacuum medium in the vacuum interrupter of the circuit breaker. The solid domain includes the current-carrying busbar, three-position switch, circuit breaker, heat sink and air chamber shell, and the fluid domain includes the air inside the air chamber. Heat flux is added to simulate the heat exchange between the air chamber shell and the outside air. The flux type selects convection heat flux-user-defined heat transfer coefficient. The heat transfer coefficient Calculated by the following formula:

[0082] ;

[0083] Where, is the outer wall surface temperature of the gas chamber shell; is the external air temperature of the air chamber shell;

[0084] S3.3. For the surface-to-surface radiation physics field, select the entire model domain as the radiation field calculation domain. Set the emitted radiation direction to be controlled by opacity. Set the air inside the gas chamber and the vacuum inside the vacuum interrupter of the circuit breaker as transparent domains, and set the rest as opaque domains.

[0085] S3.4. For the turbulent k-ε physics field, select the air domain inside the air chamber as the flow field calculation domain. Set the fluid domain physical model to compressible flow, include gravity, use reduced pressure, and have a Mach number Ma less than 0.3. Set the wall condition to no slip, add a pressure point constraint, and select a point far away from the calculation domain as the pressure point.

[0086] The settings of heat flux simulation, wall conditions, radiation transparency, current size, etc. are the settings for boundary conditions in the physical field.

[0087] For the value of the exponent m, it is 1 for surface contact, 0.7 for line contact, 0.5 for point contact, and 1 for elastic contact and plastic contact.

[0088] In S5, when meshing the simulation model, the moving and static contacts of the circuit breaker are meshed using a refined mesh, while the remaining parts, namely the current-carrying busbar, three-position switch, gas chamber housing, and the air domain inside the gas chamber, are meshed using a conventional mesh.

[0089] Fine meshing uses tetrahedrons, hexahedrons, and even octahedrons as mesh elements. This provides a high mesh density and the ability to accurately describe complex shapes, particularly the geometric morphology of three-dimensional solid boundary surfaces. Fine meshing is often used for areas requiring high-precision simulations, such as stress concentration points or areas with large temperature gradients in thermal analysis. Conventional meshing, also known as structured meshing, has a moderate mesh density and consists of regularly arranged rectangular or other polygonal elements. It is suitable for simulating large areas without overly refined problem domains.

[0090] In S6, the frequency domain-transient study frequency is set to 50 Hz, and the MUPUS solver (transient solver) is used for solution. The solution method is direct solution, and the relative tolerance is set to 0.0001.

[0091] In S8, an experimental platform is built (using known technology) to perform actual measurements on the environmentally friendly gas switchgear. During the measurement, the temperature measurement points of the current-carrying conductor are selected based on the temperature distribution of the current-carrying conductor, covering the hot spot temperature area identified in the simulation. The temperature measurement points are arranged at positions that can be contacted by the temperature sensor under the test conditions and at the joints of the current-carrying conductors. The temperature of the measurement points is collected using a contact temperature measurement method.

[0092] The temperature measurement points of the air chamber shell are selected based on the analysis of the streamline distribution inside the air chamber (i.e., airflow field analysis. Based on the turbulent k-ε physical field, the streamline distribution of the air flow inside the air chamber can be solved). When the air reaches the vicinity of the air chamber shell, the projection on the shell is the path of heat transport, and the temperature measurement points are set along the path of heat transport.

[0093] In S9, the simulation model is optimized and adjusted by adjusting the material parameters, boundary conditions, and mesh density of the meshing based on the comparison between the measurement data and the simulation results;

[0094] In S10, the multidimensional input features include operating current parameters, environmental condition parameters, gas chamber shell temperature, material parameters, geometric structure parameters and boundary conditions of simulation analysis; the hot spot temperature is the temperature of the moving and static contacts of the circuit breaker during the simulation process.

[0095] The dataset is divided into training, validation, and test sets to ensure the independence of DNN model training and validation. When constructing the dataset, real-time environmental parameters (i.e., environmental condition parameters such as humidity and air pressure) are introduced as input variables to improve the DNN model's adaptability to dynamic working conditions. In each sample, in addition to recording the hotspot temperature values ​​generated by the simulation, the residual of the heat conduction equation from the multi-physics field simulation is added as a constraint term. This constraint term is used as part of the loss function in subsequent training.

[0096] In S11, the adaptive weight distribution method based on the thermal resistance network is specifically to design a dynamic weight distribution formula to calculate the influence weight of each temperature measurement point on the hot spot temperature ,Will As the weighted coefficient of the DNN model input feature, the contribution of key temperature measurement points is improved. The dynamic weight distribution formula is:

[0097] ;

[0098] Where, is the thermal resistance from the temperature measuring point i to the hot spot; is the spatial distance from the temperature measurement point i to the hotspot; is the attenuation coefficient; exp is the exponential function; the hot spots are the moving and static contacts of the circuit breaker.

[0099] In S12, the DNN model includes three hidden layers and one output layer. The first hidden layer has 128 neurons and uses the ReLU activation function to fully extract the original features. The second hidden layer has 64 neurons to further reduce the dimension and extract features. The third hidden layer has 64 neurons to integrate features and fit nonlinear relationships. The output layer has 1 neuron to output the predicted hotspot temperature value.

[0100] In the DNN model, the loss function L that introduces physical constraints is:

[0101] ;

[0102] Where, Indicates the predicted temperature and simulation temperature The mean square error of Representing the Laplace operator of the heat conduction equation The residual of 、 is a hyperparameter used to adjust data fitting and physical constraint weights.

[0103] In S13, the input data is progressively calculated through each layer of the DNN model to obtain a predicted output, and a composite loss function is calculated. During training, the weights of each layer of the DNN model are updated based on the gradient of the loss function through a backpropagation algorithm. A dynamic feedback mechanism is also introduced during the training process. After each simulation model update, the DNN model parameters are calibrated using the latest actual measurement data. For example, thermal resistance parameters, DNN model architecture, and hyperparameters can be adjusted, including re-screening or reweighting input features, and appropriately adjusting the number of hidden layers and the number of neurons in each layer to adapt to changes in data distribution and physical property updates. This achieves a closed-loop optimization of "simulation-experimentation-algorithm". Through continuous actual data feedback, timely adjustments can be made to adapt to changes in operating conditions.

[0104] Example 2

[0105] like Figure 4 As shown, this embodiment provides another implementation method for achieving the purpose of the present invention. The difference from Example 1 is that starting from S10, the following steps are adopted:

[0106] S10. In the optimized and adjusted simulation model, input the operating current and environmental conditions, perform simulation, extract the hotspot temperature, construct a data set, and perform temperature inversion using the support vector machine algorithm SVM;

[0107] S11. Construct an objective function and a loss function of a support vector machine regression model, optimize the loss function using the constructed data set, and learn the model parameters of the support vector machine regression model;

[0108] S12, using the constructed data set as a training data set, setting the parameters of the support vector machine regression model, allowing the SVM to learn the mapping relationship between the air chamber shell temperature and the hot spot temperature, and obtaining a trained support vector machine regression model;

[0109] S13. Using the air chamber shell temperature measured during actual operation as input, the trained support vector machine regression model is used for prediction to obtain the corresponding hotspot temperature.

[0110] The support vector machine regression model is the model used in SVM, which fits the data by minimizing the prediction error and maintaining a boundary (interval) during the fitting process so that most data points fall within this boundary.

[0111] In S11, the RBF radial basis function kernel is selected as the kernel function of SVM, and the objective function of the support vector machine regression model is Expressed as:

[0112] ;

[0113] Where x is the temperature of the gas chamber shell; It is the high-dimensional feature space mapped by the kernel function (the kernel function maps the data set data to the high-order space. The RBF radial basis function kernel is suitable for complex nonlinear relationships and can capture the complex nonlinear relationship between temperatures); w and b are learning parameters;

[0114] The loss function is expressed as:

[0115] ;

[0116] Where C is the regularization parameter that controls the model complexity; i is the index of the data sample; n is the number of data samples; is the error tolerance of data sample i, which indicates the error range allowed for the sample point;

[0117] In S12, the parameters of the support vector machine regression model are set to the parameters of the RBF radial basis function kernel , regularization parameter C and error tolerance of loss function ,in The influence range of a single sample is defined in the RBF radial basis function kernel; after the parameters are determined, cross-validation is used to select the optimal parameters.

Claims

1. A hot spot temperature detection method for an environmentally friendly gas switch cabinet based on multi-physics field simulation is characterized in that The following steps are involved: S1. Use 3D drawing software to establish a geometric model for multi-physics coupling analysis of environmental gas switchgear; S2. Import the geometric model into COMSOL finite element simulation software to obtain a simulation model, and add corresponding material properties to each component of the simulation model; S3. Add the current physics field, solid and fluid heat transfer physics field, surface-to-surface radiation physics field, and turbulent k-ε physics field in COMSOL, set the current field calculation domain, temperature field calculation domain, radiation field calculation domain, and flow field calculation domain accordingly, and set the boundary conditions for electric field analysis, temperature analysis, radiation analysis, and fluid flow analysis accordingly; S4. Couple multiple physical fields: the current physical field and the solid and fluid heat transfer physical field are coupled into an electromagnetic heat field; the surface-to-surface radiation physical field and the solid and fluid heat transfer physical field are coupled into a surface-to-surface radiation heat transfer field; the solid and fluid heat transfer physical field and the turbulent k-ε physical field are coupled into a non-isothermal flow field; S5. Use the free tetrahedron meshing method to mesh the simulation model; S6. Add a frequency-domain transient study in COMSOL and use the transient solver for direct solution. S7. Post-process the simulation results to obtain the temperature distribution of the current-carrying conductor and the streamline distribution inside the air chamber; S8. Select temperature measurement points to measure the current-carrying conductor temperature and the gas chamber shell temperature during actual operation of the environmentally friendly gas switchgear; S9. Compare the measured data with the simulation results, evaluate the error of the simulation model in temperature prediction, and optimize and adjust the simulation model; S10. In the optimized and adjusted simulation model, input the operating current and environmental conditions, perform simulation, extract the hotspot temperature, construct a data sample including multi-dimensional input features and target output, and construct a data set, wherein the target output is the hotspot temperature; S11. Combining the thermal resistance characteristics of the internal structure of the switchgear, an adaptive weight distribution method based on the thermal resistance network is used to perform temperature inversion; S12. Build a deep neural network (DNN) model, introduce physical constraints into the loss function, and optimize the loss function using the constructed dataset. S13, allowing the DNN model to learn the mapping relationship between the air chamber shell temperature and the hot spot temperature, and obtaining a trained DNN model; S14. The air chamber shell temperature measured during actual operation is used as input, and the trained DNN model is used for prediction to obtain the corresponding hotspot temperature.

2. The hot spot temperature detection method for an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 1 is characterized in that: In the above-mentioned S1, the 3D drawing software adopts SOLIDWORKS, and the geometric model includes the current-carrying busbar, three-position switch, circuit breaker, heat sink, gas chamber shell of the environmentally friendly gas switch cabinet, and heat dissipation backpack. Among them, the current-carrying busbar, three-position switch and circuit breaker are current-carrying conductors. According to the actual structural parameters of the environmentally friendly gas switch cabinet, a high-precision 3D geometric model of the original size is constructed.

3. The hot spot temperature detection method for an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 2 is characterized in that: In S2, the material of the current-carrying busbar, three-position switch, and heat sink is set to copper, the material of the moving and static contacts of the circuit breaker is set to copper-chromium alloy, the material of the vacuum interrupter sleeve of the circuit breaker is set to alumina, and the material of the gas chamber shell and heat dissipation backpack is set to stainless steel. The setting method is to select from the material library.

4. The hot spot temperature detection method of an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 2 is characterized in that: In the aforementioned S3, the specific setting method of each physical field is as follows: S3.

1. For the current physical field, the current field calculation domain is selected as the current-carrying busbar, three-position switch, heat sink, and the moving and static contacts of the circuit breaker. A surface terminal is added to provide a current input surface, and a ground plane is added to provide a current outflow surface. The three-phase AC current is set to I0, I0×exp(+120[deg]×j), and I0×exp(-120[deg]×j), respectively. Where I0 is the effective current value of the environmental gas switchgear in actual operation, exp is an exponential function used to express the power of the base e of the natural logarithm, 120[deg] represents 120 degrees, which is the phase difference between every two phases in three-phase AC, and j is an imaginary unit used to represent a 90-degree phase difference in complex number representation. By adding contact impedance to the contact surface to simulate the contact resistance of the contacts and the bolt fastening surface, the contact heating during the operation of the environmental protection gas switch cabinet is simulated. The contacts are the moving and static contacts of the circuit breaker. The contact resistance calculation formula is: ； Where, K is a constant whose value is determined by the contact material; m is an index whose value is related to the contact form of the two contact surfaces; F is the contact pressure; S3.

2. For the solid and fluid heat transfer physical fields, the temperature field calculation domain selects all domains except the vacuum medium in the vacuum interrupter of the circuit breaker. The solid domain includes the current-carrying busbar, three-position switch, circuit breaker, heat sink and air chamber shell, and the fluid domain includes the air inside the air chamber. Heat flux is added to simulate the heat exchange between the air chamber shell and the outside air. The flux type selects convection heat flux-user-defined heat transfer coefficient. The heat transfer coefficient Calculated by the following formula: ； Where, is the outer wall surface temperature of the gas chamber shell; is the external air temperature of the air chamber shell; S3.

3. For the surface-to-surface radiation physics field, select the entire model domain as the radiation field calculation domain. Set the emitted radiation direction to be controlled by opacity. Set the air inside the gas chamber and the vacuum inside the vacuum interrupter of the circuit breaker as transparent domains, and set the rest as opaque domains. S3.

4. For the turbulent k-ε physics field, select the air domain inside the air chamber as the flow field calculation domain. Set the fluid domain physical model to compressible flow, include gravity, use reduced pressure, and have a Mach number Ma less than 0.

3. Set the wall condition to no slip, add a pressure point constraint, and select a point far away from the calculation domain as the pressure point.

5. The hot spot temperature detection method of an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 4 is characterized in that: In the above S5, when meshing the simulation model, the moving and static contacts of the circuit breaker are meshed using refined meshes, and the remaining parts, namely the current-carrying busbar, the three-position switch, the air chamber housing and the air domain inside the air chamber, are meshed using conventional meshes.

6. The hot spot temperature detection method of an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 1 is characterized in that: In the aforementioned S8, an experimental platform is built to perform actual measurements on the environmentally friendly gas switchgear. During the measurement, the temperature measurement points of the current-carrying conductor are selected as follows: based on the temperature distribution of the current-carrying conductor, the temperature measurement points cover the hotspot temperature area identified in the simulation. The temperature measurement points are arranged at positions that can be contacted by the temperature sensor under the test conditions and at the joints of the current-carrying conductors. The temperatures of the measurement points are collected using a contact temperature measurement method. The temperature measurement points of the air chamber shell are selected based on the analysis of the streamline distribution inside the air chamber. When the air reaches the vicinity of the air chamber shell, the projection on the shell is the path of heat transport, and the temperature measurement points are set along the path of heat transport.

7. The hot spot temperature detection method of an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 1 is characterized in that: In the above-mentioned S9, optimizing and adjusting the simulation model is to adjust the material parameters, boundary conditions and mesh density of the meshing based on the comparison between the measurement data and the simulation results; in S10, the multidimensional input features include operating current parameters, environmental condition parameters, air chamber shell temperature, material parameters, geometric structure parameters and boundary conditions of the simulation analysis; The hot spot temperature is the temperature of the moving and static contacts of the circuit breaker during the simulation process.

8. The hot spot temperature detection method of an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 1 is characterized in that: In the above S11, the adaptive weight distribution method based on the thermal resistance network is specifically to design a dynamic weight distribution formula to calculate the influence weight of each temperature measurement point on the hot spot temperature. ,Will As the weighting coefficient of the input feature of the DNN model, the dynamic weight allocation formula is: ； Where, is the thermal resistance from the temperature measuring point i to the hot spot; is the spatial distance from the temperature measurement point i to the hotspot; is the attenuation coefficient; exp is the exponential function; the hot spots are the moving and static contacts of the circuit breaker.

9. The hot spot temperature detection method of an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 1 is characterized in that: In the above-mentioned S12, the DNN model includes three hidden layers and one output layer. The first hidden layer is provided with 128 neurons and adopts the ReLU activation function; the second hidden layer is provided with 64 neurons; the third hidden layer is provided with 64 neurons; and the output layer is provided with 1 neuron to output the predicted hotspot temperature value. In the DNN model, the loss function L that introduces physical constraints is: ； Where, Indicates the predicted temperature and simulation temperature The mean square error of Representing the Laplace operator of the heat conduction equation The residual of 、 is a hyperparameter used to adjust data fitting and physical constraint weights.

10. The hot spot temperature detection method of an environmentally friendly gas switch cabinet based on multi-physics field simulation according to claim 1, characterized in that: In the above S13, during training, the weights of each layer of the DNN model are updated according to the gradient of the loss function through the back propagation algorithm. At the same time, a dynamic feedback mechanism is introduced in the training process. After each simulation model update, the parameters of the DNN model are calibrated using the latest actual measurement data.

Images