Transformer hot-spot temperature evaluation method and device based on physical information neural network

Abstract

The invention provides a transformer hot-spot temperature evaluation method and device based on a physical information neural network, and relates to the technical field of state evaluation of high-voltage power equipment. The method comprises the following steps: constructing a thermal simulation model of the transformer, wherein the thermal simulation model comprises a geometric model and a physical model; changing the environment temperature, the copper loss and the heat dissipation rate in the physical model, simulating to obtain the bottom oil temperature, the top oil temperature and the hot spot temperature under different operation conditions based on the thermal simulation model, and forming a training set; based on the training set, the environment temperature, the copper loss, the heat dissipation rate, the bottom layer oil temperature and the top layer oil temperature serve as input, the hot spot temperature serves as output, a physical information neural network model is trained through a target loss function containing physical information loss items, and a transformer hot spot temperature evaluation model is obtained; and evaluating the hot-spot temperature of the target transformer based on the transformer hot-spot temperature evaluation model. According to the invention, accurate evaluation of the hot-spot temperature of the transformer under the constraint of the thermal simulation model can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of high-voltage power equipment status assessment, and in particular to a transformer hot spot temperature assessment method and device based on a physical information neural network. Background Art

[0002] Oil-immersed transformers are prone to overheating failures when they are overloaded, have magnetic leakage, or leak oil. Evaluating the hot spot temperature of the transformer helps ensure its safe operation.

[0003] However, existing online detection methods cannot directly measure hotspot temperatures in transformer windings. Instead, they typically assess transformer overheating based on top and bottom oil temperatures. Guidelines such as GB / T 15164-94 and IEEE Std C57.91-2011 use simplified thermal models to calculate hotspot temperatures based on top oil temperature, load conditions, and ambient temperature. However, these calculations require calibration and correction for different operating conditions, which can lead to significant deviations.

[0004] For example, Chinese patent CN202211741890.2, published on June 23, 2023, discloses a method for optimizing the generation of a transformer winding index. The method obtains and constructs a transformer thermal simulation model based on the structural component relationship of the target transformer, performs dynamic simulation of a preset load type based on the transformer thermal simulation model, and records simulation data. Simulation is performed by setting several preset load rates under each preset load type. When the temperature rise of the winding temperature within a preset time is less than a preset value, the simulation is terminated. The temperature rise value of the hotspot temperature relative to the top oil temperature at each preset load rate is calculated based on the simulation data, and the temperature rise value is linearly regressed with the winding index recommended value of the IEC guideline according to a preset formula to generate an optimized value of the transformer winding index. Although this method simulates several preset load rates under each preset load type of the transformer, when the load type and load rate differ greatly from the preset load type and preset load rate, there is still a large error, and accurate assessment of the hotspot temperature of the transformer cannot be achieved. Summary of the Invention

[0005] The embodiments of the present invention provide a transformer hot spot temperature assessment method and device based on a physical information neural network to solve the current problem of difficulty in accurately assessing the transformer hot spot temperature under different operating conditions.

[0006] In a first aspect, an embodiment of the present invention provides a transformer hotspot temperature assessment method based on a physical information neural network, comprising:

[0007] Constructing a thermal simulation model of the transformer, wherein the thermal simulation model includes a geometric model determined by the geometric dimensions of the transformer and a physical model based on a heat conduction-diffusion equation, a heat source term, and thermal boundary conditions;

[0008] Changing the ambient temperature, copper loss, and heat dissipation rate in the physical model, and simulating the bottom oil temperature, top oil temperature, and hotspot temperature under different operating conditions based on the thermal simulation model to form a training set;

[0009] Based on the training set, taking the ambient temperature, the copper loss, the heat dissipation rate, the bottom oil temperature, and the top oil temperature as inputs, and the hotspot temperature as output, a physical information neural network model is trained using a target loss function including a physical information loss term to obtain a transformer hotspot temperature assessment model;

[0010] The hot spot temperature of the target transformer is evaluated based on the transformer hot spot temperature evaluation model.

[0011] In one possible implementation, the heat conduction-diffusion equation is:

[0012]

[0013] Where T is the temperature, u is the fluid velocity of the transformer oil, Q is the heat source, ρ is the material density, c p is the specific heat capacity of the material, and k is the thermal diffusion coefficient.

[0014] In a possible implementation, the heat source item includes the iron loss and copper loss of the transformer, the iron loss is a preset fixed value, and the copper loss is determined according to the load state of the transformer.

[0015] In a possible implementation, the thermal boundary conditions include ambient temperature and heat dissipation rate;

[0016] The ambient temperature is the boundary temperature at a preset distance from the transformer housing;

[0017] The heat dissipation rate includes a natural heat dissipation rate and a cooling heat dissipation rate at the bottom of the transformer. The natural heat dissipation rate is a preset fixed heat dissipation rate, and the cooling heat dissipation rate is a set value.

[0018] In a possible implementation, the target loss function includes a prediction deviation term and the physical information loss term;

[0019] The prediction deviation term is determined according to the deviation between the hotspot temperature prediction value and the actual value of the hotspot temperature of the physical information neural network model;

[0020] The physical information loss term is determined by the deviation between the temperature calculated according to the guideline method and the true value of the hot spot temperature.

[0021] In one possible implementation, the objective loss function is:

[0022]

[0023] Wherein, L is the target loss function, L1 is the prediction deviation term, L2 is the physical information loss term, w1 and w2 are the weights used to weigh the prediction deviation term and the physical information loss term, y pred is the hotspot temperature prediction value of the physical information neural network model, y0 is the actual value of the hotspot temperature, and y phy Calculate the temperature for the guideline method.

[0024] In one possible implementation, the temperature calculation formula of the guideline method is:

[0025]

[0026] Among them, y phy =θ ht The temperature is calculated by the guideline method, θ amb is the ambient temperature, Δθ top is the top oil temperature rise, R is the ratio of load loss to no-load loss, K is the load factor, x is the oil temperature rise index, H g is the temperature rise of the hot spot to the top oil, and y is the winding temperature rise index.

[0027] In one possible implementation, the target loss function including the physical information loss term is used to train the physical information neural network model to obtain the transformer hot spot temperature assessment model, including:

[0028] Training a physical information neural network model using a target loss function including a physical information loss term, and calculating an average relative error between a hotspot temperature prediction value and a true hotspot temperature value of the physical information neural network model;

[0029] When the average relative error meets a preset condition, a transformer hot spot temperature evaluation model is obtained.

[0030] In a second aspect, an embodiment of the present invention provides a transformer hotspot temperature assessment device based on a physical information neural network, comprising:

[0031] A model building module is used to build a thermal simulation model of the transformer, wherein the thermal simulation model includes a geometric model determined by the geometric dimensions of the transformer and a physical model based on a heat conduction-diffusion equation, a heat source term, and thermal boundary conditions;

[0032] A training data acquisition module is used to change the ambient temperature, copper loss and heat dissipation rate in the physical model, and simulate the bottom oil temperature, top oil temperature and hot spot temperature under different operating conditions based on the thermal simulation model to form a training set;

[0033] an evaluation model training module, configured to train a physical information neural network model based on the training set, with the ambient temperature, the copper loss, the heat dissipation rate, the bottom oil temperature, and the top oil temperature as inputs, and the hotspot temperature as output, using a target loss function including a physical information loss term, to obtain a transformer hotspot temperature evaluation model;

[0034] The hotspot temperature evaluation module is used to evaluate the hotspot temperature of the target transformer based on the transformer hotspot temperature evaluation model.

[0035] In a third aspect, an embodiment of the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method in the first aspect or any possible implementation of the first aspect is implemented.

[0036] In an embodiment of the present invention, a thermal simulation model of a transformer is first constructed. The thermal simulation model includes a geometric model determined by the geometric dimensions of the transformer and a physical model based on the heat conduction-diffusion equation, a heat source term, and thermal boundary conditions. By varying the ambient temperature, copper loss, and heat dissipation rate in the physical model, the bottom oil temperature, top oil temperature, and hotspot temperature under different operating conditions are simulated based on the thermal simulation model to form a training set. Furthermore, based on the training set, a physical information neural network model is trained using a target loss function including a physical information loss term, with the ambient temperature, copper loss, heat dissipation rate, bottom oil temperature, and top oil temperature as inputs and the hotspot temperature as output, to obtain a transformer hotspot temperature assessment model. The hotspot temperature of a target transformer is assessed based on the transformer hotspot temperature assessment model. Thus, by constructing a thermal simulation model consisting of a geometric model and a physical model based on the heat conduction-diffusion equation, a heat source term, and thermal boundary conditions, and training the physical information neural network model using a target loss function including a physical information loss term, the computational bias of the guideline method is introduced into the neural network loss function to optimize the model, thereby preventing overfitting of the neural network and ultimately achieving accurate transformer hotspot temperature assessment under the constraints of the thermal simulation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of an implementation method for transformer hot spot temperature assessment based on a physical information neural network provided by an embodiment of the present invention;

[0038] Figure 2 1 is a schematic diagram of a simulation model of a transformer provided in an embodiment of the present invention;

[0039] Figure 3 Schematic diagram of the architecture of the physical information neural network model provided by an embodiment of the present invention;

[0040] Figure 4is a training error evaluation graph provided by an embodiment of the present invention;

[0041] Figure 5 This is a timing diagram of simulation results of bottom oil temperature, top oil temperature and hotspot temperature provided by an embodiment of the present invention;

[0042] Figure 6 This is a diagram of transformer hot spot temperature assessment results provided by an embodiment of the present invention;

[0043] Figure 7 1 is a schematic structural diagram of a transformer hot spot temperature assessment device based on a physical information neural network according to an embodiment of the present invention;

[0044] Figure 8 is a schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0046] See also Figure 1 , which shows a flow chart of the implementation of the transformer hot spot temperature assessment method based on physical information neural network provided by an embodiment of the present invention, which is detailed as follows:

[0047] Step 101 : constructing a thermal simulation model of a transformer. The thermal simulation model includes a geometric model determined by the geometric dimensions of the transformer and a physical model based on a heat conduction-diffusion equation, a heat source term, and thermal boundary conditions.

[0048] For example, the geometric model of the transformer can be designed based on the actual size of the transformer, for example, a simple geometric model of the transformer can be designed based on the geometric size of the transformer (model: 100kVA / 5kV), such as Figure 2 As shown, the length, width, and height of the transformer cavity can be 1.3m, 0.7m, and 1.2m, respectively. The diameter and height of a single winding can be 0.3m and 0.8m, respectively. The three windings are centered, 0.15m from the left and right sidewalls, 0.2m from the front and back sidewalls, and 0.2m from the top and bottom. The minimum mesh size for the simulation can be set to 40mm.

[0049] For example, the physical model of the transformer may include heat source terms, thermal boundary conditions, and heat conduction-diffusion equations.

[0050] Among them, the heat source term corresponds to Figure 2 The middle winding area may include the iron loss and copper loss of the transformer. The iron loss may be a preset fixed value, such as 500W, and the copper loss may be determined according to the load state of the transformer.

[0051] Thermal boundary conditions can include ambient temperature and heat dissipation. The ambient temperature can be the boundary temperature at a preset distance from the transformer housing, for example, by setting the boundary temperature 1 meter from the transformer housing. The heat dissipation process can include natural air heat dissipation and accelerated heat dissipation by a cooling device, which can be achieved by setting different heat dissipation rates.

[0052] The heat dissipation rate may include a natural heat dissipation rate and a cooling heat dissipation rate at the bottom of the transformer. The natural heat dissipation rate may be a preset fixed heat dissipation rate, and the cooling heat dissipation rate may be a set value.

[0053] For example, the natural heat dissipation rate above and around the transformer casing is 200W / (m 2 K), the cooling rate of the bottom is adjustable.

[0054] The heat conduction-diffusion equation can be:

[0055]

[0056] Where T is the temperature, u is the fluid velocity of the transformer oil. For example, the fluid velocity of the transformer oil can be set to 0.05 m / s, Q is the heat source corresponding to the winding area, ρ is the material density, and c p is the material specific heat capacity, k is the thermal diffusivity, and the material properties of the winding, insulating oil and air can be shown in Table 1.

[0057] Table 1 Simulation material parameters

[0058]

[0059]

[0060] Step 102 , the ambient temperature, copper loss, and heat dissipation rate in the physical model are changed, and based on the thermal simulation model, the bottom oil temperature, top oil temperature, and hot spot temperature under different operating conditions are simulated to form a training set.

[0061] In this embodiment, considering that the online monitoring system of the transformer cannot directly obtain the hot spot temperature of the winding, a transformer thermal simulation model is first constructed to obtain state quantities such as the top oil temperature and the winding hot spot temperature under different operating conditions, and then a database (i.e., a training set) is constructed.

[0062] The database contains the bottom oil temperature, top oil temperature, ambient temperature, heat dissipation rate, copper loss, and hotspot temperature under different operating conditions. Different operating conditions correspond to different loads (i.e., different copper losses), different heat dissipation rates, and different ambient temperatures.

[0063] For example, when constructing a database under different operating conditions, the corresponding conditions for different operating conditions include ambient temperature (0-40°C, step size 5°C), copper loss (1000-2000W, step size 200W) and cooling rate (200-1000W / (m 2 K), step size 200W / (m 2 K)). The simulation output is the timing results of the bottom oil temperature, top oil temperature and winding hot spot temperature under different simulation conditions.

[0064] Step 103 , based on the training set, with ambient temperature, copper loss, heat dissipation rate, bottom oil temperature and top oil temperature as inputs and hotspot temperature as output, a physical information neural network model is trained using a target loss function including a physical information loss term to obtain a transformer hotspot temperature assessment model.

[0065] Optionally, the objective loss function may include a prediction bias term and a physical information loss term.

[0066] The prediction deviation term is determined according to the deviation between the hotspot temperature prediction value of the physical information neural network model and the actual value of the hotspot temperature.

[0067] The physical information loss term is determined by the deviation between the temperature calculated by the guideline method and the true value of the hotspot temperature.

[0068] For example, the objective loss function can be:

[0069]

[0070] Among them, L is the target loss function, L1 is the prediction bias term, L2 is the physical information loss term, w1 and w2 are the weights used to weigh the prediction bias term and the physical information loss term, y pred is the hotspot temperature prediction value of the physical information neural network model, y0 is the actual value of the hotspot temperature, and y phy Calculate the temperature using the guideline method.

[0071] For example, the formula for calculating the temperature using the guideline method may be:

[0072]

[0073] Among them, y phy =θ ht The temperature is calculated by the guide method, θ amb is the ambient temperature, Δθ top is the top oil temperature rise, R is the ratio of load loss to no-load loss, K is the load factor, x is the oil temperature rise index, H g is the temperature rise of the hot spot to the top oil, and y is the winding temperature rise index.

[0074] Optionally, a physical information neural network model is trained using a target loss function including a physical information loss term to obtain a transformer hot spot temperature assessment model, which may include:

[0075] The physical information neural network model is trained using an objective loss function that includes a physical information loss term. The average relative error between the hotspot temperature predictions and the true hotspot temperature values is calculated. When the average relative error meets a preset condition, the transformer hotspot temperature assessment model is obtained.

[0076] Among them, such as Figure 3 As shown, the physical information neural network model can be constructed using a multilayer perceptron neural network. The multilayer perceptron can have two hidden layers, each with 20 neurons. The learning rate can be 0.001, and the dropout rate can be 0.01. The input variables are a vector consisting of the ambient temperature, bottom oil temperature, top oil temperature, cooling rate, and copper loss. The output variable is the hotspot temperature. The objective loss function considers the calculation bias of the guidance method to improve the generalization ability of the neural network. For example, w1 can be 0.6 and w2 can be 0.4.

[0077] The temperature calculated by the guidance method is obtained according to GB / T 15164-94. The calculation formula for the hot spot temperature under the natural oil circulation cooling mode of the transformer is:

[0078] During the model training process, the prediction effect of the trained physical information neural network can be evaluated based on the average relative error (MAPE) of the database to dynamically adjust the weight coefficient of the neural network and achieve model optimization. The calculation formula for the average relative error (MAPE) is: where y i is the true value of the hotspot temperature, is the hotspot temperature prediction value of the trained physical information neural network model, and n is the number of samples.

[0079] For example, Figure 4 As shown, when training based on the database, the optimization reaches stability after 20 training steps.

[0080] In this embodiment, by designing a physical information neural network model and introducing the calculation deviation of the guidance method into the objective loss function, accurate evaluation of the transformer hot spot temperature under the constraints of the physical mechanism is achieved, which solves the defects of the transformer hot spot temperature evaluation method, such as large deviation and insufficient generalization ability. It can make up for the shortcomings of the transformer hot spot temperature evaluation method and provide a new strategy for the status evaluation of power equipment.

[0081] Step 104 : Evaluate the hotspot temperature of the target transformer based on the transformer hotspot temperature evaluation model.

[0082] For example, the following simulation conditions are set as an example: ambient temperature 20°C, copper loss 1300W, natural heat dissipation rate 5W / (m 2 K) and cooling rate 100W / (m 2 K). The bottom oil temperature, top oil temperature and hot spot temperature in the simulation results are as follows Figure 5 Based on the guideline method, the hot spot temperature calculation formula under the natural oil circulation cooling mode of the transformer is: where y phy =θ ht is the temperature calculated by the guideline method (i.e. the hot spot temperature calculated by the guideline method), θ amb is the ambient temperature, set to 20℃, Δθ top is the top oil temperature rise, R is the ratio of load loss to no-load loss, set to 5, K is the load factor, set to 1.05, x is the oil temperature rise index, set to 0.9, H g is the temperature rise of the hot spot to the top oil layer, set to 23°C, y is the winding temperature rise index, set to 1.6. Based on the physical information neural network model optimized in step 103 (i.e., the transformer hot spot temperature evaluation model), the ambient temperature of 20°C, the bottom oil temperature, the top oil temperature, and the cooling rate of 100W / (m 2 The hotspot temperature is obtained by taking the input vector consisting of 1300W of copper loss and 1300W of copper loss. Figure 6 The prediction results of this embodiment and the guideline method are presented. The prediction results of this embodiment are between the simulation results and the guideline method, indicating that this embodiment can achieve accurate evaluation of the transformer hot spot temperature under the constraints of the physical mechanism.

[0083] The method provided by an embodiment of the present invention first constructs a transformer thermal simulation model to obtain state variables such as bottom oil temperature, top oil temperature, and winding hotspot temperature under different operating conditions, thereby building a training database. Next, a physical information neural network is designed, and the calculation bias of the guideline method is introduced into the objective loss function. This method accurately assesses the transformer hotspot temperature under the constraints of physical mechanisms, ensures hotspot temperature assessment accuracy, prevents overfitting of the neural network, and effectively improves the generalization capability of the assessment method.

[0084] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0085] The following are device embodiments of the present invention. For details not fully described therein, reference may be made to the corresponding method embodiments described above.

[0086] Figure 7The following is a schematic diagram of the structure of a transformer hot spot temperature assessment device based on a physical information neural network according to an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown, which are described in detail as follows:

[0087] like Figure 7 As shown, the transformer hot spot temperature evaluation device based on physical information neural network includes: a model construction module 71, a training data acquisition module 72, an evaluation model training module 73 and a hot spot temperature evaluation module 74.

[0088] A model building module 71 is used to build a thermal simulation model of the transformer, wherein the thermal simulation model includes a geometric model determined by the geometric dimensions of the transformer and a physical model based on a heat conduction-diffusion equation, a heat source term, and thermal boundary conditions;

[0089] A training data acquisition module 72 is configured to change the ambient temperature, copper loss, and heat dissipation rate in the physical model, and simulate the bottom oil temperature, top oil temperature, and hotspot temperature under different operating conditions based on the thermal simulation model to form a training set;

[0090] An evaluation model training module 73 is configured to train a physical information neural network model based on the training set, using the ambient temperature, the copper loss, the heat dissipation rate, the bottom oil temperature, and the top oil temperature as inputs, and the hotspot temperature as output, using a target loss function including a physical information loss term, to obtain a transformer hotspot temperature evaluation model;

[0091] The hotspot temperature evaluation module 74 is configured to evaluate the hotspot temperature of the target transformer based on the transformer hotspot temperature evaluation model.

[0092] In one possible implementation, the heat conduction-diffusion equation is:

[0093]

[0094] Where T is the temperature, u is the fluid velocity of the transformer oil, Q is the heat source, ρ is the material density, c p is the specific heat capacity of the material, and k is the thermal diffusion coefficient.

[0095] In a possible implementation, the heat source item includes the iron loss and copper loss of the transformer, the iron loss is a preset fixed value, and the copper loss is determined according to the load state of the transformer.

[0096] In one possible implementation, the thermal boundary conditions include ambient temperature and heat dissipation rate; the ambient temperature is the boundary temperature at a preset distance from the transformer casing; the heat dissipation rate includes a natural heat dissipation rate and a cooling heat dissipation rate at the bottom of the transformer, the natural heat dissipation rate is a preset fixed heat dissipation rate, and the cooling heat dissipation rate is a set value.

[0097] In one possible implementation, the objective loss function includes a prediction deviation term and a physical information loss term; the prediction deviation term is determined based on the deviation between the hotspot temperature prediction value of the physical information neural network model and the actual value of the hotspot temperature; the physical information loss term is determined based on the deviation between the temperature calculated by the guideline method and the actual value of the hotspot temperature.

[0098] In one possible implementation, the objective loss function is:

[0099]

[0100] Wherein, L is the target loss function, L1 is the prediction deviation term, L2 is the physical information loss term, w1 and w2 are the weights used to weigh the prediction deviation term and the physical information loss term, y pred is the hotspot temperature prediction value of the physical information neural network model, y0 is the actual value of the hotspot temperature, and y phy Calculate the temperature for the guideline method.

[0101] In one possible implementation, the temperature calculation formula of the guideline method is:

[0102]

[0103] Among them, y phy =θ ht The temperature is calculated by the guideline method, θ amb is the ambient temperature, Δθ top is the top oil temperature rise, R is the ratio of load loss to no-load loss, K is the load factor, x is the oil temperature rise index, H g is the temperature rise of the hot spot to the top oil, and y is the winding temperature rise index.

[0104] In one possible implementation, the evaluation model training module 73 can be used to train a physical information neural network model using a target loss function that includes a physical information loss term, and calculate the average relative error between the hotspot temperature prediction value and the actual hotspot temperature value of the physical information neural network model; when the average relative error meets a preset condition, a transformer hotspot temperature evaluation model is obtained.

[0105] Figure 8 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 8As shown, the electronic device 8 of this embodiment includes a processor 80 and a memory 81. The memory 81 stores a computer program 82. When the processor 80 executes the computer program 82, the steps of the above-described method embodiments are implemented. Alternatively, when the processor 80 executes the computer program 82, the functions of the modules / units in the above-described device embodiments are implemented.

[0106] For example, the computer program 82 may be divided into one or more modules / units, which are stored in the memory 81 and executed by the processor 80 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program 82 in the electronic device 8.

[0107] The electronic device 8 may include, but is not limited to, a processor 80 and a memory 81. Those skilled in the art will appreciate that Figure 8 It is only an example of the electronic device 8 and does not constitute a limitation of the electronic device 8. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 8 may also include input and output devices, network access devices, buses, etc.

[0108] For the sake of convenience and brevity, the division of the above functional modules / units is only used as an example. In actual applications, the above functions can be assigned to different functional modules / units as needed. The above modules / units can be implemented in the form of hardware, software, or a combination of hardware and software.

[0109] In the above embodiments, the descriptions of each embodiment have their own focus. For parts not described or recorded in detail in one embodiment, please refer to the relevant descriptions of other embodiments. Unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other. The technical features of different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0110] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A transformer hot spot temperature assessment method based on physical information neural network, characterized in that: include: Constructing a thermal simulation model of the transformer, wherein the thermal simulation model includes a geometric model determined by the geometric dimensions of the transformer and a physical model based on a heat conduction-diffusion equation, a heat source term, and thermal boundary conditions; Changing the ambient temperature, copper loss, and heat dissipation rate in the physical model, and simulating the bottom oil temperature, top oil temperature, and hotspot temperature under different operating conditions based on the thermal simulation model to form a training set; Based on the training set, taking the ambient temperature, the copper loss, the heat dissipation rate, the bottom oil temperature, and the top oil temperature as inputs, and the hotspot temperature as output, a physical information neural network model is trained using a target loss function including a physical information loss term to obtain a transformer hotspot temperature assessment model; The hot spot temperature of the target transformer is evaluated based on the transformer hot spot temperature evaluation model.

2. The transformer hot spot temperature assessment method based on physical information neural network according to claim 1 is characterized in that: The heat conduction-diffusion equation is: Where T is the temperature, u is the fluid velocity of the transformer oil, Q is the heat source, ρ is the material density, c p is the specific heat capacity of the material, and k is the thermal diffusion coefficient.

3. The transformer hot spot temperature assessment method based on physical information neural network according to claim 1 is characterized in that: The heat source item includes the iron loss and copper loss of the transformer, the iron loss is a preset fixed value, and the copper loss is determined according to the load state of the transformer.

4. The transformer hot spot temperature assessment method based on physical information neural network according to claim 1 is characterized in that: The thermal boundary conditions include ambient temperature and heat dissipation rate; The ambient temperature is the boundary temperature at a preset distance from the transformer housing; The heat dissipation rate includes a natural heat dissipation rate and a cooling heat dissipation rate at the bottom of the transformer. The natural heat dissipation rate is a preset fixed heat dissipation rate, and the cooling heat dissipation rate is a set value.

5. The transformer hot spot temperature assessment method based on physical information neural network according to claim 1 is characterized in that: The target loss function includes a prediction deviation term and a physical information loss term; The prediction deviation term is determined according to the deviation between the hotspot temperature prediction value and the actual value of the hotspot temperature of the physical information neural network model; The physical information loss term is determined by the deviation between the temperature calculated according to the guideline method and the true value of the hot spot temperature.

6. The transformer hot spot temperature assessment method based on physical information neural network according to claim 5 is characterized in that: The objective loss function is: Wherein, L is the target loss function, L1 is the prediction deviation term, L2 is the physical information loss term, w1 and w2 are the weights used to weigh the prediction deviation term and the physical information loss term, y pred is the hotspot temperature prediction value of the physical information neural network model, y0 is the actual value of the hotspot temperature, and y phy Calculate the temperature for the guideline method.

7. The transformer hot spot temperature assessment method based on physical information neural network according to claim 5 is characterized in that: The calculation formula for temperature calculated by the guideline method is: Among them, y phy =θ ht The temperature is calculated by the guideline method, θ amb is the ambient temperature, Δθ top is the top oil temperature rise, R is the ratio of load loss to no-load loss, K is the load factor, x is the oil temperature rise index, H g is the temperature rise of the hot spot to the top oil, and y is the winding temperature rise index.

8. The transformer hot spot temperature assessment method based on physical information neural network according to claim 1 is characterized in that: The target loss function including the physical information loss term is used to train the physical information neural network model to obtain the transformer hot spot temperature assessment model, including: Training a physical information neural network model using a target loss function including a physical information loss term, and calculating an average relative error between a hotspot temperature prediction value and a true hotspot temperature value of the physical information neural network model; When the average relative error meets a preset condition, a transformer hot spot temperature evaluation model is obtained.

9. A transformer hot spot temperature assessment device based on physical information neural network, characterized in that: include: A model building module is used to build a thermal simulation model of the transformer, wherein the thermal simulation model includes a geometric model determined by the geometric dimensions of the transformer and a physical model based on a heat conduction-diffusion equation, a heat source term, and thermal boundary conditions; A training data acquisition module is used to change the ambient temperature, copper loss and heat dissipation rate in the physical model, and simulate the bottom oil temperature, top oil temperature and hot spot temperature under different operating conditions based on the thermal simulation model to form a training set; an evaluation model training module, configured to train a physical information neural network model based on the training set, with the ambient temperature, the copper loss, the heat dissipation rate, the bottom oil temperature, and the top oil temperature as inputs, and the hotspot temperature as output, using a target loss function including a physical information loss term, to obtain a transformer hotspot temperature evaluation model; The hotspot temperature evaluation module is used to evaluate the hotspot temperature of the target transformer based on the transformer hotspot temperature evaluation model.

10. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images