A fast calculation method for multi-physics fields of oil-immersed transformers

Abstract

The present invention discloses a fast calculation method for the multi-physics field of an oil-immersed transformer, which belongs to the field of transformer technology. The method comprises constructing a multi-physics field simulation analysis model for the transformer, establishing a three-dimensional equivalent digital model of the transformer, importing the model into simulation software, calculating the heat source density, performing multi-physics field simulation calculations of the transformer, generating a fast calculation data set for the transformer temperature field, inputting the data set into an XGBoost fast calculation model that considers spatial feature relationships, performing machine learning iterative training, and obtaining the transformer temperature field calculation results through load factor, heat source density, and ambient temperature. The present invention adopts the above-mentioned fast calculation method for the multi-physics field of an oil-immersed transformer, constructs an XGBoost model that considers spatial feature relationships in the calculation of the oil-immersed transformer temperature field, integrates the multi-physics field simulation results as input data sets into the XGBoost model, can efficiently analyze data, reduce the risk of overfitting in the model training process, and ensure the reliability of the prediction results.

Description

Technical Field

[0001] The present invention relates to the technical field of transformers, and in particular to a fast calculation method for multi-physical fields of oil-immersed transformers. Background Art

[0002] Winding temperature, a key monitoring parameter for evaluating the operating status of oil-immersed transformers, directly affects their operating efficiency and service life. Therefore, in-depth research into the distribution characteristics and evolution of transformer winding temperature fields is crucial for ensuring safe and stable operation. However, as a highly centralized and complex system, oil-immersed transformers possess complex dynamic relationships among multiple physical fields (magnetic, thermal, and fluid). Rapid and accurate calculation of transformer winding temperature fields has become a pressing challenge at the intersection of electrical engineering and computational science.

[0003] Currently, research methods for transformer temperature calculation primarily include sensor measurement, empirical formulas, thermal circuit modeling, and numerical simulation. While direct sensor measurement is accurate, it requires a large number of sensors, resulting in high costs and complex maintenance. Empirical formulas are simple to use, but their accuracy is limited and they fail to account for the transformer's structural characteristics. Thermal circuit models, based on the principle of hotspot analogy, calculate winding hotspots through circuit analysis. While they can reflect the heat transfer process, parameter settings are cumbersome. None of these methods can fully capture the transformer's internal temperature distribution, making numerical calculations the mainstream. Numerical calculations employ methods such as finite element and finite difference methods to solve the transformer's multi-physics coupling and provide a detailed analysis of temperature variations. However, the long computational cycles and high computing power requirements of multi-physics simulations limit their real-time application. To improve computational efficiency, researchers at home and abroad are exploring temperature prediction methods based on artificial intelligence algorithms. These methods, which do not require in-depth understanding of the transformer's complex structure, instead capture data characteristics and design well-designed models to achieve temperature prediction. Summary of the Invention

[0004] The purpose of this invention is to provide a fast calculation method for the multi-physical field of oil-immersed transformers, which integrates numerical calculation and machine learning technology to provide a powerful tool for transformer status monitoring and improve the digitalization and intelligence level of transformers.

[0005] To achieve the above object, the present invention provides a fast calculation method for multi-physical fields of an oil-immersed transformer, comprising the following steps:

[0006] S1. Analyze the coupling relationship between electromagnetic, fluid and temperature physical fields inside the transformer and clarify the interaction mechanism under the relationship mapping;

[0007] S2. Analyze the natural oil circulation process of the transformer. Based on the conservation equations of mass, momentum, and energy, construct a multi-physics simulation analysis model for the transformer's internal structure. Preset the boundary conditions of the transformer's oil flow velocity, gravitational acceleration, and initial temperature.

[0008] S3. Establish a three-dimensional equivalent digital model of the transformer and perform mesh division;

[0009] S4. Import the three-dimensional equivalent digital model with meshes divided in S3 into the simulation software and set parameters to simulate the actual operating conditions of the transformer. At the same time, calculate the operating losses under different working conditions using the load equivalent method and convert them into heat sources to be applied to the transformer.

[0010] S5. Calculate the ratio of the operating loss to the volume of the transformer structure as the heat source density of the transformer structure;

[0011] S6. Substitute the calculated data in S4 and S5 into the three-dimensional equivalent digital model of the transformer in S3, and perform multi-physics simulation calculation of the transformer according to the control equation in S2;

[0012] S7. Based on different working conditions, the load factor, heat source density, and ambient temperature are arranged and combined to form different initial input variables. The multi-physics field simulation calculation in S6 is repeated to generate temperature field simulation results under several working conditions, thereby forming a transformer temperature field fast calculation data set.

[0013] S8. Input the fast calculation data set generated by S7 into the XGBoost fast calculation model for iterative machine learning training. The trained model obtains the transformer temperature field calculation results through the load coefficient, heat source density and ambient temperature.

[0014] Preferably, in S2, the conservation equations of mass, momentum, and energy are:

[0015]

[0016] Where ρ represents density, t represents time, v represents fluid velocity vector, p represents fluid pressure, η represents dynamic viscosity, f represents external force, T represents temperature, k represents thermal conductivity, C p represents specific heat capacity, and Q represents heat source.

[0017] Preferably, in S4, the calculation formula for the operating loss is:

[0018]

[0019] Among them, P L and P N They represent load loss and no-load loss respectively, P represents total loss, I and I R Represents actual current and rated current respectively.

[0020] Preferably, in S5, the heat source density calculation formula is:

[0021] S = P / V (5);

[0022] S T =S0[1+β(T-T0)] (6);

[0023] Where S represents the heat source density, V represents the volume of the structural part, ST represents the heat source density at temperature T, S0 represents the heat source density at temperature T0, and β represents the temperature coefficient.

[0024] Preferably, in S7, the load factor is the ratio of the actual current to the rated current, i.e., I and I R The ratio.

[0025] Preferably, in S8, the XGBoost fast calculation model calculation process includes the following steps:

[0026] S81. A preset number of regression decision trees are used as the base learners of the XGBoost fast calculation model. The additive model is used for single model training. Each time a new decision tree is added to the model, the integrated model is as follows:

[0027]

[0028] in, represents the predicted value of the t-th decision tree model, f t (x i ) represents the structure of the t-th tree;

[0029] S82, substitute formula (7) into formula (8) and use it as the objective function, continuously modify the model, use the greedy algorithm to generate a tree for node splitting, select the optimal splitting benefit, and obtain the optimal XGBoost model.

[0030]

[0031] Among them, Obj represents the objective function, represents the loss function, Ω(f k ) represents the regularization term;

[0032] S83. Preset several position points on average for the transformer structural parts, and generate a preset number of XGBoost models at the same time. Use the temperature data of the position points as output variables and add them to the input variables of the next model. Finally, merge all the models to form an XGBoost fast calculation model that considers the spatial feature relationship. The output result of the model is the temperature field distribution result of the oil-immersed transformer.

[0033] Preferably, in S81 and S82, the input variables of the first XGBoost model formed are the load coefficient, heat source density and ambient temperature, and the output variable is the temperature of the first position point. The input variables of the second model are added to the output variables of the previous model, and so on. The input variables of the preset number of models are the load coefficient, heat source density, ambient temperature and temperature data of the preset number of position points minus 1, and the output variable is the temperature of the preset number of position points.

[0034] Therefore, the present invention adopts the above-mentioned fast calculation method for the multi-physical field of oil-immersed transformers. By constructing an oil-immersed transformer model with a multi-physical field coupling mechanism, the multi-physical field simulation calculation results of the oil-immersed transformer under complex working conditions based on the model are used as input data sets and integrated into the XGBoost machine learning prediction model. It can not only efficiently process and analyze large-scale data, but also significantly reduce the overfitting risk in the model training process through its unique gradient boosting mechanism and regularization strategy, ensuring the stability and reliability of the prediction results. In the process of constructing the XGBoost fast calculation model, the spatial feature relationship in the calculation of the oil-immersed transformer temperature field is fully considered, so that the model can more accurately reflect the dynamic change process of the internal temperature field of the transformer.

[0035] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Schematic diagram of the circulation process of natural oil circulation in an embodiment of a fast calculation method for multi-physical fields of an oil-immersed transformer according to the present invention;

[0037] Figure 2 It is a schematic diagram of an XGBoost algorithm flow chart of an embodiment of a fast calculation method for multi-physical fields of an oil-immersed transformer of the present invention. DETAILED DESCRIPTION

[0038] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0039] Unless otherwise defined, the technical or scientific terms used in the present invention shall have the usual meanings understood by persons of ordinary skill in the field to which the present invention belongs. The words "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Words such as "include" or "comprise" mean that the elements or objects preceding the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0040] Example 1

[0041] The present invention provides a fast calculation method for multi-physical fields of an oil-immersed transformer, comprising the following steps:

[0042] S1. Analyze the coupling relationship between the internal electromagnetic, fluid, temperature and other physical fields of the oil-immersed transformer during long-term service, and clarify the interaction mechanism under the mapping of these relationships.

[0043] S2. Analyze the natural oil circulation process, such as Figure 1 As shown in the figure, based on the conservation equations of mass, momentum, and energy, a multi-physics field simulation analysis model is constructed, and boundary conditions such as transformer oil flow velocity, gravitational acceleration, and initial temperature are preset.

[0044] The conservation equations of mass, momentum, and energy are:

[0045]

[0046] Where ρ represents density, t represents time, v represents fluid velocity vector, p represents fluid pressure, η represents dynamic viscosity, f represents external force, T represents temperature, k represents thermal conductivity, C p represents specific heat capacity, and Q represents heat source.

[0047] S3. Establish a three-dimensional equivalent digital model of the oil-immersed transformer and perform meshing. Reasonable meshing can better simulate the physical phenomena under the actual operating state of the oil-immersed transformer, while improving the accuracy and reliability of the calculation results in subsequent simulation analysis.

[0048] S4. Import the three-dimensional equivalent digital model with meshes divided in S3 into the simulation software and set parameters to simulate the actual operating conditions of the transformer. At the same time, calculate the operating losses under different working conditions using the load equivalent method and convert them into heat sources to be applied to the transformer. The calculation formula for operating losses is:

[0049]

[0050] Among them, P L and P N They represent load and no-load losses respectively, P represents total loss, I and I R Represents actual and rated current respectively.

[0051] S5. Calculate the ratio of operating loss to the volume of transformer structural parts and use it as the heat source density of the transformer structural parts. The heat source density calculation formula is:

[0052] S = P / V (5);

[0053] S T =S0[1+β(T-T0)] (6);

[0054] Among them, S represents the heat source density, V represents the volume of the structure, S T Represents the heat source density at temperature T, S0 represents the heat source density at temperature T0, and β represents the temperature coefficient. In particular, in the present invention, the winding heat source density is regarded as a function of temperature, and formula (6) is only used for the calculation of the winding heat source density.

[0055] S6. Substitute the calculated data in S4 and S5 into the three-dimensional equivalent digital model of the transformer in S3, and perform multi-physical field simulation calculations of the transformer according to the control equations in S2.

[0056] S7. According to different working conditions, the load factor, heat source density and ambient temperature are arranged and combined to form different initial input variables. The multi-physics field simulation calculation of S6 is repeated to generate temperature field simulation results under several working conditions and form a transformer temperature field fast calculation data set. The load factor is the ratio of the actual current to the rated current, that is, I and I R The ratio.

[0057] S8. Input the fast calculation data set generated in S7 into the XGBoost fast calculation model that considers spatial feature relationships, and perform iterative machine learning training. The trained model obtains the transformer temperature field calculation results based on the load factor, heat source density, and ambient temperature. The calculation process of the XGBoost fast calculation model includes the following steps:

[0058] S81. A preset number of regression decision trees are used as the base learners for the XGBoost fast calculation model. The preset number can be set based on experience. An additive model is used for single model training. Each time a new decision tree is added to the model, the integrated model is as follows:

[0059]

[0060] in, represents the predicted value of the t-th decision tree model, f t (x i ) represents the structure of the t-th tree.

[0061] S82, substitute formula (7) into formula (8) and use it as the objective function, continuously modify the model, use the greedy algorithm to generate a tree for node splitting, select the optimal splitting benefit, and obtain the optimal XGBoost model.

[0062]

[0063] Among them, Obj represents the objective function, represents the loss function, Ω(f k ) represents a regularization term, which is used to prevent the model from overfitting.

[0064] In S81 and S82, the input variables of the first XGBoost model formed are the load coefficient, heat source density and ambient temperature, and the output variable is the temperature of the first position point. The input variables of the second model are added to the output variables of the previous model, and so on. The input variables of the preset number of models are the load coefficient, heat source density, ambient temperature and temperature data of the preset number of positions minus 1, and the output variable is the temperature of the preset number of positions.

[0065] S83. Preset several position points on average for the transformer structure, and generate a preset number of XGBoost models at the same time. Use the temperature data of the position points as output variables and add them to the input variables of the next model. Finally, merge all the models to form an XGBoost fast calculation model that considers the spatial feature relationship. The output result of the model is the temperature field distribution result of the oil-immersed transformer. The algorithm flow is as follows: Figure 2 .

[0066] Therefore, the present invention adopts the above-mentioned fast calculation method for the multi-physical field of oil-immersed transformers. By constructing an oil-immersed transformer model with a multi-physical field coupling mechanism, the multi-physical field simulation calculation results of the oil-immersed transformer under complex working conditions based on the model are used as input data sets and integrated into the XGBoost machine learning prediction model. It can not only efficiently process and analyze large-scale data, but also significantly reduce the overfitting risk in the model training process through its unique gradient boosting mechanism and regularization strategy, ensuring the stability and reliability of the prediction results. In the process of constructing the XGBoost fast calculation model, the spatial feature relationship in the calculation of the oil-immersed transformer temperature field is fully considered, so that the model can more accurately reflect the dynamic change process of the internal temperature field of the transformer.

[0067] Finally, it should be noted that the above embodiments 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 preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A fast calculation method for multi-physics fields of oil-immersed transformers, characterized by: The following steps are involved: S1. Analyze the coupling relationship between electromagnetic, fluid and temperature physical fields inside the transformer and clarify the interaction mechanism under the relationship mapping; S2. Analyze the natural oil circulation process of the transformer. Based on the conservation equations of mass, momentum, and energy, construct a multi-physics simulation analysis model for the transformer's internal structure. Preset the boundary conditions of the transformer's oil flow velocity, gravitational acceleration, and initial temperature. S3. Establish a three-dimensional equivalent digital model of the transformer and perform mesh division; S4. Import the three-dimensional equivalent digital model with meshes divided in S3 into the simulation software and set parameters to simulate the actual operating conditions of the transformer. At the same time, calculate the operating losses under different working conditions using the load equivalent method and convert them into heat sources to be applied to the transformer. S5. Calculate the ratio of the operating loss to the volume of the transformer structure as the heat source density of the transformer structure; S6. Substitute the calculated data in S4 and S5 into the three-dimensional equivalent digital model of the transformer in S3, and perform multi-physics simulation calculation of the transformer according to the mass, momentum and energy conservation equations in S2; S7. Based on different working conditions, the load factor, heat source density, and ambient temperature are arranged and combined to form different initial input variables. The multi-physics field simulation calculation in S6 is repeated to generate temperature field simulation results under several working conditions, thereby forming a transformer temperature field fast calculation data set. S8. Input the fast calculation data set generated by S7 into the XGBoost fast calculation model for iterative machine learning training. The trained model obtains the transformer temperature field calculation results through the load coefficient, heat source density and ambient temperature.

2. The fast calculation method for multi-physical fields of an oil-immersed transformer according to claim 1, characterized in that: In S2, the conservation equations of mass, momentum, and energy are: (1)； (2)； (3)； in, represents density, Indicates time, represents the fluid velocity vector, represents the fluid pressure, represents the dynamic viscosity, represents the external force, Indicates temperature, represents the thermal conductivity, represents specific heat capacity, and Q represents heat source.

3. The fast calculation method for multi-physical fields of an oil-immersed transformer according to claim 1, characterized in that: In S4, the calculation formula for operating loss is: (4)； in, and Represent load loss and no-load loss respectively, represents the total loss, and Represents actual current and rated current respectively.

4. The fast calculation method for multi-physical fields of an oil-immersed transformer according to claim 3, characterized in that: In S5, the heat source density calculation formula is: (5)； (6)； Among them, S represents the heat source density, V represents the volume of the structure, S T represents the heat source density at temperature T, S0 represents the heat source density at temperature T0, Represents the temperature coefficient.

5. The fast calculation method for multi-physical fields of an oil-immersed transformer according to claim 3, characterized in that: In S7, the load factor is the ratio of the actual current to the rated current, that is, and The ratio.

6. The fast calculation method for multi-physical fields of an oil-immersed transformer according to claim 1, characterized in that: In S8, the XGBoost fast calculation model calculation process includes the following steps: S81. A preset number of regression decision trees are used as the base learners of the XGBoost fast calculation model. The additive model is used for single model training. Each time a new decision tree is added to the model, the integrated model is as follows: (7)； in, Indicates the t The predicted value of the decision tree model, f t (x i ) Indicates the t The structure of a tree; S82, substitute formula (7) into formula (8) and use it as the objective function, continuously modify the model, use the greedy algorithm to generate a tree for node splitting, select the optimal splitting benefit, and obtain the optimal XGBoost model. (8)； Among them, Obj represents the objective function, represents the loss function, represents the regularization term; S83. Preset several position points on average for the transformer structural parts, and generate a preset number of XGBoost models at the same time. Use the temperature data of the position points as output variables and add them to the input variables of the next model. Finally, merge all the models to form an XGBoost fast calculation model that considers the spatial feature relationship. The output result of the model is the temperature field distribution result of the oil-immersed transformer.

7. The fast calculation method for multi-physical fields of an oil-immersed transformer according to claim 6, characterized in that: In S81 and S82, the input variables of the first XGBoost model formed are the load coefficient, heat source density and ambient temperature, and the output variable is the temperature of the first position point. The input variables of the second model are added to the output variables of the previous model, and so on. The input variables of the preset number of models are the load coefficient, heat source density, ambient temperature and temperature data of the preset number of positions minus 1, and the output variable is the temperature of the preset number of positions.

Images