Old transformer hot-spot temperature prediction method based on multi-source heterogeneous data processing

Abstract

The invention discloses an old transformer hot-spot temperature prediction method based on multi-source heterogeneous data processing. The method comprises the following steps: step 1, establishing a transformer digital twinborn analysis model; 2, taking the transformer physical model as an object, obtaining optical fiber temperature measurement data, thermocouple temperature measurement data and an infrared temperature measurement picture, and proposing a heterogeneous data processing technology to carry out data extraction on the infrared temperature measurement picture; 3, finite element calculation is carried out through magnetic-thermal-fluid multi-physics field coupling, and correction is carried out through experimental data obtained in the step 2; 4, temperature data is obtained according to finite element calculation in the step 3 to serve as a data analysis basis, and then a random forest model is established for training and optimization; step 5, obtaining operation temperature data of the transformer; and step 6, performing prediction calculation on the hot spot temperature. The method has the advantages that compared with traditional hot spot temperature calculation in which a temperature field is simplified, the method can effectively avoid the inaccurate situation.

Description

Technical Field

[0001] The present invention belongs to the technical field of temperature measurement of electrical equipment, and in particular relates to a method for predicting hot spot temperature of an old transformer based on multi-source heterogeneous data processing. Background Art

[0002] The temperature of the transformer is one of the important operating parameters. The higher the temperature of the transformer, the faster the thermal aging process, resulting in reduced insulation performance of the transformer winding, thereby damaging the transformer coil; in addition, even if the insulation is not damaged, but the temperature rise is too high, the performance of the insulation material will deteriorate, and it is easy to be damaged by high voltage breakdown, resulting in transformer failure. Transformer temperature is a comprehensive issue. First of all, the loss of transformer operation is the source of transformer heating and temperature rise, so the temperature analysis of the transformer requires analysis and calculation of the loss. At present, the more mature transformer hot spot temperature research methods mainly include: transformer hot spot temperature monitoring based on temperature monitoring system, hot spot degree calculation based on empirical formula or numerical calculation method, and hot spot temperature prediction based on artificial intelligence algorithm. For old transformers with strong oil circulation, there is usually no optical fiber temperature measuring device, and only thermocouple temperature measuring device can obtain the top and bottom oil temperatures. Conventionally, the calculation model is used to simplify the thermal characteristics of the transformer, and the average of the top oil temperature and the bottom oil temperature is used as the average oil temperature to calculate and predict the transformer hot spot temperature. This method has low accuracy. Therefore, it is urgent to provide a more accurate power transformer hot spot prediction method. Summary of the invention

[0003] Purpose of the Invention

[0004] In order to solve the problem of low accuracy of hot spot temperature of old transformers in the prior art, the present invention provides a hot spot temperature prediction method for old transformers based on multi-source heterogeneous data processing. The average temperature of the transformer is predicted by combining discretized data processing of the temperature measuring device with an artificial intelligence algorithm, thereby making the hot spot temperature prediction more accurate and reliable than conventional calculation of the hot spot temperature.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for predicting hot spot temperature of old transformers based on multi-source heterogeneous data processing includes the following steps: Step 1: establishing a transformer digital twin analysis model;

[0007] Step 2: Taking the transformer physical model as the object, obtain the optical fiber temperature measurement data, thermocouple temperature measurement data and infrared temperature measurement pictures, and propose heterogeneous data processing technology to extract data from infrared temperature measurement pictures;

[0008] Step 3: Perform finite element calculations through magnetic-thermal-fluid multi-physics field coupling, and use the experimental data obtained in step 2 for correction;

[0009] Step 4: Obtain temperature data based on the finite element calculation in step 3 as a basis for data analysis, and then establish a random forest model for training and optimization;

[0010] Step 5: Use the data processed by the heterogeneous data processing technology in step 2 and the random forest model optimized in step 4 to obtain the transformer operating temperature data;

[0011] Step 6: Use the transformer operating temperature data obtained in step 5 and the temperature obtained by its own temperature measuring device to predict and calculate the hot spot temperature.

[0012] As a further description of the above scheme, the step 2 includes the following steps:

[0013] Step 2.1: Take the transformer physical model as the object and obtain the optical fiber temperature measurement data, thermocouple temperature measurement data and infrared temperature measurement pictures;

[0014] Step 2.2: Extract data from infrared temperature measurement images through heterogeneous data processing technology, that is, use an infrared thermometer to take infrared images of the transformer surface, including the temperature distribution of known temperature points and infrared images, and calibrate the positions of known temperature points, and record the values ​​and coordinates; Image processing: Convert the infrared temperature distribution map into a grayscale gradient map, divide the gradient map into a two-dimensional grid of 10*10, take 100 discrete points, number the valid discrete points, establish a discrete point coordinate system, ensure that each discrete point has its unique corresponding coordinates, and establish a linear gradient distribution of the subdivided element according to the divergence line on the discrete point and the two adjacent gradient lines where the discrete point is located, analyze the linear gradient relationship of the temperature of discrete points with different coordinates, and thus solve the temperature data of the discrete points;

[0015] As a further description of the above scheme, step 3 includes the following steps:

[0016] Step 3.1: When calculating the temperature field, the applied load mainly includes the following two aspects: heat generation rate per unit volume and ambient temperature. When analyzing the three-dimensional temperature field and fluid field of the oil-type transformer, its heat source is the core loss and high and low voltage winding loss.

[0017] Calculate the core loss P f Mainly includes hysteresis loss P h and eddy current loss P e , and its calculation formula is:

[0018]

[0019] Where J represents the eddy current density; J0 represents the correlation vector of J; J 0rmsrepresents the average value of J0 within one cycle; σ represents the conductivity; v represents the movement speed; ρ represents the material density; It represents the peak value of magnetic flux density of the i-th unit; represents the hysteresis loss of the i-th unit; N represents the number of finite element units; V (i) represents the unit volume;

[0020] Calculate the high and low voltage winding losses, the calculation formula is:

[0021]

[0022] Where P w is the winding loss; P k is the transformer short-circuit active power; β is the load factor; S js Calculate the load for the transformer; S e is the rated capacity of the transformer;

[0023] Step 3.2: The heat conduction control equation is a three-dimensional steady-state heat conduction control equation for an isotropic medium with an internal heat source. In the Cartesian coordinate system, the equation is written as:

[0024]

[0025] In the formula, T represents temperature; q represents heat flow vector; λ represents thermal conductivity; Represents the three-dimensional distance differential element of the fluid-solid coupling surface and the adiabatic surface;

[0026] Step 3.3: In the fluid domain, heat transfer is controlled by the energy transport equation. The energy conservation equation is:

[0027]

[0028] Among them, S h is the internal heat source of the fluid; Φ is the part of mechanical energy converted into thermal energy due to viscosity, which is the dissipation function; ρ is the fluid density; μ is the fluid dynamic viscosity; v is the fluid internal energy; w is the fluid heat transfer coefficient; T is the fluid temperature; p is the fluid pressure; λ is the second viscosity coefficient; S i is the heat source; p·▽U is the work done by the surface force on the fluid microelement; h is the fluid specific gravity;

[0029] Step 3.4: Correct the finite element simulation calculations of steps 3.1-3.3 using the physical transformer model operation data obtained in step 2.1, and finally obtain a reliable finite element calculation model;

[0030] Step 3.5: Using the transformer underload operation mode, overload operation mode, normal load operation mode and ambient temperature as state variables, finite element calculation is performed on the transformer to obtain the temperature data of the average oil temperature under different conditions;

[0031] As a further description of the above scheme, step 4 includes the following steps:

[0032] Step 4.1: Take the finite element calculation results in step 3.5 as the sample set;

[0033] Step 4.2: Use the random forest algorithm to predict the average temperature of transformer oil. In order to eliminate the influence of different dimensions of data of different characteristic indicators, the initial data is normalized and preprocessed. The initial data is normalized to the interval [0, 1]. The calculation expression after data normalization is:

[0034]

[0035] In the formula, x i is the temperature data of the segmentation element; x min is the minimum value of the temperature data of the segmented element; x max is the maximum value of the temperature data of the segmented element;

[0036] Step 4.3: After preprocessing the sample data, a random forest model is established and trained, and then adjusted and optimized according to the error curve to obtain a tuned random forest model.

[0037] As a further description of the above scheme, the transformer operating temperature data in step 5 is obtained by processing the infrared temperature measurement image using the heterogeneous data processing technology proposed in step 2.2, and then inputting it into the tuned random forest model in step 4.3.

[0038] As a further description of the above scheme, the prediction calculation of the hot spot temperature in step 6 is calculated as follows:

[0039]

[0040] In the formula, θ h is the hot spot temperature of the winding; θ b is the bottom oil temperature, directly obtained by the transformer thermocouple; θ a is the top oil temperature, directly obtained by the transformer thermocouple; θ om is the average oil temperature, which is input by the infrared thermometer and output by the random forest prediction model; θ o The ambient temperature is directly obtained by the temperature measuring device; K L is the load factor, i.e. the ratio of load current to rated current; w H is the exponential power of the current on the winding temperature rise; gris the temperature difference between the hot spot and the oil on the top of the winding; K T is the top oil temperature coefficient, K B is the bottom oil temperature coefficient, which is calculated based on the thermal characteristic data provided by the transformer manufacturer and is transformer model specific; w , H gr The value of is obtained through the transformer thermal characteristics data provided by the transformer manufacturer.

[0041] Advantages and effects of the present invention:

[0042] The present invention proposes a hot spot temperature prediction method for old transformers based on multi-source heterogeneous data processing. It is a temperature measurement method for old power transformers. The method obtains a transformer hot spot temperature prediction model by introducing a digital twin model and an artificial intelligence algorithm. The average temperature of the transformer oil can be obtained by using the prediction model and the data obtained by instrument temperature measurement as input. The hot spot temperature can be calculated using the formula. Compared with the simplification of the temperature field in the traditional hot spot temperature calculation, this method can effectively avoid inaccurate situations, prevent further structural failures or abnormal operation, and can greatly promote the safe and stable operation of the power system and reduce accidents and losses caused by excessive transformer temperature. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Finite element simulation analysis diagram of an embodiment of the present invention;

[0044] Figure 2 Average oil temperature prediction structure diagram based on random forest algorithm in an embodiment of the present invention;

[0045] Figure 3 The number of sample data decision trees and error curves of the embodiment of the present invention;

[0046] Figure 4 Heterogeneous discretization processing diagram of an infrared thermometer according to an embodiment of the present invention;

[0047] Figure 5 Experimental diagram of a thermocouple temperature measuring instrument according to an embodiment of the present invention;

[0048] Figure 6 Experimental diagram of the optical fiber thermometer according to an embodiment of the present invention;

[0049] Figure 7 Flowchart of an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] A method for predicting hot spot temperature of old transformers based on multi-source heterogeneous data processing includes the following steps: Step 1: establishing a transformer digital twin analysis model;

[0052] Step 2: Taking the transformer physical model as the object, obtain the optical fiber temperature measurement data, thermocouple temperature measurement data and infrared temperature measurement pictures, and propose heterogeneous data processing technology to extract data from infrared temperature measurement pictures;

[0053] Step 3: Perform finite element calculations through magnetic-thermal-fluid multi-physics field coupling, and use the experimental data obtained in step 2 for correction;

[0054] Step 4: Obtain temperature data based on the finite element calculation in step 3 as a basis for data analysis, and then establish a random forest model for training and optimization;

[0055] Step 5: Use the data processed by the heterogeneous data processing technology in step 2 and the random forest model optimized in step 4 to obtain the transformer operating temperature data;

[0056] Step 6: Use the transformer operating temperature data obtained in step 5 and the temperature obtained by its own temperature measuring device to predict and calculate the hot spot temperature.

[0057] Step 2 of the present invention comprises the following steps:

[0058] Step 2.1: Take the transformer physical model as the object and obtain the optical fiber temperature measurement data, thermocouple temperature measurement data and infrared temperature measurement pictures;

[0059] Step 2.2: Extract data from infrared temperature measurement images through heterogeneous data processing technology, that is, use an infrared thermometer to take infrared images of the transformer surface, including the temperature distribution of known temperature points and infrared images, and calibrate the positions of known temperature points, and record the values ​​and coordinates; Image processing: Convert the infrared temperature distribution map into a grayscale gradient map, divide the gradient map into a two-dimensional grid of 10*10, take 100 discrete points, number the valid discrete points, establish a discrete point coordinate system, ensure that each discrete point has its unique corresponding coordinates, and establish a linear gradient distribution of the subdivided element according to the divergence line on the discrete point and the two adjacent gradient lines where the discrete point is located, analyze the linear gradient relationship of the temperature of discrete points with different coordinates, and thus solve the temperature data of the discrete points;

[0060] Step 3 of the present invention comprises the following steps:

[0061] Step 3.1: When calculating the temperature field, the applied load mainly includes the following two aspects: heat generation rate per unit volume and ambient temperature. When analyzing the three-dimensional temperature field and fluid field of the oil-type transformer, its heat source is the core loss and high and low voltage winding loss.

[0062] Calculate the core loss P f Mainly includes hysteresis loss P h and eddy current loss P e , and its calculation formula is:

[0063]

[0064] Where J represents the eddy current density; J0 represents the correlation vector of J; J 0rms represents the average value of J0 within one cycle; σ represents the conductivity; v represents the movement speed; ρ represents the material density; It represents the peak value of magnetic flux density of the i-th unit; represents the hysteresis loss of the i-th unit; N represents the number of finite element units; V (i) represents the unit volume;

[0065] Calculate the high and low voltage winding losses, the calculation formula is:

[0066]

[0067] Where P w is the winding loss; P k is the transformer short-circuit active power; β is the load factor; S js Calculate the load for the transformer; S e is the rated capacity of the transformer;

[0068] Step 3.2: The heat conduction control equation is a three-dimensional steady-state heat conduction control equation for an isotropic medium with an internal heat source. In the Cartesian coordinate system, the equation is written as:

[0069]

[0070] In the formula, T represents temperature; q represents heat flow vector; λ represents thermal conductivity; Represents the three-dimensional distance differential element of the fluid-solid coupling surface and the adiabatic surface;

[0071] Step 3.3: In the fluid domain, heat transfer is controlled by the energy transport equation. The energy conservation equation with the fluid specific enthalpy h and temperature T as variables is:

[0072]

[0073] Among them, S h is the internal heat source of the fluid; Φ is the part of mechanical energy converted into thermal energy due to viscosity, which is the dissipation function; ρ is the fluid density; μ is the fluid dynamic viscosity; v is the fluid internal energy; w is the fluid heat transfer coefficient; T is the fluid temperature; p is the fluid pressure; λ is the second viscosity coefficient; S i is the heat source; p·▽U is the work done by the surface force on the fluid microelement; h is the fluid specific gravity;

[0074] Step 3.4: Use the physical transformer model operation data obtained in step 2.1 to correct the finite element simulation calculations in steps 3.1-3.3, and finally obtain a reliable finite element calculation model;

[0075] Step 3.5: Use the transformer underload operation mode, overload operation mode, normal load operation mode and ambient temperature as state variables to perform finite element calculation on the transformer to obtain average oil temperature data under different conditions.

[0076] Step 4 of the present invention comprises the following steps:

[0077] Step 4.1: Take the finite element calculation results in step 3.5 as the sample set;

[0078] Step 4.2: Use the random forest algorithm to predict the average temperature of transformer oil. In order to eliminate the influence of different dimensions of data of different characteristic indicators, the initial data is normalized and preprocessed. The initial data is normalized to the interval [0, 1]. The calculation expression after data normalization is:

[0079]

[0080] In the formula, x i is the temperature data of the segmentation element; x min is the minimum value of the temperature data of the segmented element; x max is the maximum value of the temperature data of the segmented element;

[0081] Step 4.3: After preprocessing the sample data, a random forest model is established and trained, and then adjusted and optimized according to the error curve to obtain a tuned random forest model.

[0082] Step 5 of the embodiment of the present invention includes the following steps:

[0083] The heterogeneous data processing technology proposed in step 2.2 is used to process the infrared temperature measurement images and input into the tuned random forest model in step 4.3 to predict the average temperature data of the transformer oil temperature.

[0084] The prediction calculation of the hot spot temperature in step 6 of the embodiment of the present invention is calculated by the following formula:

[0085]

[0086] In the formula, θ h is the hot spot temperature of the winding; θ b is the bottom oil temperature, directly obtained by the transformer thermocouple; θ a is the top oil temperature, directly obtained by the transformer thermocouple; θ om is the average oil temperature, which is input by the infrared thermometer and output by the random forest prediction model; θ o The ambient temperature is directly obtained by the temperature measuring device; K L is the load factor, i.e. the ratio of load current to rated current; w H is the exponential power of the current on the winding temperature rise; gr is the temperature difference between the hot spot and the oil on the top of the winding; K T is the top oil temperature coefficient, K B is the bottom oil temperature coefficient, which is calculated based on the thermal characteristic data provided by the transformer manufacturer and is transformer model specific; w , H gr The value of is obtained through the transformer thermal characteristics data provided by the transformer manufacturer.

[0087] Specifically, step 4 of the present invention comprises the following steps:

[0088] Step 4.1: Take the finite element calculation results in step 3.5 as the sample set;

[0089] Step 4.2: Take 10 independent elements with a load rate of 60% to 150% as the first variable; change the ambient temperature and take 10 independent elements with a load rate of 21°C to 30°C as the second variable. Then obtain 100 working condition temperature data through finite element calculation. Divide the transformer oil tank into 16 parts and the transformer radiator into 8 parts. Use the characteristic temperature values ​​of the oil tank segmentation element and the radiator segmentation element of the finite element calculation results as the sample set input, and use the average temperature of the transformer oil of the finite element calculation results as the sample set output to establish a training sample set, starting with the oil tank segmentation element from element 1 to element 16, and the radiator segmentation element from element 17 to element 24. Some sample set data are as follows:

[0090]

[0091] In order to eliminate the influence of different dimensions of data of different characteristic indicators, the initial data is normalized and preprocessed to unify the variable dimension, reduce the difficulty of model training, and prevent the model accuracy from decreasing or the network from failing to converge. The initial data is normalized to the interval [0, 1], and the calculation expression after data normalization is:

[0092]

[0093] Step 4.3: Establish a random forest prediction model, write it in a programming language, and establish 100 tree models. Figure 2 As shown in the figure. To obtain more accurate values ​​for the random forest prediction model, model tuning should be performed (see the attached figure). Figure 4 As shown in the figure, firstly, the number of decision trees and leaf trees are selected to optimize the model. According to the error curve as a reference, the number of decision trees above 200 can meet the requirements. Considering the operation speed and the computing power of the operation platform, 300 is used as the number of decision trees; according to the error curve as a reference, the number of leaves is set to 4 according to the curve distribution trend.

[0094] like Figure 1 As shown in the figure, the finite element simulation analysis of the transformer temperature field can obtain the temperature distribution trend and temperature data of the transformer oil tank shell and the radiator. The temperature data obtained by finite element calculation is used as the basis for data analysis and to generate a sample data set.

[0095] like Figure 3 As shown in the figure, the number of decision trees and the error curve of the sample data are plotted. By plotting the relationship between the number of decision trees and the error, the trade-off between model complexity and overfitting can be intuitively displayed. As the number of decision trees increases, the training error usually decreases, but the validation error may rise after a certain point, indicating the occurrence of overfitting. Analyzing this curve helps to select the appropriate number of decision trees, thereby optimizing the generalization ability of the model. Figure 4 As shown in the figure, the infrared image of the fuel tank surface taken by the infrared thermometer is converted into a grayscale gradient map. The gradient map is divided into a two-dimensional grid of 10*10, and 100 discrete points are taken. The valid discrete points are numbered and a discrete point coordinate system is established. According to the divergence line on the discrete point and the two adjacent gradient lines where the discrete point is located, the linear gradient distribution of the subdivision element is established, and the linear gradient relationship of the temperature of discrete points with different coordinates is analyzed to solve the temperature data of the discrete points.

[0096] As attached Figure 5 The figure shows the experimental diagram of the thermocouple temperature measuring instrument. The top oil temperature and the bottom oil temperature are directly obtained through the temperature measuring device of the transformer itself.

[0097] As attached Figure 6 As shown in the figure, the temperature obtained by optical fiber temperature measurement is used as the accurate reference value of the hot spot temperature. The hot spot prediction temperature is calculated by taking the average of the top oil temperature and the bottom oil temperature as the average oil temperature in the traditional calculation method; the hot spot temperature is calculated by using the average oil temperature obtained by heterogeneous data processing in the present invention. By comparison, it can be concluded that the hot spot temperature prediction proposed by the present invention is more accurate and reliable.

[0098] The specific prediction and use process of the present invention is as follows:

[0099] Step 1: Use an infrared thermometer to take an infrared image of the transformer surface, including the temperature distribution of the known temperature points and the infrared image, and calibrate the position of the known temperature points, and record the values ​​and coordinates; Image processing: Convert the infrared temperature distribution map into a grayscale gradient map, divide the gradient map into a two-dimensional grid of 10*10, take 100 discrete points, number the valid discrete points, establish a discrete point coordinate system, ensure that each discrete point has its unique corresponding coordinates, and establish a linear gradient distribution of the subdivided element according to the divergence line on the discrete point and the two adjacent gradient lines where the discrete point is located, analyze the linear gradient relationship of the temperature of discrete points with different coordinates, and thus solve the temperature data of the discrete points;

[0100] Step 2: After processing the infrared temperature measurement heterogeneous data in step 1, the discrete point temperature data is obtained and input into the random forest prediction model as input, so as to predict the average temperature data of the transformer oil temperature.

[0101] Step 3: Directly obtain the top oil temperature and the bottom oil temperature through the transformer's own temperature measuring device;

[0102] Step 4: Calculate the hot spot temperature of the transformer. The calculation formula is:

[0103]

[0104] In the formula, θ h is the hot spot temperature of the winding; θ b is the bottom oil temperature, directly obtained by the transformer thermocouple; θ a is the top oil temperature, directly obtained by the transformer thermocouple; θ om is the average oil temperature, which is input by the infrared thermometer and output by the random forest prediction model; θ o The ambient temperature is directly obtained by the temperature measuring device; K L is the load factor, i.e. the ratio of load current to rated current; w H is the exponential power of the current on the winding temperature rise; gr is the temperature difference between the hot spot and the oil on the top of the winding; K T is the top oil temperature coefficient, K B is the bottom oil temperature coefficient, which is calculated based on the thermal characteristic data provided by the transformer manufacturer and is transformer model specific; w , H gr The value of can be obtained through the transformer thermal characteristics data provided by the transformer manufacturer.

Claims

1. A method for predicting hot spot temperature of old transformers based on multi-source heterogeneous data processing, characterized in that: The following steps are involved: Step 1: Establish a transformer digital twin analysis model; Step 2: Taking the transformer physical model as the object, obtain the optical fiber temperature measurement data, thermocouple temperature measurement data and infrared temperature measurement pictures, and extract data from the infrared temperature measurement pictures through heterogeneous data processing technology; Step 3: Perform finite element calculations through magnetic-thermal-fluid multi-physics field coupling, and use the experimental data obtained in step 2 for correction; Step 4: Obtain temperature data based on the finite element calculation in step 3 as a basis for data analysis, and then establish a random forest model for training and optimization; Step 5: Use the data processed by the heterogeneous data processing technology in step 2 and the random forest model optimized in step 4 to obtain the transformer operating temperature data; Step 6: Use the transformer operating temperature data obtained in step 5 and the temperature obtained by its own temperature measuring device to predict and calculate the hot spot temperature.

2. The method for predicting hot spot temperature of old transformers based on multi-source heterogeneous data processing according to claim 1 is characterized in that: The step 2 comprises the following steps: Step 2.1: Take the transformer physical model as the object and obtain the optical fiber temperature measurement data, thermocouple temperature measurement data and infrared temperature measurement pictures; Step 2.2: Extract data from infrared temperature measurement images through heterogeneous data processing technology, that is, use an infrared thermometer to take infrared images of the transformer surface, including the temperature distribution of known temperature points and infrared images, and calibrate the positions of known temperature points, and record the values ​​and coordinates; Image processing: Convert the infrared temperature distribution map into a grayscale gradient map, divide the gradient map into a two-dimensional grid of 10*10, take 100 discrete points, number the valid discrete points, and establish a discrete point coordinate system to ensure that each discrete point has its unique corresponding coordinates. According to the divergence line on the discrete point and the two adjacent gradient lines where the discrete point is located, establish the linear gradient distribution of the subdivision element, analyze the linear gradient relationship of the temperature of discrete points with different coordinates, and thus solve the temperature data of the discrete points.

3. The method for predicting hot spot temperature of old transformers based on multi-source heterogeneous data processing according to claim 2 is characterized in that: The step 3 comprises the following steps: Step 3.1: When calculating the temperature field, the applied load mainly includes the following two aspects: heat generation rate per unit volume and ambient temperature. When analyzing the three-dimensional temperature field and fluid field of the oil-type transformer, its heat source is the core loss and high and low voltage winding loss. Calculate the core loss P f Mainly includes hysteresis loss P h and eddy current loss P e , and its calculation formula is: Where J represents the eddy current density; J0 represents the correlation vector of J; J 0rms represents the average value of J0 within one cycle; σ represents the conductivity; v represents the movement speed; ρ represents the material density; It represents the peak value of magnetic flux density of the i-th unit; represents the hysteresis loss of the i-th unit; N represents the number of finite element units; V (i) represents the unit volume; Calculate the high and low voltage winding losses, the calculation formula is: Where P w is the winding loss; P k is the transformer short-circuit active power; β is the load factor; S js Calculate the load for the transformer; S e is the rated capacity of the transformer; Step 3.2: The heat conduction control equation is a three-dimensional steady-state heat conduction control equation for an isotropic medium with an internal heat source. In the Cartesian coordinate system, the equation is written as: In the formula, T represents temperature; q represents heat flow vector; λ represents thermal conductivity; Represents the three-dimensional distance differential element of the fluid-solid coupling surface and the adiabatic surface; Step 3.3: In the fluid domain, heat transfer is controlled by the energy transport equation. The energy conservation equation is: Among them, S h is the internal heat source of the fluid; Φ is the part of mechanical energy converted into thermal energy due to viscosity, which is the dissipation function; ρ is the fluid density; μ is the fluid dynamic viscosity; v is the fluid internal energy; w is the fluid heat transfer coefficient; T is the fluid temperature; p is the fluid pressure; λ is the second viscosity coefficient; S i As heat source; is the work done by the surface force on the fluid microelement; h is the fluid specific gravity; Step 3.4: Correct the finite element simulation calculations of steps 3.1-3.3 using the physical transformer model operation data obtained in step 2.1, and finally obtain a reliable finite element calculation model; Step 3.5: Using the transformer underload operation mode, overload operation mode, normal load operation mode and ambient temperature as state variables, perform finite element calculation on the transformer to obtain the temperature data of the average oil temperature under different conditions.

4. The method for predicting hot spot temperature of old transformers based on multi-source heterogeneous data processing according to claim 1 is characterized in that: The step 4 comprises the following steps: Step 4.1: Take the finite element calculation results in step 3.5 as the sample set; Step 4.2: Use the random forest algorithm to predict the average temperature of transformer oil. In order to eliminate the influence of different dimensions of data of different characteristic indicators, the initial data is normalized and preprocessed. The initial data is normalized to the interval [0, 1]. The calculation expression after data normalization is: In the formula, x i is the temperature data of the segmentation element; x min is the minimum value of the temperature data of the segmented element; x max is the maximum value of the temperature data of the segmented element; Step 4.3: After preprocessing the sample data, a random forest model is established and trained, and then adjusted and optimized according to the error curve to obtain a tuned random forest model.

5. The method for predicting hot spot temperature of old transformers based on multi-source heterogeneous data processing according to claim 2 is characterized in that: The transformer operating temperature data in step 5 is obtained by processing the infrared temperature measurement image using the heterogeneous data processing technology proposed in step 2.2, and then inputting it into the tuned random forest model in step 4.

3.

6. The multi-source heterogeneous hotspot prediction method for old power transformers according to claim 1 is characterized in that: The prediction calculation of the hot spot temperature in step 6 is calculated as follows: In the formula, θ h is the hot spot temperature of the winding; θ b is the bottom oil temperature, directly obtained by the transformer thermocouple; θ a is the top oil temperature, directly obtained by the transformer thermocouple; θ om is the average oil temperature, which is input by the infrared thermometer and output by the random forest prediction model; θ o The ambient temperature is directly obtained by the temperature measuring device; K L is the load factor, i.e. the ratio of load current to rated current; w H is the exponential power of the current on the winding temperature rise; gr is the temperature difference between the hot spot and the oil on the top of the winding; K T is the top oil temperature coefficient, K B is the bottom oil temperature coefficient, which is calculated based on the thermal characteristic data provided by the transformer manufacturer and is transformer model specific; w , H gr The value of is obtained through the transformer thermal characteristics data provided by the transformer manufacturer.

Images