A method for detecting actual temperature rise of high-frequency magnetic component windings based on BP curve inversion

Abstract

The present invention relates to a method for detecting the actual temperature rise of a high-frequency magnetic element winding based on BP curve inversion, comprising the following steps: step S1: burying a temperature measuring component at a temperature measuring point inside a magnetic element, powering off after powering on to reach a stable temperature rise, collecting and recording temperature rise drop data after powering off; step S2: acquiring temperature rise data after the inversion time point according to a preset inversion time point, and inverting the data based on a BP neural network model to obtain the actual temperature rise of the temperature measuring point of the magnetic element winding. The present invention overcomes the additional temperature rise influence of metallic temperature rise sensors such as thermocouples and thermistors under a high-frequency magnetic field, and realizes accurate measurement of the hot spot temperature rise inside the magnetic element winding under a high-frequency magnetic field.

Description

Technical Field

[0001] The invention relates to the field of high-frequency magnetic component detection, and in particular to a method for detecting actual temperature rise of a high-frequency magnetic component winding based on BP curve inversion and reverse deduction. Background Art

[0002] The core and winding of high-frequency magnetic components will produce losses during operation, causing the temperature of the magnetic components to rise. Excessive temperature will affect the service life of high-frequency magnetic components, winding insulation, core magnetic permeability and safety of use. In order to ensure its stable and safe operation and meet safety requirements, it is necessary to measure the internal temperature rise of the running magnetic components. Thermocouples and thermistors are one of the commonly used methods. However, in the high-frequency magnetic field environment inside the magnetic components, the metal parts of thermocouples and thermistors will produce eddy current losses, causing additional temperature rise, resulting in distortion of temperature rise measurement, and this additional temperature rise is difficult to obtain accurately. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a method for detecting the actual temperature rise of high-frequency magnetic component windings based on BP curve inversion, so as to overcome the additional temperature rise influence of metallic temperature rise sensor devices such as thermocouples and thermistors under high-frequency magnetic fields, thereby realizing accurate measurement of the hot spot temperature rise inside the magnetic component windings under high-frequency magnetic fields.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A method for detecting the actual temperature rise of a high-frequency magnetic component winding based on BP curve inversion and reverse deduction, characterized in that it comprises the following steps:

[0006] Step S1: bury the temperature measuring component at the temperature measuring point inside the magnetic element, turn off the power after the power is turned on and the temperature rise reaches a stable level, and collect and record the temperature rise and drop data after the power is turned off;

[0007] Step S2: According to the preset reverse time point, the temperature rise data after the reverse time point is obtained, and the actual temperature rise of the temperature measuring point of the magnetic component winding is obtained by data inversion based on the BP neural network model.

[0008] Furthermore, the preset reverse time point is t seconds after power failure. .

[0009] Furthermore, the temperature measurement component includes a plurality of thermocouples or thermistors, which are evenly arranged at the winding gaps between the primary side and the secondary side for data collection.

[0010] Furthermore, the BP neural network input data is the temperature data measured by the thermocouple, and the output data is the data after inversion and inference using the neural network, as follows:

[0011] The data is normalized in the input layer and nonlinearly calculated in the hidden layer. The actual model is a nonlinear mapping, so a sigmoid activation function is superimposed after the hidden layer. The sigmoid activation function is , through mapping, the function output is a function that conforms to the trend of the cooling model ;

[0012] The corresponding output signal is generated from the output layer, and the error is obtained by comparing it with the actual output signal. Then, through the back propagation of the error, the weights and thresholds of the network are continuously adjusted until the output result is within the predetermined error range.

[0013] Compared with the prior art, the present invention has the following beneficial effects:

[0014] The present invention overcomes the additional temperature rise influence of metallic temperature rise sensor devices such as thermocouples and thermistors under high-frequency magnetic fields, and realizes accurate measurement of hot spot temperature rise inside magnetic element windings under high-frequency magnetic fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0016] Figure 2 It is a cooling heat circuit model in one embodiment of the present invention;

[0017] Figure 3 is the magnetic field intensity distribution of the winding window in one embodiment of the present invention;

[0018] Figure 4 It is the location where the hot spot temperature appears in one embodiment of the present invention;

[0019] Figure 5 It is a hotspot temperature acquisition code setting in one embodiment of the present invention;

[0020] Figure 6 This is a full flow chart of a BP neural network in one embodiment of the present invention;

[0021] Figure 7 It is a schematic diagram of winding temperature rise predicted by neural network data inversion in one embodiment of the present invention;

[0022] Figure 8 It is a data optimization process in one embodiment of the present invention;

[0023] Fig. 9 It is a BP neural network statement in one embodiment of the present invention. DETAILED DESCRIPTION

[0024] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0025] Please refer to Figure 1The present invention provides a method for detecting the actual temperature rise of a high-frequency magnetic component winding based on BP curve inversion, which is characterized by comprising the following steps:

[0026] Step S1: bury the temperature measuring component at the temperature measuring point inside the magnetic element, turn off the power after the power is turned on and the temperature rise reaches a stable level, and collect and record the temperature rise and drop data after the power is turned off;

[0027] Step S2: According to the preset reverse time point, the temperature rise data after the reverse time point is obtained, and the actual temperature rise of the temperature measuring point of the magnetic component winding is obtained by data inversion based on the BP neural network model.

[0028] In this embodiment, a thermocouple is used to measure the temperature rise of the high-frequency magnetic element winding. Before power failure, the high-frequency magnetic element winding has a winding loss heat source, and the thermocouple wire has a magnetic field induced eddy current loss heat source. After power failure, the thermocouple and the winding begin the heat dissipation and cooling process at the same time. The thermocouple dissipates heat and cools down to the surrounding environment (i.e., the winding measurement point), and the winding measurement point itself also dissipates heat and cools down to the external space. According to the heat transfer process, a simplified thermal circuit model equivalent to cooling is established, such as Figure 2 shown.

[0029] In the figure is the thermal resistance between the thermocouple and the winding temperature measurement point, is the thermal resistance between the winding temperature measurement point and the environment, is the winding temperature at the moment of power failure, is the winding temperature, is the thermocouple temperature at the moment of power failure, is the thermocouple temperature, is the heat capacity of the winding, the ambient temperature is the ambient temperature.

[0030] According to Ampere's loop law NI=HL, the magnetic field distribution on both sides of each conductor layer can be calculated. The magnetic field strength induced by the magnetic element between the primary and secondary sides is different at different positions. The point with the largest magnetic field strength has a large eddy current loss magnetic field strength. The outermost thermocouple is theoretically in a magnetic field strength of 0, such as Figure 3 Preferably, it is found that the hot spot of the magnetic element appears between the primary winding and the secondary winding, and five thermocouples are evenly distributed in the gap between the primary and secondary windings for data collection.

[0031] After data collection, all data are compared, and the data with the highest temperature rise are selected in the collection system, and then BP neural network training is performed.

[0032] Because the end of the thermocouple is small in size, the loss density is large, and the additional temperature rises, the thermocouple cools down quickly after power failure. However, as time goes by, the temperature difference between the thermocouple and the winding decreases, the thermocouple cools down slower, and the temperature rise measured by the thermocouple tends to the actual temperature of the winding temperature measurement point. For the cooling process curve after the magnetic element is powered off, the front end slope of the temperature drop curve is mainly due to the thermocouple being heated by the magnetic field, and the back end curve is mainly due to the winding temperature drop. Therefore, the back end data with a slower cooling rate after power failure can be used for data inversion and reverse deduction, and the temperature rise of the magnetic element winding temperature measurement point when the power is just off can be fitted and deduced, that is, the stable temperature rise when the winding is heated.

[0033] In this embodiment, preferably, reference Figure 6 , the input data of the BP neural network is the temperature data measured by the thermocouple, and the output data is the data after inversion using the neural network. The details are as follows: the data is normalized in the input layer and nonlinearly calculated in the hidden layer. Because the actual model is a nonlinear mapping, it is necessary to superimpose a sigmoid activation function after the hidden layer. The sigmoid activation function is , through mapping, the function output is a function that conforms to the trend of the cooling model The corresponding output signal is generated from the output layer, and the error is obtained by comparing it with the actual output signal. Then, the weights and thresholds of the network are continuously adjusted through the back propagation of the error. When the output result is within the predetermined error range, the network training ends.

[0034] The BP neural network is used to perform data inversion and forward push, and the temperature rise of the temperature measurement point of the magnetic component winding just after power failure is fitted and deduced, that is, the stable temperature rise of the winding when it is heated, such as Figure 7 shown.

[0035] Among them, the sampling time of the latter data has an impact on the data inversion. If the data is intercepted for a long time after the power failure, the data volume is small and the error is large. If the data is intercepted for a short time after the power failure, it is difficult to judge whether the thermocouple has completed the heat dissipation. Therefore, the key step in the inversion of the actual temperature of the winding is to optimize the measurement position, set the inversion point according to the actual situation, and improve the accuracy of the inversion. Therefore, on the basis of using the BP neural network, it is necessary to combine the heat dissipation of the thermocouple under the high-frequency magnetic component and the heat dissipation of the coil winding to set the data collection point.

[0036] Because the end of the thermocouple is small, the heat dissipation is faster, while the winding heat dissipation is slower, according to the experimental data prediction and the heat dissipation speed of the thermocouple to the surrounding environment, in this embodiment, preferably, the program time optimization range is set to The sampling interval of each point is 1 second, and on this basis, preferably, the slope of every three points is required to meet , ensure that the data segment is mainly based on the winding temperature drop, such as Figure 8 It helps to obtain faster convergence speed and more accurate data.

[0037] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A method for detecting the actual temperature rise of high-frequency magnetic component windings based on BP curve inversion, characterized in that: The following steps are involved: Step S1: bury the temperature measuring component at the temperature measuring point inside the magnetic element, turn off the power after the power is turned on to reach a stable temperature rise, and collect and record the temperature rise and drop data after the power is turned off; Step S2: according to the preset reverse time point, obtain the temperature rise data after the reverse time point, and perform data inversion based on the BP neural network model to obtain the actual temperature rise of the temperature measuring point of the magnetic element winding; The preset reverse time point is t seconds after power failure, 120s≤t≤360s; The temperature measurement component includes a plurality of thermocouples or thermistors, which are evenly arranged at the winding gaps between the primary side and the secondary side for data collection; The BP neural network input data is the temperature data measured by the thermocouple, and the output data is the data after inversion using the neural network, as follows: The data is normalized in the input layer and nonlinearly calculated in the hidden layer. The actual model is a nonlinear mapping, so a sigmoid activation function is superimposed after the hidden layer. The sigmoid activation function is Through mapping, the function output is made into a function that conforms to the trend of the cooling model The corresponding output signal is generated from the output layer, and the error is obtained by comparing it with the actual output signal. Then, through the back propagation of the error, the weights and thresholds of the network are continuously adjusted until the output result is within the predetermined error range.

2. A high-frequency magnetic component winding actual temperature rise detection system based on BP curve inversion and reverse deduction, characterized in that: The system is implemented by a method for detecting the actual temperature rise of a high-frequency magnetic component winding based on BP curve inversion as described in claim 1, comprising a plurality of thermocouples or thermistors, a temperature collector and a host computer; the thermocouples or thermistors are evenly arranged at the winding gap between the primary and secondary sides for data collection, and the data is transmitted to the host computer through the temperature collector.

Images