A high-frequency transformer thermal management method and system based on data-driven technology and multi-modal data fusion, a storage medium and an equipment
By integrating data-driven technology and multimodal data fusion, and combining heterogeneous data from multiple sources such as electrical, thermal, and environmental data, accurate prediction of hot spot temperature and active thermal management of high-frequency transformers are achieved. This solves the problem of insufficient accuracy and reliability of thermal management in existing technologies, and improves thermal management efficiency and transformer safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID JIANGSU ELECTRIC POWER CO XUZHOU POWER SUPPLY CO
- Filing Date
- 2026-04-02
- Publication Date
- 2026-07-03
AI Technical Summary
Existing technologies struggle to effectively integrate heterogeneous data from multiple sources, resulting in insufficient accuracy and reliability in the thermal management of high-frequency transformers. The thermal management strategies lack initiative and cannot effectively prevent hotspot temperatures from becoming too high.
By employing data-driven technology and multimodal data fusion methods, hotspot temperatures are predicted through multimodal data acquisition, preprocessing, and intelligent algorithms. Based on the prediction results, proactive thermal management strategies are implemented, and precise regulation is achieved by combining heat dissipation modules and control modules.
This improves the accuracy of hot spot temperature prediction in high-frequency transformers and the reliability of thermal management decisions, enabling a shift from passive to active response and enhancing thermal management efficiency and transformer operation safety.
Smart Images
Figure CN122333341A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of transformer technology, specifically to a high-frequency transformer thermal management method, system storage medium, and device based on data-driven technology and multimodal data fusion. Background Technology
[0002] A high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion.
[0003] Therefore, we propose a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion to address the problems mentioned above. Summary of the Invention
[0004] The purpose of this invention is to provide a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion, so as to solve the problems currently in the market mentioned in the background.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion, comprising:
[0006] S1 Data Acquisition: Acquires operating data of the high-frequency transformer through a multi-modal data acquisition module;
[0007] S2 Data Processing: The multimodal data collected by S1 is input as data features into the data-driven module for data preprocessing;
[0008] S3 Data Comparison: The host computer control module receives the preprocessed data output by the data-driven module in S2, predicts the hot spot temperature of the high-frequency transformer through its integrated intelligent algorithm, and compares it with the set expected temperature to give a thermal management strategy.
[0009] S4 thermal management strategy execution: The heat dissipation module executes the heat dissipation strategy output by the host computer, and the drive control cabinet controls the high-frequency transformer to execute the drive strategy output by the host computer;
[0010] S5 Data Feedback: The operating data of the high-frequency transformer is collected again by S1 to determine whether it meets expectations. If it does not meet expectations, S3 is executed to re-formulate the thermal management strategy.
[0011] Preferably, the data acquired in the S1 data acquisition of the high-frequency transformer includes: electrical parameters, environmental parameters, and temperature parameters.
[0012] Preferably, the data preprocessing in S2 is as follows:
[0013] S21 Data Synchronization: Synchronizes the sampling time of multi-source data output from different sensors;
[0014] S22 Data Cleaning: Noise removal, missing value imputation, and outlier detection of raw data;
[0015] S23 Data Fusion: Integrates electrical, environmental, and temperature data through a multi-source data fusion algorithm;
[0016] Among them, the fusion method is feature fusion or decision fusion;
[0017] S24 Data Feature Extraction: Extract data features reflecting the thermal state and potential thermal risks of the transformer.
[0018] Preferably, the sampling time synchronization in S21 data synchronization is as follows:
[0019] By stretching or compressing to reach the same time axis range, a cumulative cost matrix is constructed using a dynamic time warping algorithm. The minimum path for aligning time series is then calculated, and the cost matrix is expressed in the following formula:
[0020] D(i,j) = C(i,j) + min{D(i-1,j), D(i,j-1), D(i-1,j-1)}
[0021] In the formula, D is the current cumulative cost, and C is the local cost at each step.
[0022] Preferably, the specific steps for filling missing values in the S22 data cleaning process are as follows:
[0023] Interpolation algorithms are used for filling; common interpolation methods include linear interpolation and cubic spline interpolation.
[0024] The formula for constructing a linear polynomial using linear interpolation is as follows:
[0025] y=ax+b
[0026] The formula for constructing a cubic polynomial for cubic spline interpolation is as follows:
[0027] y=ax 3 +bx 2 +cx+d.
[0028] Preferably, outlier detection in S22 data cleaning is as follows:
[0029] The multimodal data acquisition module calculates the actual acquired signal value by detecting the duration of the high-level signal;
[0030] By calculating the mean and standard deviation of the fragment data sequence, outliers in the collected data are detected, removed, and missing values are filled.
[0031] Preferably, the intelligent algorithm used in the S3 data comparison is either MLP or KAN, two different neural network architectures.
[0032] Preferably, the thermal management strategy executed by the S4 thermal management strategy is as follows:
[0033] The heat dissipation strategy and drive strategy are customized based on the preprocessed electrical parameters, environmental parameters and temperature parameters output by the data drive module;
[0034] The customized logic includes: maintaining hot spot temperatures below safe thresholds, optimizing heat dissipation and power consumption to improve system efficiency, and extending the lifespan of high-frequency transformers;
[0035] The temperature data from the first two time points is collected using a PID control algorithm. The power of the heat dissipation module is adjusted by calculating the difference between the two data points. This ensures that the temperature rise fluctuation |T1-T0| within each data collection period is less than the set value ∆T, thus realizing customized logic.
[0036] A system for executing the high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion, comprising:
[0037] The multimodal data acquisition module is used to acquire the multimodal operating data of the high-frequency transformer;
[0038] The data-driven module is used to efficiently fuse, clean, synchronize, and extract features from heterogeneous multimodal data acquired from the feature input multimodal data acquisition module;
[0039] The host computer control module receives preprocessed data output from the data drive module, predicts the hot spot temperature of the high-frequency transformer based on intelligent algorithms, compares it with the set expected temperature, and provides a thermal management strategy.
[0040] The heat dissipation module is used to implement the heat dissipation strategy of the high-frequency transformer based on the heat management strategy given by the host computer control module.
[0041] The drive control cabinet control is used to execute the drive strategy given by the host computer control module for thermal management.
[0042] Preferably, the heat dissipation module includes a fan and / or a liquid cooling system.
[0043] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion.
[0044] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion.
[0045] Compared with the prior art, the beneficial effects of the present invention are:
[0046] This invention integrates multi-source heterogeneous data, including electrical, thermal, and environmental data, through multi-modal data fusion to provide more comprehensive and accurate information on the operating status of high-frequency transformers, thereby improving the accuracy of hotspot temperature prediction and the reliability of thermal management decisions.
[0047] An active thermal management strategy is proposed. Based on the prediction results, the system can regulate the heat dissipation device, transforming from a passive response to an active response. This effectively avoids excessively high hot spot temperatures, improves thermal management efficiency, and enhances the safety of transformer operation. Attached Figure Description
[0048] The accompanying drawings, as part of this invention, are provided to further illustrate the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention, but do not constitute an undue limitation thereof. Clearly, the drawings described below are merely some embodiments, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.
[0049] Figure 1 This is a flowchart of a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to the present invention.
[0050] Figure 2 This is a schematic diagram of the KAN neural network architecture mentioned in the data-driven module of the high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion in this invention.
[0051] Figure 3 This is a schematic diagram of the structure of a high-frequency transformer thermal management system based on data-driven technology and multimodal data fusion according to the present invention. Detailed Implementation
[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0053] Example 1
[0054] like Figure 1 As shown, a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion includes:
[0055] S1 Data Acquisition: Acquires operating data of the high-frequency transformer through the multimodal data acquisition module.
[0056] As further explanation: the data collected from the high-frequency transformer includes electrical parameters, environmental parameters, and temperature parameters.
[0057] In this embodiment, electrical parameters such as transformer terminal current, active power, reactive power, apparent power, power factor, load factor, and operating switching frequency are monitored in real time using high-precision current sensors (including but not limited to Hall effect sensors), voltage sensors, and power analyzers. These parameters directly reflect the operating conditions and internal losses of the high-frequency transformer.
[0058] It should be noted that: for current sensors, such as Rogowski coil type current sensors, the spacing between them should be increased, and the position of the device being measured should be fixed to reduce measurement errors. Voltage sensors typically use high-voltage differential probes, and the probe measurement spacing should be as close as possible to reduce errors caused by the distribution coefficient between probes in high-frequency environments. When deploying current and voltage sensors on the same power analyzer to monitor the power parameters of the transformer under test, the aforementioned electrical parameter measurement sensors must be reliably grounded, or a reliable isolation transformer should be used for isolated measurements.
[0059] In this embodiment, environmental parameters are collected using ambient temperature sensors, humidity sensors, and airflow velocity sensors to assess the temperature, humidity, and heat dissipation airflow around the high-frequency transformer. Environmental factors significantly impact the transformer's heat dissipation capacity and thermal equilibrium.
[0060] It should be noted that environmental sensors should be deployed in the same space as the transformer to reduce measurement errors caused by spatial distance.
[0061] In this embodiment, temperature parameters are monitored in real time by embedding high-precision temperature sensors (such as thermocouples or PT100 resistance thermometers) at locations inside the high-frequency transformer (such as at the windings, the middle of the core column, the middle of the core yoke, and the surface of the insulating material). These characteristic temperature points are important bases for assessing the thermal state of the high-frequency transformer and predicting hot spot temperatures.
[0062] It should be noted that in this embodiment, all data collected by the sensors is converted into optical signals, and then transmitted to the data driver module via optical isolation to a reliable communication interface (such as RS485, Ethernet, wireless communication module, etc.).
[0063] S2 Data Processing: The multimodal data collected by S1 is input as data features into the data-driven module for data preprocessing.
[0064] As further explanation: Data preprocessing in S2 is as follows:
[0065] S21 Data Synchronization: Synchronizes the sampling time of multi-source data output from different sensors.
[0066] In this embodiment: Since different sensors may have different sampling times, this module adopts timestamp alignment technology, including but not limited to, to ensure the synchronization of all data in the time dimension, providing a consistent basis for subsequent fusion analysis.
[0067] It should be noted that the sampling time synchronization in S21 data synchronization is as follows:
[0068] Due to differences in measurement timescales (e.g., voltage measurements at high frequencies have a timescale of μs, while temperature parameter measurements have a timescale of h), time series are stretched or compressed to reach the same time axis range. The Dynamic Time Warping (DTW) algorithm is used to construct a cumulative cost matrix and calculate the minimum path to align the time series. The cost matrix is expressed in the following formula:
[0069] D(i,j) = C(i,j) + min{D(i-1,j), D(i,j-1), D(i-1,j-1)}
[0070] In the formula, D is the current cumulative cost, and C is the local cost at each step.
[0071] S22 Data Cleaning: Noise removal, missing value imputation, and outlier detection of raw data.
[0072] It should be noted that there is a lot of electromagnetic interference in the test environment, and the acquisition module is prone to data loss during data transmission. Missing values need to be filled in to maintain data continuity and improve the accuracy of subsequent model training.
[0073] Therefore, in this embodiment, the specific steps for filling missing values in the S22 data cleaning process are as follows:
[0074] Interpolation algorithms are used for filling; common interpolation methods include linear interpolation and cubic spline interpolation. The formula for constructing a linear polynomial using linear interpolation is as follows:
[0075] y=ax+b
[0076] The formula for constructing a cubic polynomial for cubic spline interpolation is as follows:
[0077] y=ax 3 +bx 2 +cx+d.
[0078] It should be noted that electromagnetic interference in the experiment can not only easily lead to data loss, but also easily interfere with the data transmission process of the acquisition module. The acquisition module usually calculates the actual acquired signal value by detecting the duration of the high-level signal. However, the high-level time has low anti-disturbance capability, and abnormal values are easily generated in the acquired data.
[0079] Therefore, in this embodiment, outlier detection in data cleaning S22 is specifically as follows:
[0080] The multimodal data acquisition module calculates the actual acquired signal value by detecting the duration of the high-level signal; it detects outliers in the acquired data by calculating the mean and standard deviation of the data segment sequence, and removes and fills in missing values.
[0081] S23 Data Fusion: Integrates electrical, environmental, and temperature data through a multi-source data fusion algorithm;
[0082] Among them, the fusion methods are feature fusion (concatenating or transforming features of different types of data) or decision fusion (weighting or voting on the prediction results of different models).
[0083] S24 Data Feature Extraction: Extract data features reflecting the thermal state and potential thermal risks of the transformer.
[0084] It should be noted that the data characteristics reflecting the thermal state and potential thermal risks of transformers can be based on the common knowledge of those skilled in the art and / or data analysis, and new, representative characteristics can be extracted or constructed from the original data, including but not limited to: calculating the total loss, loss density, temperature gradient, temperature change rate, etc. These characteristics can more effectively reflect the thermal state and potential thermal risks of transformers.
[0085] S3 Data Comparison: The host computer control module receives the preprocessed data output by the data-driven module in S2, predicts the hot spot temperature of the high-frequency transformer through its integrated intelligent algorithm, and compares it with the set expected temperature to give a thermal management strategy.
[0086] As a further explanation: Figure 2 As shown, the intelligent algorithms used in the S3 data comparison are two different neural network architectures: MLP and / or KAN.
[0087] It should be noted that the key difference between the KAN neural network and the traditional MLP neural network is that, in this neural network, each edge is not set to a fixed scalar weight, but rather a univariate learnable function ϕ. ij (x) is typically a linear combination of B-spline basis functions (given by the Cox–de Boor recursive formula, the specific recursive formula is not described here); while the nodes in the network only need to sum the outputs of all edges in the previous layer. The KAN neural network architecture can be viewed as "input dimension × output dimension × B-spline basis", which can achieve the same training accuracy of MLP with fewer parameters.
[0088] The acquired data, processed by the data-driven module, is initialized. The dimensions in the B-spline basis functions can be controlled, approximately initialized as ϕ(x)≈x. The optimizer used in the model is typically Adam / AdamW, with a learning rate set slightly lower than that of MLP models (e.g., an initial learning rate of 1e-3), because the gradients on each edge in the KAN neural network architecture are more complex. The cost function can be set using the traditional MSE function plus a cost component based on the fundamental laws that the training physical quantities must satisfy (e.g., for the temperature parameter in the training model, the cost component due to the rate of temperature change λ=max(0,|∂T / ∂t|-τ0)). 2 , where τ0 is assumed to be the maximum rate of change.
[0089] S4 thermal management strategy execution: The heat dissipation module executes the heat dissipation strategy output by the host computer, and the drive control cabinet controls the high-frequency transformer to execute the drive strategy output by the host computer.
[0090] As further explanation: The thermal management strategy executed by the S4 thermal management strategy is as follows: the heat dissipation strategy and the driving strategy are customized based on the preprocessed electrical parameters, environmental parameters and temperature parameters output by the data driving module; among them, the customized logic includes: maintaining the hot spot temperature below the safety threshold, optimizing heat dissipation power consumption to improve system efficiency, and extending the life of the high-frequency transformer; the temperature data of the first two time points are collected through the PID control algorithm, and the power of the heat dissipation module is adjusted by the difference, so as to control the temperature rise fluctuation |T1-T0| within each collection period to be less than the set value ∆T, thereby realizing the customized logic.
[0091] In this embodiment, the heat dissipation module acts as the actuator of the thermal management system, responsible for actively regulating the heat dissipation device of the high-frequency transformer based on the output of the host computer control module and in conjunction with the thermal management strategy. According to the hot spot temperature predicted by the host computer control module, the heat dissipation module can dynamically adjust the operating parameters of the heat dissipation device. For example, if the hot spot temperature is predicted to rise and approach a threshold, the system can proactively increase the start-up of the coolant pump or adjust the coolant flow rate. Typically, a PID control algorithm is used to achieve precise and smooth control of the heat dissipation device, avoiding significant sudden temperature increases.
[0092] S5 Data Feedback: The operating data of the high-frequency transformer is collected again by S1 to determine whether it meets expectations. If it does not meet expectations, S3 is executed to re-formulate the thermal management strategy.
[0093] In this embodiment: Simulation verification platform: A high-frequency transformer electromagnetic-thermal coupling simulation model is constructed using finite element simulation software (such as Comsol Multiphysics, Ansys Maxwell / Fluent). For example, a transformer model drawn by engineering software such as Solidworks or CAD is imported into the Comsol platform. Given the voltage and current excitation waveforms, the corresponding thermal conductivity coefficient expression is provided. The simulation circuit needs to provide stray impedance and distributed capacitance present in the actual test circuit, enabling the platform to generate hotspot temperature data obtained under difficult experimental conditions. This data serves as a supplement or verification to the host computer control module and can be used to verify the effectiveness of different thermal management strategies.
[0094] Example 2
[0095] like Figure 3 As shown, a system is used to implement a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion, comprising:
[0096] The multimodal data acquisition module is used to acquire the multimodal operation data of the high-frequency transformer.
[0097] It should be noted that this data includes at least the temperature information of the high-frequency transformer (such as local winding temperature, core column temperature, epoxy resin material temperature, etc.), the electrical parameters of the high-frequency transformer (such as terminal current, power, load rate, switching frequency, etc.), and environmental parameters (such as ambient temperature, humidity, ambient airflow velocity, etc.). By deploying various high-precision sensors, such as thermocouples, PT100 RTDs, current / voltage sensors, power analyzers, and environmental sensors, the diversity and accuracy of the data are ensured.
[0098] The data-driven module is the data hub of the system, used to efficiently fuse, clean, synchronize, and extract features from heterogeneous multimodal data acquired from the feature input multimodal data acquisition module.
[0099] It should be noted that the methods in the data-driven module include, but are not limited to, time series alignment, missing value imputation, and outlier detection. Through neural network-based artificial intelligence algorithms, multimodal data are integrated into a high-quality comprehensive feature dataset, and feature engineering methods (such as principal component analysis) are applied to extract features strongly correlated with the temperature state of the high-frequency transformer, thereby improving the accuracy and robustness of subsequent prediction models.
[0100] The host computer control module is the core intelligent prediction part of the system. It receives preprocessed data from the data-driven module, predicts the hot spot temperature of the high-frequency transformer based on intelligent algorithms, compares it with the set expected temperature, and then provides a thermal management strategy.
[0101] The heat dissipation module is used to implement the heat dissipation strategy of the high-frequency transformer, which is given by the host computer control module for heat management.
[0102] As further explanation: The heat dissipation module includes a fan and / or a liquid cooling system. The liquid cooling system includes a coolant pump, a controllable cooling device, etc. The controllable cooling device can adjust the heat dissipation method according to the heat dissipation strategy, achieving proactive thermal management, avoiding temperature overshoot or hysteresis response, thereby maintaining the high-frequency transformer temperature within the optimal operating range, effectively preventing overheating damage, extending equipment life, and optimizing overall operating efficiency.
[0103] The drive control cabinet control is used to execute the drive strategy given by the host computer control module for thermal management.
[0104] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion.
[0105] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of a high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion are disclosed.
[0106] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
Claims
1. A high-frequency transformer thermal management method based on data-driven technology and multi-modal data fusion, characterized in that, include: S1 Data Acquisition: Acquires operating data of the high-frequency transformer through a multi-modal data acquisition module; S2 Data Processing: The multimodal data collected by S1 is input as data features into the data-driven module for data preprocessing; S3 Data Comparison: The host computer control module receives the preprocessed data output by the data-driven module in S2, predicts the hot spot temperature of the high-frequency transformer through its integrated intelligent algorithm, and compares it with the set expected temperature to give a thermal management strategy. S4 thermal management strategy execution: The heat dissipation module executes the heat dissipation strategy output by the host computer, and the drive control cabinet controls the high-frequency transformer to execute the drive strategy output by the host computer; S5 Data Feedback: The operating data of the high-frequency transformer is collected again by S1 to determine whether it meets expectations. If it does not meet expectations, S3 is executed to re-formulate the thermal management strategy.
2. The high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to claim 1, characterized in that: The data collected in the S1 data acquisition process for the high-frequency transformer includes: electrical parameters, environmental parameters, and temperature parameters.
3. The high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to claim 2, characterized in that: Data preprocessing in S2 is as follows: S21 Data Synchronization: Synchronizes the sampling time of multi-source data output from different sensors; S22 Data Cleaning: Noise removal, missing value imputation, and outlier detection of raw data; S23 Data Fusion: Integrates electrical, environmental, and temperature data through a multi-source data fusion algorithm; Among them, the fusion method is feature fusion or decision fusion; S24 Data Feature Extraction: Extract data features reflecting the thermal state and potential thermal risks of the transformer.
4. The high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to claim 3, characterized in that: The specific steps for sampling time synchronization in S21 data synchronization are as follows: By stretching or compressing to reach the same time axis range, a cumulative cost matrix is constructed using a dynamic time warping algorithm. The minimum path for aligning time series is then calculated, and the cost matrix is expressed in the following formula: D(i,j) = C(i,j) + min{D(i-1,j), D(i,j-1), D(i-1,j-1)} In the formula, D is the current cumulative cost, and C is the local cost at each step.
5. The high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to claim 3, characterized in that: The specific steps for filling missing values in S22 data cleaning are as follows: Interpolation algorithms are used for filling; common interpolation methods include linear interpolation and cubic spline interpolation. The formula for constructing a linear polynomial using linear interpolation is as follows: y=ax+b The formula for constructing a cubic polynomial for cubic spline interpolation is as follows: y=ax 3 +bx 2 +cx+d。 6. The high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to claim 3, characterized in that: The outlier detection process in S22 data cleaning is as follows: The multimodal data acquisition module calculates the actual acquired signal value by detecting the duration of the high-level signal; By calculating the mean and standard deviation of the fragment data sequence, outliers in the collected data are detected, removed, and missing values are filled.
7. The high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to any one of claims 1-6, characterized in that: The intelligent algorithms used in the S3 data comparison employ two different neural network architectures: MLP and / or KAN.
8. The high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion according to claim 7, characterized in that: The specific thermal management strategy implemented by the S4 thermal management strategy is as follows: The heat dissipation strategy and drive strategy are customized based on the preprocessed electrical parameters, environmental parameters and temperature parameters output by the data drive module; The customized logic includes: maintaining hot spot temperatures below safe thresholds, optimizing heat dissipation and power consumption to improve system efficiency, and extending the lifespan of high-frequency transformers; The temperature data from the first two time points is collected using a PID control algorithm. The power of the heat dissipation module is adjusted by calculating the difference between the two data points. This ensures that the temperature rise fluctuation |T1-T0| within each data collection period is less than the set value ∆T, thus realizing customized logic.
9. A system, characterized by The method for implementing the high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion as described in any one of claims 1-8 includes: The multimodal data acquisition module is used to acquire the multimodal operating data of the high-frequency transformer; The data-driven module is used to efficiently fuse, clean, synchronize, and extract features from heterogeneous multimodal data acquired from the feature input multimodal data acquisition module; The host computer control module receives preprocessed data output from the data drive module, predicts the hot spot temperature of the high-frequency transformer based on intelligent algorithms, compares it with the set expected temperature, and provides a thermal management strategy. The heat dissipation module is used to implement the heat dissipation strategy of the high-frequency transformer based on the heat management strategy given by the host computer control module. The drive control cabinet control is used to execute the drive strategy given by the host computer control module for thermal management.
10. The system of claim 9, wherein: The heat dissipation module includes a fan and / or a liquid cooling system. 11.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-10. When the processor executes the computer program, it implements the steps of the high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion as described in any one of claims 1-8.
12. A computer readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the high-frequency transformer thermal management method based on data-driven technology and multimodal data fusion as described in any one of claims 1-8.