High-precision voltage regulating system of power frequency non-partial discharge test transformer

Abstract

The invention discloses a high-precision voltage regulating system of a power frequency no-partial-discharge test transformer, and the system comprises a heterogeneous multi-stage power module which is formed by connecting an H-bridge unit group and a matrix converter in parallel, the H-bridge unit group outputs a fundamental voltage component, and the matrix converter compensates a high-frequency harmonic component; the multi-physics field digital twinning unit integrates an electromagnetic field solver, a thermodynamic model and a mechanical deformation simulation module and predicts the multi-physics field coupling state of the insulation system in real time; the dynamic impedance matching network comprises an adjustable inductor array and a digital capacitor bank, and automatically adjusts filtering parameters according to a load impedance measurement value; and the control module is used for deploying a deep learning model and is used for optimizing a modulation strategy and a protection threshold on line. According to the method, the voltage regulation precision and the waveform purity can be improved, the stability under the wide-range working condition is guaranteed, the breakdown hidden danger is recognized in advance through multi-physics field coupling simulation, and partial discharge development is restrained from the source.

Description

Technical Field

[0001] The present invention relates to the field of voltage regulation systems, in particular to a high-precision voltage regulation system for a power frequency partial discharge-free test transformer. Background Art

[0002] Traditional power frequency test transformers face the following technical bottlenecks in high voltage testing:

[0003] Insufficient voltage regulation accuracy: Conventional voltage regulation systems rely on a single power module, making it difficult to balance fundamental voltage stability with high-frequency harmonic suppression, resulting in output waveform distortion and increased risk of partial discharge.

[0004] Poor load adaptability: Impedance matching networks are mostly designed with fixed parameters and cannot dynamically respond to load changes. This can easily cause voltage fluctuations and energy reflections, affecting test stability.

[0005] Insulation status monitoring lags: Traditional sensors have low spatial resolution and rely on point-based monitoring, making them unable to capture the electric field strength, temperature rise, and mechanical deformation coupling effects of the entire winding in real time, resulting in delayed insulation degradation warnings.

[0006] Rigid control strategy: PID or rule base control is used, with fixed parameters and reliance on manual experience. It is difficult to cope with complex working conditions, and the dynamic response speed and accuracy are insufficient.

[0007] Passive fault suppression: Existing systems mostly rely on post-fault protection and lack active identification and suppression mechanisms for partial discharge precursors, which limits equipment life and test safety.

[0008] Aiming at the above problems, a high-precision voltage regulation system for power frequency non-partial discharge test transformer is proposed. Summary of the Invention

[0009] The present invention solves the problems existing in the prior art through the following technical solutions, which include:

[0010] A heterogeneous multi-stage power module consists of an H-bridge unit group and a matrix converter in parallel, where the H-bridge unit group outputs the fundamental voltage component and the matrix converter compensates for the high-frequency harmonic components;

[0011] Multi-physics digital twin unit, integrating electromagnetic field solver, thermodynamic model and mechanical deformation simulation module, to predict the multi-physics coupling state of insulation system in real time;

[0012] Dynamic impedance matching network, including adjustable inductor array and digital capacitor bank, automatically adjusts filter parameters based on load impedance measurement;

[0013] The control module deploys a deep learning model for online optimization of modulation strategies and protection thresholds;

[0014] The fiber optic sensing unit arranges the fiber grating array along the transformer winding to monitor the local temperature rise and mechanical deformation in real time.

[0015] Furthermore, the matrix converter adopts a three-phase-single-phase direct AC / AC topology, its switching device is SiCMOSFET, the switching frequency is ≥100kHz, and the carrier phase difference with the H-bridge unit group is set to Δθ=2π / N, where N is the number of parallel branches.

[0016] Furthermore, the calculation process of the power distribution ratio of the H-bridge unit group and the matrix converter is as follows:

[0017]

[0018] Where V base is the fundamental voltage amplitude, V comp is the harmonic compensation voltage amplitude, I h / I1 is the ratio of the hth harmonic current to the fundamental current.

[0019] Furthermore, the process of the multi-physics digital twin unit performing coupling calculation is as follows:

[0020] Step a: Calculate the winding electric field intensity distribution E(x, y, z) using Maxwell equations;

[0021] Step b: Based on the heat conduction equation Solve the temperature field T(x,y,z,t), where Q loss is the loss power density;

[0022] Step c: Import the temperature field data into the mechanical deformation model to calculate the stress distribution σ of the insulation material ij .

[0023] Furthermore, when any point is predicted to satisfy E(x,y,z)>0.8E breakdown or σ ij >σ yield When the voltage reaches 300V, a three-level warning signal is generated and the target voltage curve is adjusted in conjunction.

[0024] Furthermore, the control logic of the dynamic impedance matching network is:

[0025] The amplitude of the injected current is V test =0.05 Vrated The test signal is used to measure the phase difference Δφ of the response current;

[0026] According to Z load =V test / (I test ·e jΔφ ) Calculate the load impedance and solve the optimal filter parameter Lopt with C opt .

[0027] Furthermore, the deep learning model is a dual-channel convolutional neural network (CNN), the first channel of which inputs voltage and current waveform data, the second channel inputs multi-physical field prediction results, and the output layer generates modulation mode selection instructions and PID parameter adjustment amounts.

[0028] Furthermore, the CNN training process adopts a transfer learning strategy, using finite element simulation data in the pre-training stage and injecting measured fault samples in the fine-tuning stage;

[0029] Furthermore, the spatial positioning error of the distributed optical fiber sensing unit is ≤1 cm, the temperature resolution is ≤0.1° C., and breakpoint positioning is achieved through optical time domain reflectometry technology.

[0030] Furthermore, the distributed optical fiber sensing unit further comprises:

[0031] A dynamic wavelength demodulation module dynamically adjusts the sampling frequency and demodulation accuracy of the fiber Bragg grating according to the electric field intensity distribution predicted by the multi-physics field digital twin unit;

[0032] A multi-source data fusion engine that inputs fiber optic sensor data, ultrasonic partial discharge signals, and vibration acceleration information into a spatiotemporal attention mechanism model to generate an insulation defect probability heat map.

[0033] The self-healing feedback loop triggers the heterogeneous power module to reconstruct the output waveform to suppress the development of partial discharge when it is detected that the probability of insulation defect exceeds the associated threshold.

[0034] Compared with the existing technology, the present invention has the following advantages: the high-precision voltage regulation system of the power frequency partial discharge-free test transformer has high-precision voltage regulation and harmonic suppression. Through a heterogeneous multi-stage power module (H-bridge unit group and matrix converter in parallel), it can achieve independent control of the fundamental voltage component and high-frequency harmonic components, significantly improve the voltage regulation accuracy, and reduce the risk of partial discharge.

[0035] The multi-physics digital twin unit integrates electromagnetic field, thermodynamics, and mechanical deformation models to simulate the insulation system status in real time, providing early warning of electric field overload or material stress exceeding the limit, and adjusting the voltage curve in a coordinated manner to enhance system safety. The dynamic impedance matching network automatically calculates the load impedance by injecting test signals and optimizes the filtering parameters in real time to ensure stable operation of the system under different loads and reduce harmonic interference.

[0036] The control module uses a dual-channel convolutional neural network (CNN) to fuse voltage and current waveforms with multi-physics field data, online optimizes modulation strategies and protection thresholds, and combines transfer learning strategies to improve model generalization capabilities, enabling efficient identification and active suppression of fault samples.

[0037] Distributed fiber optic sensing units, combined with dynamic wavelength demodulation and multi-source data fusion, accurately generate insulation defect heat maps. A self-healing feedback loop triggers waveform reconstruction based on defect probability, outputting anti-phase harmonics to offset electric field distortion, actively suppressing the development of partial discharge and extending equipment life.

[0038] Dynamically adjust the fiber Bragg grating sampling frequency and demodulation accuracy to avoid redundant data collection and reduce the system computing load. The spatiotemporal attention mechanism model integrates multi-source sensor data to improve the efficiency and accuracy of defect location.

[0039] Through the three-level early warning mechanism, reverse modulation instructions and dynamic adjustment of voltage levels, passive protection is upgraded to active suppression, significantly improving the equipment test life and long-term operation reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0041] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0042] like Figure 1 As shown, this embodiment provides a technical solution: a high-precision voltage regulation system for a power frequency non-partial discharge test transformer, comprising:

[0043] A heterogeneous multi-stage power module consists of an H-bridge unit group and a matrix converter in parallel, where the H-bridge unit group outputs the fundamental voltage component and the matrix converter compensates for the high-frequency harmonic components;

[0044] Multi-physics digital twin unit, integrating electromagnetic field solver, thermodynamic model and mechanical deformation simulation module, to predict the multi-physics coupling state of insulation system in real time;

[0045] Dynamic impedance matching network, including adjustable inductor array and digital capacitor bank, automatically adjusts filter parameters based on load impedance measurement;

[0046] The control module deploys a deep learning model for online optimization of modulation strategies and protection thresholds;

[0047] The fiber optic sensing unit arranges the fiber grating array along the transformer winding to monitor the local temperature rise and mechanical deformation in real time.

[0048] The matrix converter adopts a three-phase to single-phase direct AC / AC topology, its switching device is SiC MOSFET, the switching frequency is ≥100kHz, and the carrier phase difference with the H-bridge unit group is set to Δθ=2π / N, where N is the number of parallel branches;

[0049] The use of SiC MOSFET as the switching device, combined with a switching frequency design of ≥100kHz, significantly reduces conduction loss and switching loss, improves the overall system efficiency, and supports rapid compensation of high-frequency harmonics, improving the output waveform quality.

[0050] The three-phase to single-phase direct AC / AC topology reduces intermediate conversion links, simplifies the circuit structure, and reduces system complexity. At the same time, it improves the ability to work in conjunction with the H-bridge unit group, ensuring accurate separation and control of the fundamental and harmonic components.

[0051] The phase stagger strategy is used to reduce harmonic superposition and electromagnetic interference between parallel branches, improve system stability and output consistency, and avoid power fluctuations caused by phase conflicts.

[0052] The high temperature resistance and high-frequency switching capability of SiC MOSFET enable the system to operate stably under high temperature and high load conditions, reducing heat dissipation pressure, extending the life of key components, and enhancing the long-term reliability of the system.

[0053] High-frequency switching combined with optimized phase control shortens system response time, enables rapid dynamic adjustment to load changes and harmonic components, and ensures voltage regulation accuracy and partial discharge suppression effects.

[0054] The calculation process of the power distribution ratio of the H-bridge unit group and the matrix converter is as follows:

[0055]

[0056] Where V base is the fundamental voltage amplitude, V comp is the harmonic compensation voltage amplitude, I h / I1 is the ratio of the hth harmonic current to the fundamental current;

[0057] By quantifying the distribution ratio of fundamental voltage and harmonic compensation voltage through a formula and combining it with the proportion of harmonic current, the power of the H-bridge unit group and the matrix converter is optimized. This ensures the stability of the fundamental output while efficiently compensating for high-frequency harmonics and avoiding overload risks.

[0058] Dynamically adjust the compensation voltage amplitude based on the harmonic current ratio to specifically weaken specific subharmonic components, significantly reduce the probability of partial discharge, and enhance the purity and safety of the system output waveform;

[0059] Formula parameters are associated with load conditions in real time, supporting dynamic adjustment of power distribution ratio, adapting to complex and changing test environments, and ensuring the system's voltage regulation accuracy and stability under a wide range of loads;

[0060] Provides a clear mathematical model as a design basis, reduces traditional trial-and-error debugging costs, accelerates system parameter configuration and performance optimization, and improves project implementation efficiency;

[0061] The power allocation ratio calculation results can be used as input parameters of the deep learning model to assist in generating better modulation strategies, achieve deep integration of algorithms and hardware, and further improve system response speed and energy efficiency.

[0062] The process of performing coupling calculations by the multi-physics digital twin unit is as follows:

[0063] Step a: Calculate the winding electric field intensity distribution E(x, y, z) using Maxwell equations;

[0064] Step b: Based on the heat conduction equation Solve the temperature field T(x,y,z,t), where Q loss is the loss power density;

[0065] Step c: Import the temperature field data into the mechanical deformation model to calculate the stress distribution σ of the insulation material ij ;

[0066] Maxwell's equations accurately calculate the electric field intensity distribution, combined with the heat conduction equation to solve the temperature field. Furthermore, a mechanical deformation model is introduced to analyze stress, enabling multi-physics coupled simulation of electromagnetic, thermal, and mechanical deformation. This integrated modeling approach comprehensively captures system behavior under complex operating conditions, significantly improving prediction accuracy.

[0067] Real-time calculation of electric field strength and material stress distribution can identify potential risks such as local electric field excess or mechanical stress exceeding the limit in advance, provide data support for the generation of three-level warning signals, and avoid insulation system failure.

[0068] The temperature field data is linked with the mechanical deformation model to quantify the impact of temperature rise on the deformation of the insulation material, accurately predict the fatigue and aging trends of the material, and provide a theoretical basis for optimizing heat dissipation design and extending equipment life. The coupled simulation results are input into the control module in real time, and the modulation strategy and protection threshold are dynamically adjusted to achieve closed-loop control from simulation prediction to active regulation, thereby improving the system response speed and reliability. High-precision multi-physics field simulation replaces some physical experiments, shortening the R&D cycle and reducing trial and error costs, while providing direct data support for design optimization.

[0069] When any point is predicted to satisfy E(x,y,z)>0.8E breakdown or σ ij>σ yield When the voltage reaches 300V, a three-level warning signal is generated and the target voltage curve is adjusted in a coordinated manner;

[0070] When the predicted electric field strength exceeds the safety threshold or the material stress exceeds the yield limit, a three-level warning signal is immediately generated, triggering the system to intervene in advance to avoid irreversible damage caused by insulation breakdown or mechanical deformation, significantly reducing the probability of failure.

[0071] The target voltage curve is adjusted in a coordinated manner. By actively reducing the local electric field strength or releasing the voltage in the area of ​​stress concentration, the system operation status is balanced, ensuring that the test transformer operates within a safety margin and extending the service life of the equipment.

[0072] The three-level warning signals correspond to different risk levels (such as warning, alarm, and emergency), support differentiated control strategies (such as voltage fine-tuning, current limiting protection, and emergency shutdown), and improve the system's adaptability and response accuracy to complex working conditions.

[0073] Real-time linkage with multi-physics field digital twin units forms a closed loop from simulation prediction to early warning triggering to dynamic regulation, realizing full process automation from risk identification to active suppression, and enhancing system reliability. In high-voltage tests, the development of local discharge is suppressed through real-time voltage regulation, avoiding chain reactions caused by insulation failure, and ensuring the safety of the test environment and operators.

[0074] The control logic of the dynamic impedance matching network is:

[0075] The amplitude of the injected current is V test =0.05 Vrated The test signal is used to measure the phase difference Δφ of the response current;

[0076] According to Z load =V test / (I test ·e jΔφ ) Calculate the load impedance and solve the optimal filter parameter L opt with C opt ;

[0077] By injecting a small-amplitude test signal, the response current phase difference Δφ is accurately measured without interfering with the normal operation of the main system, achieving real-time monitoring of the load impedance and avoiding system fluctuations or additional losses caused by large signal injection in traditional methods.

[0078] Based on the load impedance calculation results, the optimal inductor and capacitor parameters are automatically solved, and the filtering characteristics of the dynamic impedance matching network are adjusted in real time to effectively suppress harmonic interference and ensure the purity of the output waveform and system stability;

[0079] Through the dynamic impedance matching mechanism, the system can quickly adapt to different load impedances (such as capacitive, inductive or nonlinear loads), eliminate voltage distortion or energy reflection caused by impedance mismatch, and improve voltage regulation accuracy and equipment compatibility;

[0080] Combined with a high-frequency matrix converter (≥100kHz), the dynamically adjusted filter network can accurately match the high-frequency harmonic compensation requirements, reduce the interference of high-frequency noise on partial discharge detection, and improve the reliability of test results;

[0081] Load impedance data can be input into the control module's deep learning model as a key parameter for optimizing the modulation strategy, enabling deep linkage between hardware adjustment and algorithm decision-making, significantly improving system response speed and energy efficiency.

[0082] Real-time impedance matching reduces device stress (such as capacitor overvoltage and inductor saturation) caused by impedance mutation or harmonic overload, reduces the aging rate of key components, and extends the overall service life of the system.

[0083] The deep learning model is a dual-channel convolutional neural network (CNN), the first channel of which inputs voltage and current waveform data, the second channel inputs multi-physical field prediction results, and the output layer generates modulation mode selection instructions and PID parameter adjustment amounts.

[0084] The CNN training process adopts a transfer learning strategy, using finite element simulation data in the pre-training stage and injecting measured fault samples in the fine-tuning stage;

[0085] The dual-channel convolutional neural network (CNN) captures electrical dynamic characteristics in real time through the first channel (voltage and current waveforms), and simultaneously analyzes the health status of the insulation system through the second channel (multi-physics field prediction results), realizing multi-dimensional electrical, mechanical and thermal data fusion, and improving the globality and accuracy of control decisions.

[0086] Based on the deep learning model, modulation mode selection instructions are generated in real time to adaptively match high-frequency harmonic compensation requirements with fundamental voltage regulation requirements, significantly improving waveform quality and voltage regulation efficiency, and reducing the risk of partial discharge.

[0087] The output layer dynamically adjusts PID parameters (such as proportional gain and integral time), breaking through the limitations of traditional fixed parameters, achieving precise closed-loop control for nonlinear loads and transient conditions, reducing overshoot and oscillation, and enhancing system stability.

[0088] In the pre-training stage, finite element simulation data is used to build a basic model, and in the fine-tuning stage, actual fault samples are introduced to significantly reduce the data acquisition cost and improve the model's generalization ability and robustness to actual complex working conditions (such as partial discharge and abnormal temperature rise).

[0089] Combined with multi-physics field prediction results, CNN can identify the correlation characteristics between voltage waveform distortion and insulation degradation in advance, trigger PID parameter pre-adjustment or modulation strategy switching, and achieve a transition from post-fault protection to pre-fault suppression.

[0090] The dual-channel structure uses feature-level parallel computing to ensure real-time performance while taking into account high-resolution analysis of multi-source data, thus avoiding control deviations driven by single sensor data.

[0091] The spatial positioning error of the distributed optical fiber sensing unit is ≤1 cm, the temperature resolution is ≤0.1° C., and the breakpoint positioning is achieved through the optical time domain reflectometry (OTDR) technology.

[0092] The spatial positioning error of the distributed fiber optic sensing unit is ≤1cm, and combined with a temperature resolution of ≤0.1℃, it can accurately capture subtle changes in local temperature rise and mechanical deformation of the transformer winding, providing highly sensitive early warning for early faults (such as partial discharge and insulation aging) to avoid the expansion of hidden dangers.

[0093] Optical Time Domain Reflectometry (OTDR) technology is used to detect fiber breakpoints in real time, shortening troubleshooting time, reducing the risk of data loss due to sensor link interruption, and ensuring the continuity and reliability of the monitoring system.

[0094] Arranging fiber grating arrays along the transformer windings enables continuous monitoring of the temperature and strain of the entire winding area, breaking through the spatial limitations of traditional point sensors, avoiding monitoring blind spots, and fully understanding the equipment's operating status.

[0095] Fiber optic sensing uses optical signal transmission and is naturally immune to electromagnetic interference. Fiber optic materials are corrosion-resistant and aging-resistant, adaptable to complex working conditions, and reduce maintenance frequency and costs.

[0096] High-precision sensor data is input into the multi-physics field digital twin unit in real time to calibrate the temperature field and stress field prediction results of the simulation model, thereby improving the confidence of the digital twin and the decision-making accuracy of active control.

[0097] The high-resolution spatiotemporal data of optical fiber sensing can be deeply integrated with ultrasonic partial discharge signals and vibration information to construct a multi-dimensional fault feature library and enhance the accuracy of generating insulation defect probability heat maps.

[0098] The distributed optical fiber sensing unit further comprises:

[0099] A dynamic wavelength demodulation module dynamically adjusts the sampling frequency and demodulation accuracy of the fiber Bragg grating according to the electric field intensity distribution predicted by the multi-physics field digital twin unit;

[0100] Breaking through the traditional fixed sampling mode, it dynamically optimizes the configuration of sensor resources based on the electric field prediction results, taking into account both accuracy and efficiency, avoiding redundant data collection, and reducing the system computing load;

[0101] A multi-source data fusion engine that inputs fiber optic sensor data, ultrasonic partial discharge signals, and vibration acceleration information into a spatiotemporal attention mechanism model to generate an insulation defect probability heat map.

[0102] Constructing spatiotemporal feature tensors α, β and γ are adaptive weight coefficients;

[0103] Calculating regional risk weights through attention mechanism Among them, Q, V and K correspond to the query vector, key vector and value vector, combining physical sensor data with AI attention mechanism to accurately locate potential discharge points;

[0104] a self-healing feedback loop that, when it is detected that the probability of an insulation defect exceeds an associated threshold, triggers the heterogeneous power module to reconstruct the output waveform to suppress the development of partial discharge;

[0105] A self-healing feedback loop marks high-risk areas in the digital twin model based on defect heat map coordinates;

[0106] Generate reverse modulation instructions to make the matrix converter output anti-phase harmonics to offset the electric field distortion in the area;

[0107] Continuously monitor the rate of change of discharge capacity and gradually reduce the overall voltage level if it does not converge;

[0108] Upgrading from passive protection to active suppression significantly extends the test life of equipment.

[0109] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0110] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0111] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A high-precision voltage regulation system for power frequency partial discharge-free test transformers, characterized in that: include: A heterogeneous multi-stage power module consists of an H-bridge unit group and a matrix converter in parallel, where the H-bridge unit group outputs the fundamental voltage component and the matrix converter compensates for the high-frequency harmonic components; Multi-physics digital twin unit, integrating electromagnetic field solver, thermodynamic model and mechanical deformation simulation module, to predict the multi-physics coupling state of insulation system in real time; Dynamic impedance matching network, including adjustable inductor array and digital capacitor bank, automatically adjusts filter parameters based on load impedance measurement; The control module deploys a deep learning model for online optimization of modulation strategies and protection thresholds; The fiber optic sensing unit arranges the fiber grating array along the transformer winding to monitor the local temperature rise and mechanical deformation in real time.

2. The high-precision voltage regulation system for a power frequency non-partial discharge test transformer according to claim 1, characterized in that: The matrix converter adopts a three-phase to single-phase direct AC / AC topology, its switching device is SiC MOSFET, the switching frequency is ≥100kHz, and the carrier phase difference with the H-bridge unit group is set to Δθ=2π / N, where N is the number of parallel branches.

3. The high-precision voltage regulation system for a power frequency non-partial discharge test transformer according to claim 1, characterized in that: The calculation process of the power distribution ratio of the H-bridge unit group and the matrix converter is as follows: Where V base is the fundamental voltage amplitude, V comp is the harmonic compensation voltage amplitude, I h / I1 is the ratio of the hth harmonic current to the fundamental current.

4. The high-precision voltage regulation system for a power frequency non-partial discharge test transformer according to claim 1, characterized in that: The process of performing coupling calculations by the multi-physics digital twin unit is as follows: Step a: Calculate the winding electric field intensity distribution E(x, y, z) using Maxwell equations; Step b: Based on the heat conduction equation Solve the temperature field T(x,y,z,t), where Q loss is the loss power density; Step c: Import the temperature field data into the mechanical deformation model to calculate the stress distribution σ of the insulation material ij .

5. The high-precision voltage regulation system for power frequency non-partial discharge test transformer according to claim 4, characterized in that: When any point is predicted to satisfy E(x,y,z)>0.8E breakdown or σ ij >σ yield When the voltage reaches 300V, a three-level warning signal is generated and the target voltage curve is adjusted in conjunction.

6. The high-precision voltage regulation system for power frequency non-partial discharge test transformer according to claim 1, characterized in that: The control logic of the dynamic impedance matching network is: The amplitude of the injected current is V test =0.05 Vrated The test signal is used to measure the phase difference Δφ of the response current; According to Z load =V test / (I test ·e jΔφ ) Calculate the load impedance and solve the optimal filter parameter L opt with C opt .

7. The high-precision voltage regulation system for power frequency non-partial discharge test transformer according to claim 1, characterized in that: The deep learning model is a dual-channel convolutional neural network, whose first channel inputs voltage and current waveform data, the second channel inputs multi-physical field prediction results, and the output layer generates modulation mode selection instructions and PID parameter adjustment amounts.

8. The high-precision voltage regulation system for power frequency non-partial discharge test transformer according to claim 7, characterized in that: The CNN training process adopts a transfer learning strategy, using finite element simulation data in the pre-training stage and injecting measured fault samples in the fine-tuning stage.

9. The high-precision voltage regulation system for power frequency non-partial discharge test transformer according to claim 1, characterized in that: The spatial positioning error of the distributed optical fiber sensing unit is ≤1 cm, the temperature resolution is ≤0.1°C, and breakpoint positioning is achieved through optical time domain reflection technology.

10. The high-precision voltage regulation system for power frequency non-partial discharge test transformer according to claim 9, characterized in that: The distributed optical fiber sensing unit further comprises: A dynamic wavelength demodulation module dynamically adjusts the sampling frequency and demodulation accuracy of the fiber Bragg grating according to the electric field intensity distribution predicted by the multi-physics field digital twin unit; A multi-source data fusion engine that inputs fiber optic sensor data, ultrasonic partial discharge signals, and vibration acceleration information into a spatiotemporal attention mechanism model to generate an insulation defect probability heat map. The self-healing feedback loop triggers the heterogeneous power module to reconstruct the output waveform to suppress the development of partial discharge when it is detected that the probability of insulation defect exceeds the associated threshold.

Images