Intelligent low-loss optimization system and method for transformer core

Abstract

The application relates to a transformer core intelligent low-loss optimization system and method, and belongs to the technical field of transformer core loss control. The system comprises a multilevel dynamic monitoring module, a material characteristic identification module, a three-stage magnetic field optimization control module, a reconfigurable magnetic field regulation and control module and a self-adaptive optimization and maintenance module. A giant magnetoresistance sensor array is deployed to obtain multidimensional monitoring data flow; a pulse magnetization test is adopted to construct a multidimensional loss model; a threshold-driven three-stage magnetic field optimization control strategy is implemented; the magnetic flux density is accurately regulated and controlled through reconfigurable unit coding and an auxiliary excitation coil matrix; and self-adaptive loss minimization is carried out in combination with a genetic algorithm and a particle swarm optimization. The application can realize accurate control and optimization of transformer core loss, significantly improve energy utilization efficiency and reduce harmonic additional loss.

Description

Technical Field

[0001] This invention relates to the field of power system equipment optimization, and in particular to an intelligent low-loss optimization system and method for transformer cores, belonging to the field of transformer core loss control and intelligent operation and maintenance technology. Background Technology

[0002] As a key piece of equipment in the power system, the core loss of a transformer directly affects the overall energy utilization efficiency. The transformer core is mainly composed of stacked grain-oriented silicon steel sheets, and its no-load loss mainly includes hysteresis loss and eddy current loss, which are important factors affecting transformer efficiency.

[0003] Traditional transformer core design relies primarily on static optimization and empirical formulas, typically employing methods such as increasing the silicon content of silicon steel sheets, reducing the thickness of silicon steel sheets, and improving stacking processes to reduce losses. For example, using high-permeability non-oriented silicon steel or laser marking technology to treat the core material can reduce the resistance to magnetic domain movement, thereby lowering hysteresis losses.

[0004] Currently, advanced transformer core optimization technologies typically employ a single magnetic field monitoring and open-loop control strategy. This involves installing a small number of magnetic field sensors on the core surface to acquire local magnetic field distribution information and making simple adjustments based on preset parameters. While this method improves core performance to some extent, it lacks dynamic adaptability and global optimization capabilities. In actual operation, factors such as temperature changes, magnetic flux density fluctuations, and harmonic interference alter the magnetic characteristics of the core. Traditional methods cannot respond to these changes in real time, resulting in no-load losses typically hovering between 1.0-1.5 W / kg, which is insufficient to meet the requirements of ultra-high-efficiency transformers. Especially under conditions of severe grid harmonic pollution, the additional core losses increase significantly, causing the actual operating efficiency to fall far below the design value.

[0005] The existing technology has the following shortcomings: it lacks an intelligent dynamic control mechanism and cannot adaptively optimize for real-time changes in the properties of the core material; it lacks the ability to monitor and predict the degree of core aging and cannot detect potential problems in advance; in addition, in the face of complex and ever-changing power grid environment, especially harmonic pollution conditions, the existing technology does not have effective compensation measures and cannot minimize losses to the greatest extent. Summary of the Invention

[0006] The purpose of this invention is to overcome the aforementioned problems in the prior art and provide an intelligent low-loss optimization system and method for transformer cores. Through the synergistic effect of a multi-level dynamic monitoring network, intelligent characteristic identification, a three-stage magnetic field optimization control strategy, and a reconfigurable magnetic field control unit, the system achieves precise control and optimization of transformer core losses, while also providing predictive maintenance functions.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for intelligent low-loss optimization of transformer cores includes the following steps: The structural characteristics and magnetic circuit distribution information of the transformer core are obtained, and a giant magnetoresistive sensor array is deployed at key locations in the core to obtain a multi-dimensional monitoring data stream that includes magnetic field distribution, temperature distribution and harmonic interference. The multidimensional monitoring data stream is subjected to pulse magnetization test to obtain the BH magnetization curve and loss characteristics of the core material in real time. A multidimensional loss model of temperature-frequency-magnetic induction intensity is constructed to obtain the real-time characteristic parameter set of the core material. Based on the real-time characteristic parameter set of the core material, the magnetic field optimization is divided into three stages: conventional optimization, critical optimization and emergency control. A preemption threshold is set and limited preemption control is implemented to obtain the magnetic field optimization control instruction set. The magnetic field optimization control instruction set is reconfigurable unit encoded, and the magnetic flux density is precisely controlled through the auxiliary excitation coil matrix coordinated control to obtain the optimal magnetic field distribution scheme; Based on the optimal magnetic field distribution scheme, a hybrid optimization strategy of genetic algorithm and particle swarm optimization is used to adaptively minimize losses. Combined with the analysis of hysteresis loop variation trend, a core aging prediction model is established to obtain comprehensive optimized control results and predictive maintenance suggestions.

[0008] Optionally, a giant magnetoresistive sensor array can be deployed at key locations in the core to obtain a multi-dimensional monitoring data stream containing magnetic field distribution, temperature distribution, and harmonic interference, including: Based on the structural characteristics and magnetic circuit distribution information of the transformer core, a giant magnetoresistive sensor array with a measurement accuracy of 0.01T is deployed at key locations of the core to obtain high-precision real-time monitoring data of the magnetic field inside the core. Temperature distribution monitoring is performed on the high-precision real-time monitoring data. A temperature monitoring subsystem is constructed using thermocouples and infrared thermal imaging technology to obtain temperature distribution data with a temperature monitoring accuracy of ±0.5℃. Based on the temperature distribution data, the power grid voltage waveform is collected in real time by a harmonic analyzer and a voltage waveform monitoring device to analyze the harmonic content and distribution characteristics and obtain harmonic interference detection data. The harmonic interference detection data is acquired synchronously from multiple sources using time synchronization technology. The heterogeneous data is preprocessed and fused using a multi-sensor data fusion algorithm to obtain the multi-dimensional monitoring data stream containing magnetic field distribution, temperature distribution, and harmonic interference.

[0009] Optionally, pulse magnetization testing is performed on the multidimensional monitoring data stream to construct a multidimensional loss model of temperature-frequency-magnetic induction intensity, thereby obtaining a real-time characteristic parameter set of the core material, including: Based on the multidimensional monitoring data stream, a pulse magnetization signal of a specific waveform is applied to the iron core, and magnetic field strength and magnetic induction intensity data are collected simultaneously to obtain the real-time BH magnetization curve. The real-time BH magnetization curve is decomposed into total core loss, and the hysteresis loss, eddy current loss and additional loss are calculated separately to obtain the core loss component calculation results. Based on the calculation results of the core loss components, a model is established to show the influence of temperature on permeability, hysteresis loss coefficient and resistivity, so as to realize the temperature correction of the core material properties and obtain the temperature-corrected material property parameters. The temperature-corrected material property parameters are fused to construct a multi-dimensional loss model that includes temperature, frequency, and magnetic induction intensity. The model accuracy is improved through small-sample testing and iterative optimization to obtain the real-time property parameter set of the core material.

[0010] Optionally, based on the real-time characteristic parameter set of the core material, limited preemption control is implemented to obtain a magnetic field optimization control command set, including: Based on the real-time characteristic parameter set of the core material, and combined with the system operating status, the magnetic field optimization control is divided into three stages: conventional optimization, critical optimization and emergency control. Corresponding triggering conditions and optimization objectives are set to obtain a three-stage task division scheme. A preemption threshold is set for the three-stage task division scheme, including a loss rate threshold, a temperature threshold, and a harmonic interference threshold. The threshold is dynamically adjusted according to the system operating status to obtain dynamic preemption threshold parameters. Based on the dynamic preemption threshold parameter, when a high-priority task request is detected, it is determined whether to interrupt the currently executing low-priority task, implement limited preemption control to avoid wasting system resources, and obtain the preemption control execution instruction. A smooth switching process is designed for the preemptive control execution instructions to ensure stable system operation during the switching process between different optimization stages, avoid voltage fluctuations and transient processes caused by the switching process, and obtain the magnetic field optimization control instruction set.

[0011] Optionally, the magnetic field optimization control instruction set is reconfigurable unit encoded to obtain the optimal magnetic field distribution scheme, including: Based on the magnetic field optimization control instruction set, multiple independently controllable magnetic field control units are designed. Each unit includes an auxiliary excitation coil, a power drive circuit, and a local controller, resulting in a reconfigurable magnetic field control unit group. The reconfigurable magnetic field control unit group is encoded, and a unique code is assigned to each magnetic field control unit. Based on the optimization objective and the core state, the global magnetic field distribution mode is obtained by combining different unit activation modes. Based on the global magnetic field distribution pattern, an auxiliary excitation coil matrix is ​​deployed at key locations in the core. The magnitude and direction of the current in each coil are controlled according to the instructions of the magnetic field control unit, resulting in precise control of the magnetic flux density. The results of the precise control of magnetic flux density are monitored in real time. The actual magnetic field distribution is compared with the target distribution. Based on the deviation, closed-loop control is implemented to continuously optimize the magnetic field distribution and obtain the optimal magnetic field distribution scheme.

[0012] Optionally, a temperature monitoring subsystem is constructed using thermocouples and infrared thermal imaging technology to obtain temperature distribution data with a monitoring accuracy of ±0.5℃, including: Based on the high-precision real-time monitoring data, K-type thermocouple sensors are arranged in key parts of the core to measure the local temperature changes of the core and obtain contact temperature measurement data. The contact temperature measurement data is supplemented by infrared thermal imaging technology. A high-resolution infrared thermal imager is used to perform non-contact temperature monitoring on the iron core surface to obtain an image of the temperature distribution on the iron core surface. Based on the temperature distribution image of the core surface, a temperature field numerical model is established, and the internal temperature distribution of the core is calculated by the finite element analysis method to obtain three-dimensional temperature field distribution data. Temperature gradient analysis is performed on the three-dimensional temperature field distribution data to identify temperature anomaly areas and hotspot locations, a temperature change trend prediction model is established, and the temperature distribution analysis results are obtained. Based on the temperature distribution analysis results, a temperature monitoring alarm threshold is set. When the temperature exceeds the safe range, an early warning signal is triggered, enabling timely detection and handling of temperature anomalies, and obtaining temperature distribution data with a temperature monitoring accuracy of ±0.5℃.

[0013] Optionally, the real-time BH magnetization curve is decomposed into total core loss to obtain the core loss component calculation results, including: Based on the real-time BH magnetization curve, the area and shape parameters of the hysteresis loop are extracted, and the hysteresis loss coefficient is calculated. The hysteresis loss is proportional to the nth power of the magnetic induction intensity and the value of n ranges from 1.6 to 2.2, thus obtaining the hysteresis loss calculation data. Eddy current loss analysis is performed on the hysteresis loss calculation data. Based on the resistivity and thickness parameters of the silicon steel sheet, the eddy current loss is calculated, where the eddy current loss is proportional to the square of the frequency and the square of the magnetic induction intensity, and the eddy current loss calculation data is obtained. Based on the eddy current loss calculation data, the additional losses caused by harmonics and magnetic domain effects are analyzed, wherein the additional losses are proportional to the 1.5th power of the frequency and the 1.5th power of the magnetic induction intensity, and the additional loss calculation data is obtained. The additional loss calculation data is subjected to loss correction processing. Taking into account the influence of factors such as temperature, stress and aging on each loss component, a correction coefficient model is established to obtain the corrected loss component data. Based on the corrected loss component data, the hysteresis loss, eddy current loss, and additional loss are superimposed to calculate the total core loss value, thus completing the core loss component calculation result.

[0014] Optionally, a preemption threshold is set for the three-stage task partitioning scheme to obtain dynamic preemption threshold parameters, including: Based on the three-stage task division scheme, the loss rate threshold is set to 110% of the standard value, the temperature threshold is 85℃, and the harmonic interference threshold is 5% for the conventional optimization stage, thus obtaining the threshold parameters for the conventional optimization stage. The threshold parameters of the conventional optimization stage are set to the threshold parameters of the critical optimization stage. The loss rate threshold is set to 120% of the standard value, the temperature threshold is set to 95℃, and the harmonic interference threshold is set to 8%, thus obtaining the threshold parameters of the critical optimization stage. Based on the threshold parameters of the critical optimization stage, the loss rate threshold is set to 135% of the standard value, the temperature threshold is 105℃, and the harmonic interference threshold is 12% for the emergency control stage, thus obtaining the threshold parameters for the emergency control stage. A dynamic adjustment algorithm is designed for the threshold parameters in the emergency control phase. The threshold parameters are corrected in real time according to seasonal changes, load characteristics and historical operation data to obtain an adaptive threshold adjustment strategy. Based on the adaptive threshold adjustment strategy, a threshold parameter database is established to record the optimal threshold combination under different operating conditions, thereby realizing the intelligent optimization and self-learning function of the threshold parameters and obtaining the dynamic preemption threshold parameters.

[0015] Optionally, an auxiliary excitation coil matrix can be deployed at key locations in the core to obtain precise control of the magnetic flux density, including: Based on the global magnetic field distribution pattern, multi-turn auxiliary excitation coils are arranged at key positions such as the core column, yoke, and clamps to form a three-dimensional spatially distributed coil matrix network, thus obtaining the spatial layout scheme of the auxiliary excitation coils. A current control algorithm is designed for the spatial layout scheme of the auxiliary excitation coil. An independent current regulator is configured for each coil to achieve precise current control within the range of ±10A, and the coil current control parameters are obtained. Based on the coil current control parameters, a magnetic field superposition mathematical model is established to calculate the spatial distribution and superposition effect of the magnetic field generated by each coil, and the magnetic field distribution calculation results are obtained. The magnetic field distribution calculation results are optimized by using the least squares method and constraint optimization algorithm to solve for the optimal coil current combination under the condition of satisfying voltage stability requirements, and obtain the optimal excitation current scheme. Based on the optimal excitation current scheme, the current of each coil is precisely controlled by the PWM power drive circuit, the change in magnetic flux density is monitored in real time and closed-loop feedback adjustment is performed to obtain the precise control result of magnetic flux density.

[0016] This invention also provides an intelligent low-loss optimization device for transformer cores, comprising: The multi-level dynamic monitoring module is used to deploy a giant magnetoresistive sensor array at key locations in the iron core to acquire a multi-dimensional monitoring data stream that includes magnetic field distribution, temperature distribution, and harmonic interference. The material property identification module is used to perform pulse magnetization testing on the multidimensional monitoring data stream, construct a multidimensional loss model of temperature-frequency-magnetic induction intensity, and output a real-time characteristic parameter set of the core material. The three-stage magnetic field optimization control module is used to implement a threshold-driven three-stage magnetic field optimization control strategy based on the real-time characteristic parameter set of the core material and output a magnetic field optimization control instruction set. The reconfigurable magnetic field control module is used to encode the magnetic field optimization control instruction set into reconfigurable units, and outputs the optimal magnetic field distribution scheme through the coordinated control of the auxiliary excitation coil matrix. The adaptive optimization and maintenance module is used to minimize adaptive losses based on the optimal magnetic field distribution scheme using a hybrid optimization strategy, establish a core aging prediction model, and output comprehensive optimization control results and predictive maintenance suggestions.

[0017] The beneficial effects of this invention include: 1. By adopting a threshold-driven three-stage magnetic field optimization control strategy, the magnetic field optimization is divided into three stages: normal optimization, critical optimization, and emergency regulation. This achieves a limited preemption mechanism, ensuring that system resources are prioritized for processing high-priority tasks and improving system response efficiency. 2. A magnetic field control architecture based on reconfigurable units is adopted. By assigning a unique code to each magnetic field control unit, the combined activation mode of different control units is realized, and the required global magnetic field distribution mode is synthesized, so as to achieve high-precision magnetic field control at low cost. 3. Intelligent identification technology for core material characteristics combining pulse magnetization testing and multidimensional loss model, real-time acquisition of BH magnetization curve and construction of multidimensional loss model including temperature, frequency and magnetic induction intensity, improving the accuracy of loss calculation; 4. By integrating adaptive loss minimization and predictive maintenance into a system, and combining a hybrid optimization strategy of genetic algorithm and particle swarm optimization, loss minimization is achieved while ensuring transformer performance. At the same time, the degree of core aging is predicted by analyzing the trend of hysteresis loop variation. 5. Implement harmonic loss compensation technology by generating reverse harmonic flux through auxiliary excitation coil to counteract the impact of harmonics on the iron core, effectively reducing additional losses in harmonic pollution environments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the overall architecture of the intelligent low-loss optimization system for transformer cores of the present invention; Figure 2 This is a schematic diagram illustrating the deployment of the multi-level dynamic monitoring network of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0021] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0022] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0023] 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, not all, of the embodiments of the present invention. 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.

[0024] like Figure 1 As shown, the intelligent low-loss optimization system for transformer cores provided by this invention includes a multi-level dynamic monitoring module, a material characteristic identification module, a three-stage magnetic field optimization control module, a reconfigurable magnetic field regulation module, and an adaptive optimization and maintenance module. The modules are connected through a data bus to form a closed-loop control system.

[0025] Example 1: like Figure 1 and Figure 2 As shown, the specific steps of the method for deploying a giant magnetoresistive sensor array at key locations in the iron core to obtain a multi-dimensional monitoring data stream including magnetic field distribution, temperature distribution, and harmonic interference are as follows: S1: Based on the structural characteristics and magnetic circuit distribution information of the transformer core, a giant magnetoresistive sensor array with a measurement accuracy of 0.01T is deployed at key locations in the core to obtain high-precision real-time monitoring data of the magnetic field inside the core. Specifically, in step S1, based on the structural characteristics of the transformer core, high-precision magnetic flux density sensors based on the giant magnetoresistance effect are installed at various key parts of the core, such as the core columns, yoke, and joints, forming a sensor array network. These sensors have a measurement accuracy of up to 0.01T and a response time of less than 1μs, enabling them to capture minute changes in the magnetic field inside the core in real time. The sensor array is rationally arranged according to the magnetic circuit distribution of the core, ensuring that the magnetic field information in key magnetic circuit areas can be accurately monitored.

[0026] S2: Monitor the temperature distribution of high-precision real-time monitoring data. Construct a temperature monitoring subsystem using thermocouples and infrared thermal imaging technology to obtain temperature distribution data with a temperature monitoring accuracy of ±0.5℃. In step S2, the temperature distribution monitoring subsystem consists of two parts: thermocouples and infrared thermal imaging. Thermocouples are mainly installed inside the iron core in locations that are difficult to observe directly, while infrared thermal imaging technology is used to monitor the temperature distribution on the surface of the iron core. The two technologies complement each other, jointly constructing a complete temperature field of the iron core, with a temperature monitoring accuracy of ±0.5℃.

[0027] S3: Based on temperature distribution data, the grid voltage waveform is collected in real time by a harmonic analyzer and a voltage waveform monitoring device to analyze the harmonic content and distribution characteristics and obtain harmonic interference detection data. In step S3, the real-time harmonic interference detection module acquires the grid voltage waveform through voltage waveform monitoring devices installed on the primary and secondary sides of the transformer, and analyzes its harmonic content and distribution characteristics using a fast Fourier transform algorithm. The system can identify harmonic components from the fundamental wave to the 50th harmonic, providing a basis for subsequent harmonic interference compensation.

[0028] S4: Time synchronization technology is used to collect multi-source data synchronously for harmonic interference detection data. Multi-sensor data fusion algorithm is used to preprocess and fuse heterogeneous data to obtain a multi-dimensional monitoring data stream containing magnetic field distribution, temperature distribution and harmonic interference.

[0029] In step S4, the system employs GPS time synchronization technology to ensure that all monitoring data have a unified time reference. The multi-sensor data fusion algorithm, based on Kalman filtering and evidence theory, preprocesses and fuses heterogeneous data such as magnetic fields, temperature, and harmonics, eliminating noise and redundant information to form a multi-dimensional data stream describing the real-time operating status of the core.

[0030] In Example 1, the construction of a multi-level dynamic monitoring network for the transformer core is the foundation of the entire system. First, based on the structural characteristics and magnetic circuit distribution of the transformer core, the system deploys a high-precision giant magnetoresistive (GMR) sensor array at key locations within the core. These sensors operate based on the GMR effect, achieving a measurement accuracy of 0.01T and a response time of less than 1μs, enabling them to capture minute changes in the magnetic field within the core in real time. The sensor layout fully considers the non-uniform magnetic flux density distribution, focusing on monitoring key points in the magnetic field distribution, such as the central region of the core columns, the connection between the yoke and the core columns, and joints, forming a comprehensive monitoring network.

[0031] The temperature distribution monitoring subsystem consists of two parts: contact measurement and non-contact measurement. Contact measurement uses K-type thermocouple sensors, installed inside the core in locations difficult to observe directly, such as inside clamps and between laminations. Non-contact measurement utilizes a high-resolution infrared thermal imager to perform thermal imaging scans of the core surface. These two technologies complement each other, jointly constructing a complete temperature field for the core. The temperature monitoring accuracy reaches ±0.5℃, accurately reflecting the temperature distribution in various parts of the core and providing reliable data for subsequent temperature compensation and overheat warning.

[0032] The real-time harmonic interference detection module acquires grid voltage waveforms by installing high-precision voltage waveform monitoring devices on the primary and secondary sides of the transformer. The system employs a 32-bit high-speed A / D converter with a sampling rate of 100kHz, capable of capturing harmonic components up to the 50th order. The acquired waveform data is processed using a Fast Fourier Transform algorithm to analyze the amplitude, phase, and distribution characteristics of each harmonic. The harmonic analysis accuracy reaches 0.1%, providing an accurate basis for subsequent harmonic loss assessment and compensation.

[0033] Multi-source data acquisition and fusion processing are crucial for achieving comprehensive monitoring. The system employs GPS time synchronization technology to ensure all sensor data have a unified time reference, with a synchronization accuracy better than 1μs. The multi-sensor data fusion algorithm, based on improved Kalman filtering and DS evidence theory, preprocesses and fuses heterogeneous data such as magnetic fields, temperature, and harmonics. Through noise suppression, outlier detection, and information complementarity, the algorithm improves data quality, eliminates redundant information, and forms a high-quality, multi-dimensional data stream describing the real-time operating status of the core. This fused data retains its individual characteristics while also demonstrating the correlations between them, providing a solid foundation for subsequent intelligent analysis and optimized control.

[0034] Example 2: The specific steps for performing pulse magnetization tests on the multidimensional monitoring data stream, constructing a multidimensional loss model of temperature-frequency-magnetic induction intensity, and obtaining the real-time characteristic parameter set of the core material are as follows: S1: Based on the multi-dimensional monitoring data stream, a pulse magnetization signal of a specific waveform is applied to the iron core, and magnetic field strength and magnetic induction intensity data are collected at the same time to obtain the real-time BH magnetization curve; Pulse magnetization testing technology applies a pulse magnetization signal of a specific waveform (including square wave, triangular wave, and sine wave) to the iron core, while simultaneously acquiring data on magnetic field strength H and magnetic induction intensity B, and plotting a real-time BH magnetization curve. Unlike traditional static measurements, this method can obtain dynamic magnetization characteristics under the actual operating conditions of the transformer, more closely approximating real-world working conditions.

[0035] S2: Decompose the total core loss of the real-time BH magnetization curve, and calculate the three parts of hysteresis loss, eddy current loss and additional loss respectively to obtain the core loss component calculation results. The core loss components are decomposed using a modified Berotti three-component loss model, which decomposes the total core loss into three parts: hysteresis loss, eddy current loss, and additional loss. Hysteresis loss is proportional to the nth power of the magnetic flux density (the value of n is adjusted between 1.6 and 2.2 according to material properties), eddy current loss is proportional to the square of the frequency and the square of the magnetic flux density, and additional loss is proportional to the 1.5th power of the frequency and the 1.5th power of the magnetic flux density. This decomposition method allows for a more accurate analysis of the causes and influencing factors of various types of losses.

[0036] S3: Based on the calculation results of the core loss components, establish a model of the influence of temperature on permeability, hysteresis loss coefficient and resistivity, realize the temperature correction of the core material properties, and obtain the temperature-corrected material property parameters. The temperature-corrected model obtains the effect function of temperature on permeability, hysteresis loss coefficient, and resistivity by fitting experimental data. As temperature increases, the resistivity of the silicon steel sheet increases, and eddy current loss decreases; however, at the same time, the permeability decreases, and hysteresis loss increases. By establishing an accurate temperature-corrected model, the system can accurately calculate the actual loss of the iron core under different temperature conditions.

[0037] S4: The temperature-corrected material property parameters are fused to construct a multi-dimensional loss model that includes temperature, frequency, and magnetic induction intensity. The model accuracy is improved through small-sample testing and model iteration optimization to obtain a real-time property parameter set of the core material.

[0038] The multidimensional loss model comprehensively considers three main variables: temperature, frequency, and magnetic induction intensity. It employs a hybrid modeling approach combining support vector regression and neural networks to construct a high-precision core loss prediction model. After small-sample testing and iterative optimization, the model's loss prediction error under different operating conditions is controlled within 3%.

[0039] In Example 2, intelligent identification of core material properties and construction of a multi-dimensional loss model are among the core functions of the system. The pulse magnetization test technology applies a carefully designed pulse magnetization signal to the core while simultaneously acquiring data on magnetic field strength H and magnetic induction intensity B, and plots the BH magnetization curve in real time. The system can generate pulse signals of various waveforms, including square waves, triangular waves, and sine waves, with a frequency range from 50Hz to 1kHz, and the amplitude can be adjusted between 30% and 120% of the rated magnetic flux density. Compared with traditional static measurements, this dynamic testing method can more realistically reflect the magnetization characteristics of the transformer under actual operating conditions.

[0040] The core loss component decomposition employs an improved Bertotti three-component loss model, accurately decomposing the total core loss into three parts: hysteresis loss, eddy current loss, and additional loss. For hysteresis loss, the system analyzes the closed loop area of ​​the BH curve to extract characteristic parameters of the hysteresis loop, such as coercivity, remanence, and hysteresis loop shape factor, and accurately calculates the hysteresis loss coefficient. The model considers the nonlinear relationship between hysteresis loss and magnetic induction intensity, and the Steinmetz exponent n is dynamically adjusted between 1.6 and 2.2 according to material properties, significantly improving the accuracy of hysteresis loss calculation.

[0041] The eddy current loss calculation fully considers the physical properties of silicon steel sheets. The system establishes a detailed eddy current loss calculation model based on the thickness, stacking method, and resistivity of the silicon steel sheets. The model accurately reflects the relationship between eddy current loss and the square of the frequency and the square of the magnetic flux density, while also considering the influence of the silicon steel sheet surface coating, interlayer insulation, and degree of compression on the eddy current distribution. The system also optimizes for cases with complex frequency components, enabling accurate calculation of eddy current losses under non-sinusoidal excitation.

[0042] The influence of temperature on the properties of core materials cannot be ignored. A temperature-corrected model was established, and the effects of temperature on permeability, hysteresis loss coefficient, and resistivity were obtained through fitting with a large amount of experimental data. As temperature increases, the resistivity of silicon steel sheets increases, theoretically reducing eddy current losses; however, at the same time, the permeability decreases, increasing hysteresis losses. The model accurately describes these contradictory influencing factors, providing accurate loss correction coefficients under different temperature conditions, making loss calculations closer to reality.

[0043] The multidimensional loss model is an advanced mathematical model that comprehensively considers the influence of three main variables—temperature, frequency, and magnetic induction intensity—on core loss. The system employs a hybrid modeling approach combining support vector regression and deep neural networks to construct a high-precision core loss prediction model. Support vector regression handles accurate fitting in small sample regions, while deep neural networks provide global modeling capabilities and nonlinear expression. Model training utilizes an online incremental learning method, continuously optimizing its parameters as operational data accumulates. After rigorous testing, the model's loss prediction error under different operating conditions is controlled within 3%, providing a reliable theoretical foundation for subsequent optimization and control.

[0044] This intelligent identification technology for core material properties has significant advantages over traditional methods: First, it can acquire material properties in real time during normal transformer operation without requiring shutdown testing; second, the multi-dimensional loss model comprehensively considers various factors affecting core loss, greatly improving prediction accuracy; finally, the system has self-learning capabilities, continuously improving identification accuracy with accumulated operating experience and adapting to the changing characteristics of core materials over time. These advantages lay a solid foundation for achieving precise control of core loss.

[0045] Example 3: like Figure 1 As shown, the specific steps for implementing limited preemption control to obtain the magnetic field optimization control command set based on the real-time characteristic parameter set of the core material are as follows: S1: Based on the real-time characteristic parameter set of the core material, combined with the system operating status, the magnetic field optimization control is divided into three stages: conventional optimization, critical optimization and emergency control. Corresponding triggering conditions and optimization objectives are set to obtain a three-stage task division scheme. Specifically, in step S1, the three-stage task division scheme divides the magnetic field optimization control into three stages based on the severity and urgency of the core operating state: routine optimization (low priority) mainly targets efficiency improvement under normal operating conditions; critical optimization (medium priority) intervenes when the core parameters are close to the warning threshold; and emergency control (high priority) handles the abnormal state when the core parameters exceed the safe range.

[0046] S2: Set preemption thresholds for the three-stage task division scheme, including loss rate threshold, temperature threshold, and harmonic interference threshold. Dynamically adjust the thresholds according to the system operating status to obtain dynamic preemption threshold parameters. In step S2, the system sets three types of preemption thresholds: loss rate threshold, temperature threshold, and harmonic interference threshold. For the normal optimization phase, the loss rate threshold is set to 110% of the standard value, the temperature threshold to 85℃, and the harmonic interference threshold to 5%. For the critical optimization phase, the loss rate threshold is set to 120% of the standard value, the temperature threshold to 95℃, and the harmonic interference threshold to 8%. For the emergency control phase, the loss rate threshold is set to 135% of the standard value, the temperature threshold to 105℃, and the harmonic interference threshold to 12%. The system adjusts these thresholds in real time based on seasonal changes, load characteristics, and historical operating data.

[0047] S3: Based on the dynamic preemption threshold parameter, when a high-priority task request is detected, it determines whether to interrupt the currently executing low-priority task, implements limited preemption control to avoid wasting system resources, and obtains the preemption control execution instruction; In step S3, the limited preemption mechanism determines whether to interrupt a low-priority task by judging the relationship between the current system state and a preset threshold. When a high-priority task request is detected, the system evaluates the cost and benefit of interrupting the low-priority task. Preemption is only implemented when the benefit significantly outweighs the cost, thus avoiding system instability and resource waste caused by frequent switching.

[0048] S4: The preemptive control execution instruction is designed with smooth switching processing to ensure stable system operation during the switching process between different optimization stages, avoid voltage fluctuations and transient processes caused by the switching process, and obtain the magnetic field optimization control instruction set.

[0049] In step S4, to ensure a smooth transition between different optimization stages, the system adopts a gradual control strategy. By setting a transition period and gradual parameters, the control parameters change smoothly during the switching process, avoiding voltage fluctuations and transient processes caused by sudden changes. The final generated magnetic field optimization control instruction set contains complete information such as optimization objectives, control parameters, execution timing, and feedback mechanisms.

[0050] Example 3 details the implementation process of the threshold-driven three-stage magnetic field optimization control strategy. The core idea of ​​this strategy is to divide the magnetic field optimization control into three stages with different priorities based on the severity and urgency of the core's operating status. By setting a preemption threshold, resources are rationally allocated to ensure that the system prioritizes the most urgent tasks under limited computing resources.

[0051] First, the system divides its tasks into three phases based on the real-time characteristic parameter set of the core material and the current operating status. The routine optimization phase targets the normal operating state and is a low-priority task. Its main objective is to reduce normal operating losses and improve energy efficiency by finely adjusting the magnetic field distribution while ensuring the safe and stable operation of the transformer. The critical optimization phase targets the state where the core parameters are close to the warning threshold and is a medium-priority task. Its objective is to prevent further deterioration of the state and avoid triggering emergency control through more aggressive control strategies. The emergency control phase targets the abnormal state where the core parameters exceed the safe range and is a high-priority task. Its objective is to take decisive measures to quickly bring the system back to the safe operating range and prevent potential faults.

[0052] Setting the preemption threshold is crucial to the three-stage control strategy. The system sets three core thresholds for each stage: loss rate threshold, temperature threshold, and harmonic interference threshold. The loss rate threshold measures the ratio of actual core loss to standard loss, reflecting energy efficiency. The temperature threshold monitors the temperature of critical core components to prevent overheating damage. The harmonic interference threshold assesses the impact of grid harmonics on the core. For the normal optimization stage, relatively lenient thresholds are set: loss rate not exceeding 110% of the standard value, temperature not exceeding 85℃, and harmonic interference not exceeding 5%. The thresholds for the critical optimization stage are more stringent: loss rate threshold is 120% of the standard value, temperature threshold is 95℃, and harmonic interference threshold is 8%. The emergency control stage represents the system's extreme operating state: loss rate reaches 135% of the standard value, temperature approaches 105℃, and harmonic interference reaches 12%.

[0053] This threshold setting considers multiple key factors for the safe operation of transformers, forming a multi-dimensional comprehensive evaluation system. The system dynamically adjusts these thresholds based on seasonal changes, load characteristics, and historical operating data. For example, during the high temperatures of summer, the temperature threshold is appropriately lowered; during peak industrial load periods, the harmonic interference threshold is appropriately raised; and the loss rate threshold is adjusted accordingly based on the degree of core aging. This dynamic adjustment capability makes the system more adaptable to complex and changing operating environments.

[0054] Limited preemption is a guarantee of efficient system operation. When a high-priority task is detected, the system will not blindly interrupt the currently executing low-priority task, but will make a reasonable decision by assessing the costs and benefits of interruption. The system considers three key factors: the difference in task urgency, the current task's completion status, and the switching cost. Preemption will only be implemented when the urgency of the high-priority task is significantly higher than the current task, and the switching cost is acceptable. This mechanism avoids system instability and resource waste caused by frequent switching, improving overall control efficiency.

[0055] To ensure a smooth transition between different optimization stages, the system is designed with a smooth transition mechanism. Traditional abrupt switching can lead to sudden changes in excitation current, causing voltage fluctuations and transient processes, affecting grid stability. The system employs a gradual control strategy, setting transition periods and gradual parameters to ensure smooth changes in control parameters during the switching process. The switching speed is dynamically adjusted according to the urgency level: faster switching speeds for urgent tasks and smoother switching speeds for routine tasks. This smooth transition mechanism ensures that the transformer maintains stable operation even when task priorities change.

[0056] The final generated magnetic field optimization control instruction set contains complete control information: optimization objectives (such as minimizing losses, harmonic suppression, temperature equalization, etc.), control parameters (excitation current values ​​and phases of each control unit), execution timing (the sequence and time intervals of control actions), and feedback mechanism (parameters and response speed of closed-loop control). These detailed control instructions provide clear guidance for the next stage of magnetic field regulation.

[0057] Example 4: The specific steps for encoding the magnetic field optimization control instruction set into reconfigurable units to obtain the optimal magnetic field distribution scheme are as follows: S1: Based on the magnetic field optimization control instruction set, multiple independently controllable magnetic field control units are designed. Each unit includes an auxiliary excitation coil, a power drive circuit, and a local controller, resulting in a reconfigurable magnetic field control unit group. Specifically, in step S1, the reconfigurable magnetic field control unit is the core execution component of this system. Each control unit comprises three main parts: an auxiliary excitation coil for generating a local magnetic field; a power drive circuit responsible for providing precise current control; and a local controller that receives central control commands and executes local closed-loop control. The system designs 8-24 control units based on the transformer capacity and core structure, placing them at key locations within the core.

[0058] S2: Encode the reconfigurable magnetic field control unit group, assign a unique code to each magnetic field control unit, and obtain the global magnetic field distribution mode by combining different unit activation modes according to the optimization goal and the core state; In step S2, a unique binary code is assigned to each magnetic field control unit. By combining activation modes of different units, various desired global magnetic field distribution patterns can be synthesized. For example, for a system with 8 control units, 2^8 = 256 different basic distribution patterns can be generated. Theoretically, an infinite number of magnetic field distributions can be synthesized by adjusting the excitation intensity of each unit. The system selects the most suitable activation mode based on the optimization objective and the core state.

[0059] S3: Based on the global magnetic field distribution mode, an auxiliary excitation coil matrix is ​​deployed at key locations in the iron core. The magnitude and direction of the current in each coil are controlled according to the instructions of the magnetic field control unit to obtain precise control results of magnetic flux density. In step S3, the auxiliary excitation coil matrix is ​​deployed at key locations on the core according to a preset spatial layout, including the core pillars, yoke, and clamps. Each excitation coil is wound with multiple turns of insulated wire. By precisely controlling the magnitude and direction of the current in the coil, a directional auxiliary magnetic field is generated, which superimposes with the main magnetic field to form an optimal magnetic field distribution. The power drive circuit can provide current control with an accuracy of 0.01A within a range of ±10A.

[0060] S4: Real-time feedback monitoring of the precise control results of magnetic flux density, comparison of actual magnetic field distribution with target distribution, and continuous optimization of magnetic field distribution based on deviation through closed-loop control to obtain the optimal magnetic field distribution scheme.

[0061] In step S4, the system uses the giant magnetoresistive sensor array deployed in step 1 to monitor the effect of magnetic field control in real time, compares the actual magnetic field distribution with the target distribution, calculates the deviation, and generates a correction signal. Based on these deviation signals, the closed-loop control system continuously adjusts the excitation current of each control unit, so that the actual magnetic field distribution gradually approaches the target distribution, ultimately forming the optimal magnetic field distribution scheme.

[0062] Example 4 details the collaborative control process of the reconfigurable magnetic field control unit. The core innovation of this method lies in decomposing the traditional monolithic magnetic field control into multiple independently controllable control units. By combining and activating different units, precise control of the magnetic field distribution in the iron core can be achieved.

[0063] The reconfigurable magnetic field control unit is the core execution component of the system. Each control unit consists of three parts: an auxiliary excitation coil, a power drive circuit, and a local controller. The auxiliary excitation coil uses a high-permeability magnetic core and multi-turn insulated wire winding to generate a directional local magnetic field; the power drive circuit, based on H-bridge topology and PWM control technology, provides bidirectional adjustable excitation current; the local controller is responsible for receiving central control commands, executing local closed-loop control, and monitoring coil temperature and current status. The system designs an appropriate number of control units, typically 8 to 24, based on the transformer capacity and core structure, strategically placing them in key locations within the core.

[0064] Control unit encoding and mode synthesis are key technologies for achieving reconfigurable control. The system assigns a unique binary code to each magnetic field control unit. By combining different units to activate modes, various desired global magnetic field distribution patterns can be synthesized. This encoding method borrows from the "polarization-coded reconfigurable unit" technology in the radio frequency field, transforming the complex magnetic field control problem into a unit encoding problem. For example, for a system with 16 control units, 2^16 = 65536 different basic distribution patterns can be generated. By adjusting the excitation intensity and direction of each unit, theoretically an infinite number of magnetic field distributions can be synthesized. The system selects the most suitable activation mode from a preset mode library or calculates a new optimal mode in real time, based on the current optimization objective and core state.

[0065] The auxiliary excitation coil matrix is ​​the physical basis for achieving precise control of magnetic flux density. These coils are deployed at key locations in the core according to a pre-defined spatial layout, including the core pillars, yoke, and clamps. Each excitation coil is wound with multiple turns of insulated wire, and the number of turns, size, and installation position of the coil are carefully designed to maximize the control effect. The system generates a directional auxiliary magnetic field by precisely controlling the magnitude and direction of the current in the coil, which superimposes with the main magnetic field to form an optimal magnetic field distribution. The power drive circuit adopts high-precision digital current control technology, which can provide current control with an accuracy of 0.01A within a range of ±10A, with a response time of less than 1ms, meeting the requirements of real-time control.

[0066] The magnetic field superposition mathematical model is the theoretical foundation of the system design. Based on Maxwell's equations and the BH characteristic curve of the core material, the model accurately describes the spatial distribution and superposition effect of the magnetic fields generated by each excitation coil. The model considers the nonlinear characteristics of the core material, magnetic saturation effect, edge effect, and mutual inductance between adjacent coils. Through a combination of finite element analysis and experimental verification, a high-precision magnetic field distribution prediction model was established, with the error controlled within 5%. This model enables the system to accurately predict the magnetic field distribution under arbitrary coil current combinations, providing a reliable basis for optimized control.

[0067] Solving for the optimal excitation current scheme is a multi-objective optimization problem. The system needs to find the coil current combination that best approximates the target magnetic field distribution while meeting voltage stability requirements. The optimization solution employs a combination of a modified least squares method and the Lagrange multiplier method, considering multiple constraints such as coil current limitations, core saturation limitations, and power consumption limitations. To accelerate the solution process, the system uses an iterative algorithm based on gradient descent and introduces a pre-computed lookup table technique, simplifying the complex optimization problem into table lookup and interpolation operations, achieving a millisecond-level solution speed to meet the requirements of real-time control.

[0068] Closed-loop feedback control is the key mechanism for ensuring precise control of magnetic flux density. The system uses a giant magnetoresistive sensor array deployed in step 1 to monitor the effect of magnetic field control in real time, comparing the actual magnetic field distribution with the target distribution, calculating the deviation, and generating a correction signal. The closed-loop control system employs an improved PID control algorithm, combined with feedforward compensation and model predictive control strategies, enabling rapid response to magnetic field changes, suppression of external disturbances, and maintenance of system stability. The control loop bandwidth reaches 200Hz, sufficient to handle common disturbances and load changes in the power grid. Through this high-precision closed-loop control, the system can control the deviation between the actual magnetic field distribution and the target distribution within 2%, achieving precise control of magnetic flux density.

[0069] This magnetic field control architecture based on reconfigurable units has significant advantages: First, it decomposes the complex global control problem into multiple simple local control problems, reducing system complexity; second, through unit encoding and pattern synthesis, it can achieve high-precision magnetic field control with low hardware cost; finally, the system has good scalability and fault tolerance, and the overall system can still operate normally even if some control units fail. These characteristics make this method promising for application in the field of transformer core optimization control.

[0070] Example 5

[0071] Based on the optimal magnetic field distribution scheme, the specific steps for adaptive loss minimization to obtain comprehensive optimized control results and predictive maintenance suggestions include: Step A: Based on the optimal magnetic field distribution scheme, a hybrid optimization strategy combining genetic algorithm and particle swarm optimization is adopted to seek the operating state with minimum loss in real time under the constraints of voltage regulation accuracy and transient stability. Specifically, in step A, adaptive loss minimization employs a hybrid optimization strategy combining genetic algorithms and particle swarm optimization. The genetic algorithm performs a global search to identify potential optimal solution regions; the particle swarm optimization then conducts a localized, refined search within these regions, rapidly converging to the optimal solution. The optimization process must satisfy two constraints: voltage regulation accuracy must be no less than ±0.5%, and transient stability indicators must meet the power system stability requirements. Through this hybrid optimization strategy, the system can seek the operating state with minimum losses in real time within a dynamically changing operating environment.

[0072] Step B: When the grid harmonics exceed the standard, the auxiliary excitation coil is controlled to generate reverse harmonic flux according to the harmonic frequency and amplitude characteristics to counteract the influence of harmonics on the iron core and reduce additional losses. In step B, harmonic loss compensation technology is a major innovation of this system. When the harmonic analyzer detects excessive harmonics in the power grid, the system analyzes the frequency and amplitude characteristics of the harmonics and calculates the required reverse harmonic flux. Then, by controlling the auxiliary excitation coil to generate a reverse harmonic flux of a specific frequency and amplitude, it cancels out the flux caused by the power grid harmonics, significantly reducing the additional losses caused by harmonics. Experiments show that this technology can reduce additional losses caused by harmonics by up to 60%.

[0073] Step C: By monitoring the changing trend of the hysteresis loop, including parameters such as coercivity, remanence, and maximum permeability, an aging condition assessment model for the core material is established to predict the aging degree and remaining life of the silicon steel sheet. In step C, the core aging prediction model is established by monitoring the changing trend of the hysteresis loop. As the core material ages, its hysteresis loop undergoes characteristic changes: coercivity increases, remanence decreases, and maximum permeability declines. Through long-term monitoring and trend analysis of these parameters, the system establishes an assessment model for the aging state of the core material, which can accurately predict the aging degree and remaining service life of the silicon steel sheets, providing a basis for predictive maintenance.

[0074] Step D: Based on the core aging prediction results and system operation data, use decision support algorithms to generate predictive maintenance recommendations, including maintenance time windows, key component replacement recommendations, and operating parameter optimization schemes, and output comprehensive optimization control results and predictive maintenance recommendations.

[0075] In step D, the predictive maintenance recommendation system generates detailed maintenance recommendations based on core aging predictions and historical operating data, using a case-based reasoning decision support algorithm. These recommendations include optimal maintenance time windows, a list of critical components requiring replacement, and optimized operating parameters for specific aging conditions. Compared to traditional periodic maintenance strategies, predictive maintenance can reduce unnecessary downtime, extend equipment lifespan, and lower maintenance costs.

[0076] Example 5 details the implementation method of an integrated system for adaptive loss minimization and predictive maintenance. This system organically combines loss optimization with equipment health management, enabling it to not only reduce transformer core losses in real time but also predict the degree of core aging, provide scientific maintenance recommendations, and comprehensively improve the transformer's operating efficiency and service life.

[0077] Adaptive loss minimization employs a hybrid optimization strategy combining genetic algorithms and particle swarm optimization. Genetic algorithms possess powerful global search capabilities, finding potential optimal solution regions within a vast solution space, but their convergence speed is relatively slow. Particle swarm optimization, on the other hand, boasts faster local convergence speed but is prone to getting trapped in local optima. The system combines the advantages of both algorithms: the genetic algorithm handles the global search, identifying potential optimal solution regions; while the particle swarm optimization performs a refined local search within these regions, quickly converging to the optimal solution. The optimization process first encodes the core's operating parameters, forming chromosome and particle representations. Then, the fitness function is set as the reciprocal of the total core loss. Iterative optimization is performed through operations such as crossover, mutation, and position updates, ultimately finding the operating state with minimum loss.

[0078] During the optimization process, the system needs to meet two key constraints: voltage regulation accuracy of no less than ±0.5% to ensure the voltage output quality of the transformer; and transient stability indicators to meet the power system stability requirements to avoid system oscillations caused by the optimization process. These constraints are incorporated into the fitness function using a penalty function method, ensuring that the final optimization result minimizes losses while meeting system operation requirements. Practice shows that this hybrid optimization strategy can reduce additional losses by 3-5% compared to a single algorithm, while increasing the convergence speed by about 40%, and can adapt to the real-time optimization needs in the dynamic operating environment of transformers.

[0079] Harmonic loss compensation technology is a major innovation of this system. Traditional transformers experience significant additional losses in their cores under severe harmonic pollution conditions, which conventional optimization methods cannot address. This system proposes an active harmonic compensation strategy: when the harmonic analyzer detects excessive grid harmonics, the system analyzes the frequency spectrum and amplitude characteristics of the harmonics, using Fourier analysis to determine the main harmonic components. Then, through the reconfigurable magnetic field control unit in Example 4, the auxiliary excitation coil is controlled to generate reverse harmonic flux with a specific frequency and amplitude, which cancels out the flux caused by grid harmonics. The compensation accuracy depends on the accuracy of harmonic identification and the control precision of the auxiliary magnetic field; it typically compensates for the 3rd to 13th main harmonics, achieving a compensation efficiency of 60-85%. This technology is particularly effective in transformer applications in industrial areas and areas with concentrated large electronic equipment, effectively reducing additional losses caused by harmonics.

[0080] The core aging prediction model is the core of predictive maintenance. As transformer core materials age, their hysteresis loop characteristics change. The system establishes a database by long-term monitoring of key hysteresis loop parameters (such as coercivity, remanence, and maximum permeability) and analyzes their trends. Specifically, a modified Preisach model is used to describe the hysteresis characteristics, combined with a Long Short-Term Memory (LSTM) network for time-series prediction, to construct a core material aging status assessment model. The model inputs historical monitoring data and current operating parameters, and outputs the predicted future aging curve and remaining service life. Experiments show that the model's remaining service life prediction error is controlled within ±10%, providing a reliable basis for transformer maintenance decisions.

[0081] The predictive maintenance recommendation system generates detailed maintenance suggestions based on core aging predictions and historical operating data, utilizing a case-based reasoning decision support algorithm. The system first establishes a maintenance case library, recording optimal maintenance strategies and effectiveness evaluations under different aging conditions. Then, it extracts features from the current state and searches for similar cases in the library. Finally, based on the maintenance strategies of similar cases, it modifies them according to the specific current situation to form personalized maintenance recommendations. These recommendations include three aspects: the optimal maintenance time window, where the system recommends the most suitable maintenance time based on load forecasts and aging rates, avoiding peak power consumption periods and minimizing downtime impact; a list of critical components requiring replacement, where the system accurately identifies components that need replacement based on aging condition assessments, avoiding over-maintenance or under-maintenance; and optimized operating parameters for specific aging conditions, providing temporary operating parameter adjustment suggestions to slow down the aging process when immediate maintenance is not possible.

[0082] This predictive maintenance strategy offers significant advantages over traditional periodic maintenance: reduced unnecessary downtime and increased equipment utilization; extended equipment lifespan and lower maintenance costs; and early detection of potential problems to prevent sudden failures. Practical applications show that predictive maintenance can reduce transformer maintenance costs by approximately 20% while extending lifespan by 8-15%.

[0083] In summary, the adaptive loss minimization and predictive maintenance integrated system, through advanced optimization algorithms, harmonic compensation technology, and aging prediction models, achieves optimized management of the transformer core throughout its entire lifecycle. This not only improves short-term operating efficiency but also ensures long-term reliability, representing the future development direction of intelligent operation and maintenance of transformers.

[0084] Example 6: like Figure 2 As shown, the specific steps for deploying a giant magnetoresistive sensor array at key locations in the iron core, constructing a temperature monitoring subsystem using thermocouples and infrared thermal imaging technology, and obtaining temperature distribution data with a monitoring accuracy of ±0.5℃ are as follows: S1: Based on high-precision real-time monitoring data, K-type thermocouple sensors are arranged in key parts of the iron core to measure local temperature changes in the iron core and obtain contact temperature measurement data. K-type thermocouple sensors are placed at key locations in the core, such as the core pillars, yoke, and joints. These sensors provide temperature measurements with an accuracy of ±0.5°C. The thermocouple sensors convert temperature into a voltage signal through the thermoelectric effect, which is then amplified, filtered, and transmitted to the data acquisition system. This contact-based measurement method provides precise temperature data for critical locations within the core.

[0085] S2: To supplement the contact temperature measurement data with infrared thermal imaging technology, a high-resolution infrared thermal imager is used to perform non-contact temperature monitoring on the iron core surface to obtain an image of the iron core surface temperature distribution. To obtain more comprehensive temperature distribution information, the system uses a high-resolution infrared thermal imager to scan the core surface and generate a surface temperature distribution image. The infrared thermal imager has a resolution of 640×480 pixels and a temperature resolution better than 0.05℃, enabling it to capture minute temperature differences on the core surface. This non-contact measurement method supplements areas that thermocouples cannot cover, providing more comprehensive temperature distribution information.

[0086] S3: Based on the temperature distribution image of the iron core surface, establish a temperature field numerical model, calculate the internal temperature distribution of the iron core using the finite element analysis method, and obtain three-dimensional temperature field distribution data; Based on acquired surface temperature data, combined with the core's material properties, geometry, and environmental conditions, the system establishes a three-dimensional temperature field numerical model. Using finite element analysis, the temperature distribution inside the core is calculated, yielding complete three-dimensional temperature field data. This method can extrapolate internal temperatures that are difficult to measure directly, providing a possibility for a comprehensive understanding of the core's temperature state.

[0087] S4: Perform temperature gradient analysis on the three-dimensional temperature field distribution data, identify temperature anomaly areas and hotspot locations, establish a temperature change trend prediction model, and obtain temperature distribution analysis results; The system performs temperature gradient analysis on three-dimensional temperature field data, calculates the rate and direction of temperature change, and identifies abnormal temperature regions and hotspot locations. It also establishes a temperature change trend prediction model, based on historical temperature data and current operating conditions, to predict temperature change trends over a future period, providing a basis for preventative control.

[0088] S5: Based on the temperature distribution analysis results, set the temperature monitoring alarm threshold. When the temperature exceeds the safe range, trigger the early warning signal to realize the timely detection and handling of temperature anomalies, and obtain temperature distribution data with a temperature monitoring accuracy of ±0.5℃.

[0089] The system sets temperature monitoring alarm thresholds based on the characteristics of the core material and safe operation requirements. When the detected temperature exceeds the safe range, the system triggers an early warning signal, alerting maintenance personnel to pay attention and take necessary cooling measures. Simultaneously, this temperature data is transmitted to the three-stage magnetic field optimization control module, serving as a crucial basis for adjusting the magnetic field distribution.

[0090] Specifically, Embodiment 6 of this application details the construction process of a temperature monitoring subsystem, which is an important component of a multi-level dynamic monitoring network. This system is responsible for accurately monitoring the temperature distribution of the iron core and providing reliable data for temperature compensation and overheating early warning.

[0091] The contact temperature measurement system employs K-type thermocouple sensors, strategically deployed at key locations within the iron core. These K-type thermocouples are made of chromium-aluminum alloy, offering a wide operating temperature range (-200℃ to 1260℃), high accuracy (±0.75%), and excellent linearity and stability. The sensors are primarily deployed in the core areas of the iron core, such as the center of the core column, the center of the yoke, and high-loss areas like joints. They also cover internal locations with poor heat dissipation, such as inside clamps and between laminations. Each thermocouple sensor is equipped with an individual signal conditioning circuit, including cold junction compensation, amplification, and filtering to ensure the accuracy of the measurement signal. Signal acquisition utilizes a 16-bit high-precision A / D converter with a temperature resolution better than 0.1℃, meeting the requirements for high-precision temperature monitoring.

[0092] The non-contact temperature monitoring system employs advanced infrared thermal imaging technology to comprehensively scan the iron core surface. The system uses a high-resolution infrared thermal imager with a pixel resolution of 640×480 and a thermal sensitivity better than 0.05℃, capable of capturing minute temperature differences on the iron core surface. The thermal imager utilizes autofocus and autocalibration technology to ensure image clarity and measurement accuracy. To obtain a complete temperature distribution on the iron core surface, the system features an electrically driven rotating platform, enabling the thermal imager to scan the iron core from multiple angles and construct a complete surface temperature distribution map through image stitching technology. The infrared thermal imaging system is also equipped with an automatic emissivity correction function, adjusting the emissivity parameter in real time based on the iron core surface material characteristics and environmental conditions to improve measurement accuracy.

[0093] A numerical model of the temperature field is crucial for obtaining a complete temperature distribution within the iron core. Due to technological and spatial limitations, it is impossible to install temperature sensors at every location inside the core. The system establishes a three-dimensional numerical model of the iron core's temperature field based on the finite element method. The model inputs include the core's geometry, material thermal properties (such as thermal conductivity and specific heat capacity), surface boundary conditions, and temperature measurements at known points. By solving the heat conduction equations, the system can calculate the temperature value at any location within the core, constructing a complete three-dimensional temperature field distribution. The model employs adaptive meshing technology, using a denser mesh in regions with large temperature gradients to improve computational accuracy; simultaneously, parallel computing technology is introduced to keep the computation time within one minute, meeting the requirements for near real-time monitoring.

[0094] Temperature gradient analysis is a crucial method for identifying potential hotspots and temperature anomalies. The system calculates the spatial gradient vector of the temperature field to identify areas of most dramatic temperature changes; these areas are typically potential hotspots or regions with blocked heat flow. Areas with temperature gradients exceeding a preset threshold (usually 5°C / cm) are marked as key monitoring targets. The system also analyzes the temporal rate of change of the temperature field to identify areas with abnormally rapid temperature increases, which may indicate problems such as localized overheating or insulation aging. Based on these analyses, the system establishes a temperature change trend prediction model, employing the Autoregressive Integral Moving Average (ARIMA) algorithm to predict the temperature change trends of various parts of the core over the next few hours, providing a basis for preventative control.

[0095] The temperature monitoring and early warning system is the last line of defense for ensuring the safe operation of transformers. Based on the characteristics of the core material and safe operating requirements, the system sets multiple temperature monitoring alarm thresholds. Generally, when the local temperature exceeds 85℃, the system will issue a preliminary warning, advising maintenance personnel to monitor the area; when the temperature exceeds 95℃, the system will issue a general warning, suggesting measures such as reducing the load or increasing cooling; when the temperature exceeds 105℃, the system will issue an emergency warning, requiring immediate emergency measures and, if necessary, initiating a protective shutdown procedure. Warning information is not only sent to the field control system but also pushed to the mobile terminals of maintenance personnel via a remote communication module to ensure timely response. Simultaneously, this temperature data is transmitted in real time to the three-stage magnetic field optimization control module, serving as a crucial basis for adjusting the magnetic field distribution and achieving closed-loop control.

[0096] The system is designed with comprehensive data storage and analysis capabilities. All temperature monitoring data is stored in a time-series database at a specific sampling frequency (normally once per minute, increasing to once every 10 seconds under abnormal conditions), forming a complete historical record. The system provides rich data visualization tools, including temperature distribution heatmaps, temperature change trend charts, and marking of abnormal temperature areas, helping maintenance personnel intuitively understand the temperature status of the core. The system also provides advanced data analysis functions, capable of uncovering hidden patterns in the temperature data, such as the relationship between load changes and temperature response, and the impact of ambient temperature on core temperature rise, providing data support for optimizing transformer operation strategies.

[0097] This multi-layered, multi-technical temperature monitoring subsystem achieves comprehensive and accurate monitoring of transformer core temperature, with an industry-leading accuracy of ±0.5℃. The system not only promptly detects temperature anomalies and prevents overheating damage, but also provides accurate data support for temperature compensation and optimized control of core losses, making it a crucial component of the intelligent low-loss optimization system for transformer cores.

[0098] Example 7: The specific steps for decomposing the total core loss of the real-time BH magnetization curve and obtaining the core loss component calculation results are as follows: S1: Based on the real-time BH magnetization curve, the area and shape parameters of the hysteresis loop are extracted, and the hysteresis loss coefficient is calculated. The hysteresis loss is proportional to the nth power of the magnetic induction intensity and the value of n ranges from 1.6 to 2.2, thus obtaining the hysteresis loss calculation data. S2: Perform eddy current loss analysis on the hysteresis loss calculation data. Based on the resistivity and thickness parameters of silicon steel sheet, calculate the eddy current loss, where the eddy current loss is proportional to the square of the frequency and the square of the magnetic induction intensity, and obtain the eddy current loss calculation data. S3: Based on the eddy current loss calculation data, analyze the additional losses caused by harmonics and magnetic domain effects, where the additional losses are proportional to the 1.5th power of the frequency and the 1.5th power of the magnetic induction intensity, and obtain the additional loss calculation data. S4: Perform loss correction processing on the additional loss calculation data, consider the influence of factors such as temperature, stress and aging on each loss component, establish a correction coefficient model, and obtain the corrected loss component data. S5: Based on the corrected loss component data, the hysteresis loss, eddy current loss and additional loss are superimposed to calculate the total core loss value, thus completing the core loss component calculation results.

[0099] By analyzing the real-time BH magnetization curve, the system first extracts the area and shape parameters of the hysteresis loop. The area of ​​the hysteresis loop directly corresponds to the hysteresis loss per unit volume of core material within one magnetization cycle. The system uses the improved Steinmetz equation to calculate the hysteresis loss, i.e., hysteresis loss Ph = kh·f·Bmn, where kh is the hysteresis loss coefficient, f is the frequency, Bm is the maximum magnetic flux density, and n is the Steinmetz exponent, which is adjusted between 1.6 and 2.2 according to different material properties. The system automatically determines the most suitable value of n by analyzing the shape of the hysteresis loop, thus improving the calculation accuracy.

[0100] Eddy current loss is the loss caused by the induced current flowing through the material in the iron core. The system analyzes this loss based on the classic eddy current loss calculation formula Pe=ke·f²·Bm²·d² / ρ, where ke is the eddy current loss coefficient, d is the thickness of the silicon steel sheet, and ρ is the material resistivity. Eddy current loss is proportional to the square of the frequency and the square of the magnetic flux density, and is also related to the thickness and resistivity of the silicon steel sheet. The system continuously updates the eddy current loss calculation results by monitoring changes in resistivity in real time (which is significantly affected by temperature).

[0101] Additional losses are mainly caused by the damping of magnetic domain wall movement and microscopic eddy currents, which are often ignored or simplified in traditional models. This system uses Pa = ka·f1.5·Bm1.5 to calculate additional losses, where ka is the additional loss coefficient. When the core is affected by harmonics, the additional losses increase significantly. The system uses a harmonic analyzer to detect the harmonic situation of the power grid, analyzes the contribution of each harmonic to the additional losses, and obtains more accurate calculation results for additional losses.

[0102] The system considers the effects of factors such as temperature, stress, and aging on various loss components and establishes a correction coefficient model. Increased temperature increases material resistivity, reducing eddy current losses, but simultaneously decreases permeability and increases hysteresis losses. Mechanical stress affects the magnetic domain structure and alters hysteresis characteristics. Material aging caused by long-term operation increases various losses. By establishing correction coefficient models for these factors, the system corrects the initially calculated loss values ​​to obtain results closer to reality.

[0103] Finally, the system superimposes the corrected hysteresis loss, eddy current loss, and additional loss to obtain the total core loss value Pt = Ph + Pe + Pa. This decomposition calculation method not only improves the accuracy of loss calculation but also provides a more targeted basis for subsequent optimization control, because different types of losses require different control strategies for optimization.

[0104] Example 7 details the design and implementation method of an intelligent load adaptation and core multi-condition collaborative optimization system. This system overcomes the limitations of traditional transformer optimization under a single load condition, achieving adaptive optimization of the transformer core under different load conditions, and significantly improving the overall operating efficiency of the transformer.

[0105] Load characteristic identification and classification is the first step of the system. Using high-precision current and voltage transformers installed on the primary and secondary sides of the transformer, the system collects load current and voltage data in real time at a sampling rate of 10kHz, enabling it to capture rapid load changes. Based on the collected data, the system extracts multi-dimensional load characteristics, including basic features (active power, reactive power, power factor, etc.), time-domain features (load mean, standard deviation, peak factor, etc.), and frequency-domain features (harmonic content, frequency distribution, etc.). These features comprehensively reflect the electrical characteristics and dynamic behavior of the load. The system uses an unsupervised learning algorithm (combining K-means clustering and hierarchical clustering) to automatically classify the loads, grouping similar load patterns together to form a load characteristic database. After long-term data accumulation, the system can typically classify loads into 5-8 typical categories, such as light load, heavy load, balanced load, unbalanced load, linear load, and nonlinear load.

[0106] For each type of load characteristic, the system constructs a dedicated core optimization model. Under light load conditions (load rate below 30%), core loss accounts for a high proportion of total loss, and the system focuses on optimizing hysteresis loss and controlling the excitation current to reduce the core magnetization. Under medium load conditions (load rate 30%-70%), the system adopts a balancing strategy, simultaneously optimizing core loss and load loss. Under heavy load conditions (load rate exceeding 70%), the system emphasizes temperature equalization control to prevent local overheating. For nonlinear loads, the system enhances harmonic suppression to reduce additional losses caused by harmonics. For fluctuating loads, the system employs a predictive control strategy to adjust the core operating state in advance to cope with upcoming load changes. This classification optimization method can reduce additional losses by 5-8% compared to general optimization algorithms, significantly improving the transformer's energy efficiency.

[0107] Load forecasting and dynamic programming are the core functions of the system. The system employs deep learning to build a load forecasting model, integrating Long Short-Term Memory (LSTM) networks and attention mechanisms to accurately predict load changes over the next 24 hours. The forecasting model fully considers time factors (such as intraday, weekly, and seasonal variations), historical load data, and external factors (such as weather, temperature, and special events), achieving a prediction accuracy of over 95% under normal conditions. Based on the load forecasting results, the system uses a dynamic programming algorithm to plan the optimal control strategy, dividing the 24 hours into multiple time periods and selecting the most suitable optimization mode for each period to minimize the total daily loss. This forward-looking control strategy avoids the "short-sightedness" problem of traditional methods, achieving globally optimal control on a 24-hour scale.

[0108] Smooth transition control under multiple operating conditions is a key technology to ensure stable system operation. When the load condition changes, abrupt changes in the core optimization strategy can cause transient fluctuations in the system, affecting power quality. The system is designed with a smooth transition mechanism based on fuzzy logic control and a gradually changing function to achieve smooth switching of the optimization strategy. Specifically, near the critical point of the load condition, the system does not immediately switch to the new optimization mode, but gradually adjusts the control parameters according to a preset smoothing function, allowing the system state to smoothly transition along the optimal trajectory. The smoothing function is designed to consider the load change rate, the core temperature change rate, and system stability requirements, ensuring both smooth transition and avoiding excessively prolonged transition time. Practical application shows that this smooth control mechanism can control voltage fluctuations during the switching process within ±0.2%, meeting power quality standards.

[0109] The system is designed with a comprehensive performance evaluation and self-optimization mechanism. By comparing core losses, temperature distribution, and voltage quality indicators before and after optimization, the system evaluates the optimization effect in real time and calculates the percentage improvement in energy efficiency and economic benefits. These evaluation results are not only displayed in real time but also stored in the system database as historical performance records. The system employs a reinforcement learning algorithm, treating each optimization control as an "attempt" and the optimization effect as a "reward signal." Through continuous trial and learning, the system gradually improves its optimization strategy under various loads. This self-learning mechanism enables the system to continuously adapt to changing operating environments and aging equipment characteristics, maintaining long-term high-efficiency operation.

[0110] The intelligent load adaptation and multi-condition collaborative optimization system for transformer cores enables "location-specific and time-specific" optimized control of the core, overcoming the limitations of traditional optimization methods. The system not only copes with complex and ever-changing load environments but also predicts future load changes and plans optimal control strategies in advance, minimizing core losses across all time periods and operating conditions. Practical application data shows that this system can save an additional 3-5% of energy consumption compared to traditional single-condition optimization methods, resulting in considerable annual economic benefits for large substations and industrial users.

[0111] Example 8: The specific steps for setting a preemption threshold and obtaining dynamic preemption threshold parameters for the three-stage task partitioning scheme are as follows: S1: Based on the three-stage task division scheme, the loss rate threshold is set to 110% of the standard value, the temperature threshold is 85℃, and the harmonic interference threshold is 5% for the conventional optimization stage, thus obtaining the threshold parameters for the conventional optimization stage. S2: Set the threshold parameters for the critical optimization stage for the threshold parameters of the conventional optimization stage. Set the loss rate threshold to 120% of the standard value, the temperature threshold to 95℃, and the harmonic interference threshold to 8% to obtain the threshold parameters for the critical optimization stage. S3: Based on the threshold parameters of the critical optimization stage, the loss rate threshold is set to 135% of the standard value, the temperature threshold is 105℃, and the harmonic interference threshold is 12% for the emergency control stage, thus obtaining the threshold parameters for the emergency control stage. S4: Design a dynamic adjustment algorithm for threshold parameters during the emergency control phase. Correct each threshold parameter in real time based on seasonal changes, load characteristics and historical operating data to obtain an adaptive threshold adjustment strategy. S5: Based on an adaptive threshold adjustment strategy, a threshold parameter database is established to record the optimal threshold combination under different operating conditions, thereby realizing intelligent optimization and self-learning functions of threshold parameters and obtaining dynamic preemption threshold parameters.

[0112] The system sets relatively lenient threshold parameters for the routine optimization phase: a loss rate threshold of 110% of the standard value, a temperature threshold of 85℃, and a harmonic interference threshold of 5%. When system operating parameters fall below these thresholds, routine optimization tasks are executed to improve the daily operating efficiency of the transformer. These threshold settings ensure that the system has sufficient margin for optimization adjustments and will not frequently trigger high-priority tasks.

[0113] The threshold parameters for the critical optimization phase are more stringent than those for the regular phase: the loss rate threshold is 120% of the standard value, the temperature threshold is 95℃, and the harmonic interference threshold is 8%. When the system operating parameters exceed the thresholds of the regular phase but are below the thresholds of the critical phase, the system will execute the critical optimization task, adopting a more aggressive control strategy to prevent further deterioration. The critical phase task has medium priority and can preempt the regular optimization task.

[0114] The threshold parameters for the emergency control phase represent the system's extreme operating conditions: the loss rate threshold is 135% of the standard value, the temperature threshold is 105℃, and the harmonic interference threshold is 12%. When the system operating parameters exceed the critical phase thresholds, an emergency control task will be immediately executed, taking decisive measures to pull the system back to a safe operating range. The emergency control task has the highest priority and can preempt all other tasks.

[0115] The system is designed with a dynamic adjustment algorithm that automatically adjusts threshold parameters based on changes in the environment and operating conditions. For example, during periods of high summer temperatures, the temperature threshold is appropriately lowered; during peak industrial load periods, the harmonic interference threshold is appropriately raised; and the loss rate threshold is adjusted accordingly based on analysis of equipment aging data. This dynamic adjustment capability makes the system more adaptable to complex and ever-changing operating environments.

[0116] The system has established a threshold parameter database, recording the optimal threshold combinations under different operating conditions. Through the accumulation and analysis of long-term operating data, the system can identify the most suitable threshold settings for different operating conditions and automatically apply these empirical values ​​when similar conditions recur. This self-learning function makes the system's preemptive decision-making increasingly accurate, reduces unnecessary task switching, and improves overall control efficiency.

[0117] Example 8 details the design and implementation method of a self-organizing magnetic field zoning control system. This system innovatively divides the transformer core into multiple control zones, implementing differentiated control strategies for different zones to achieve refined management of the core's magnetic field, further reducing core losses and improving transformer operating efficiency.

[0118] The core-zoned control architecture is the core design concept of this system. Traditional transformers treat the entire core as a single control object, making it difficult to address the differentiated needs of different regions within the core. This system, based on the core's structural characteristics and magnetic field distribution patterns, divides the core into multiple control zones, typically including the core column center zone, core column edge zone, yoke center zone, yoke edge zone, and joint zone. Each zone is equipped with an independent sensor array and control unit, forming a self-organizing control network. The zoning considers both the natural boundaries of the physical structure and the differences in magnetic flux density distribution, ensuring a relatively uniform magnetic field distribution within each zone, facilitating the implementation of a unified control strategy.

[0119] Identifying the magnetic field characteristics of each zone is fundamental to implementing differentiated control. The system conducts in-depth analysis of the magnetic field characteristics of each zone, extracting key characteristic parameters, including average magnetic flux density, magnetic flux density gradient, magnetic field direction consistency, and magnetic field time stability. Based on these characteristics, the system classifies each zone, such as high magnetic flux density regions, magnetic field gradient transition regions, and magnetic field disorder regions. The system also analyzes the loss characteristics of each zone, including the proportion of hysteresis loss, eddy current loss, and additional loss, serving as an important basis for optimized control. This zone characteristic identification method can reveal subtle differences in the magnetic field distribution within the iron core, providing a scientific basis for refined control.

[0120] The system implements differentiated control strategies for zones with varying characteristics. For high flux density zones (such as the center of the core column), the system focuses on controlling the magnetic induction intensity to prevent local magnetic saturation. For magnetic field gradient transition zones (such as the connection between the core column and the yoke), the system optimizes flux transfer efficiency to reduce losses caused by increased magnetic reluctance. For areas with disordered magnetic fields (such as joints and lamination defects), the system enhances local magnetic field shaping to suppress abnormal eddy currents. Control parameters for each zone, such as the magnitude, direction, and waveform of the auxiliary excitation current, are independently optimized to form a zone-specific control scheme. This zone-specific differentiated control can address specific problems in different parts of the core, resulting in a more significant effect than a globally unified control.

[0121] A zoned control coordination mechanism is crucial for ensuring overall system optimization. While each zone has its own independent control unit, they do not operate in isolation but rather coordinate closely through a collaborative mechanism. The system establishes a collaborative control framework between zones, including information sharing mechanisms, boundary condition coordination, and maintenance of global goal consistency. Control units in adjacent zones exchange status information in real time via a high-speed communication network, including magnetic field strength, temperature, and control actions, ensuring that control decisions are based on complete information. At zone boundaries, the system implements special boundary coordination control to ensure a smooth transition of the magnetic field and avoid local magnetic field distortion caused by control incoordination. The system also establishes a global coordination controller responsible for maintaining the consistency of overall control goals, coordinating control actions across zones, and resolving potential conflicts.

[0122] Self-organizing learning and evolution are advanced features of the system. The system treats each partition control unit as an agent, constructing a multi-agent self-organizing control network. Each agent not only executes a preset control strategy but also continuously learns and evolves through operational experience. The learning process is based on the Q-learning algorithm and an experience replay mechanism. By trying different control actions and observing the system response, agents gradually establish a state-action-reward mapping and optimize their own control strategies. The evolution mechanism is based on genetic algorithms, allowing successful control strategies from different partitions to exchange and learn from each other, eliminating ineffective strategies, retaining and strengthening effective strategies, and achieving collective intelligence improvement across the entire control network. This self-organizing learning enables the system to adapt to equipment aging and changes in the operating environment, maintaining long-term optimization performance.

[0123] The evaluation of zone control effectiveness employs a multi-index comprehensive evaluation method. The system assesses the control effectiveness of each zone from four dimensions: loss reduction, temperature uniformity, control stability, and response speed. Loss reduction is calculated by comparing zone losses before and after optimization; temperature uniformity is evaluated using the temperature standard deviation within the zone; control stability is measured by the fluctuation amplitude during the control process; and response speed is determined by the time interval between the control action and the manifestation of the effect. Based on these four indicators, the system calculates a comprehensive score, periodically ranks the control strategies of each zone, and identifies best practices and areas for improvement. The evaluation results are not only used for real-time display but also serve as important feedback signals for self-organizing learning, driving continuous system optimization.

[0124] Practice has shown that the self-organizing magnetic field zoning control system can reduce core losses by an additional 3-6% compared to traditional global unified control, especially under non-uniform load and non-sinusoidal excitation conditions, where the effect is even more significant. Through refined zoning and differentiated control, the system achieves "locally adapted" core magnetic field management, representing an advanced development direction for transformer core optimization control.

[0125] Example 9: The specific steps for deploying an auxiliary excitation coil matrix at key locations in the iron core to obtain precise control of magnetic flux density are as follows: S1: Based on the global magnetic field distribution mode, multi-turn auxiliary excitation coils are arranged at key positions such as the core column, yoke and clamps to form a three-dimensional spatially distributed coil matrix network, thus obtaining the spatial layout scheme of the auxiliary excitation coils. S2: Design a current control algorithm for the spatial layout scheme of the auxiliary excitation coil, configure an independent current regulator for each coil, achieve precise current control within the range of ±10A, and obtain the coil current control parameters; S3: Based on the coil current control parameters, establish a mathematical model of magnetic field superposition, calculate the spatial distribution and superposition effect of the magnetic field generated by each coil, and obtain the magnetic field distribution calculation results; S4: Optimize the magnetic field distribution calculation results by using the least squares method and constraint optimization algorithm to solve for the optimal coil current combination under the voltage stability requirements, and obtain the optimal excitation current scheme. S5: Based on the optimal excitation current scheme, the current of each coil is precisely controlled by the PWM power drive circuit, the change of magnetic flux density is monitored in real time and closed-loop feedback is performed to obtain the precise control result of magnetic flux density.

[0126] Based on the structural characteristics and magnetic circuit distribution of the iron core, the system arranges multi-turn auxiliary excitation coils at key locations such as the core columns, yoke, and clamps. These coils are wound with high-temperature resistant insulated wire, and the number of turns is determined according to the required magnetic field strength and positional characteristics, typically between 20 and 100 turns. The coils are arranged to form a three-dimensional spatial matrix network, ensuring precise control of the magnetic field at each key part of the iron core.

[0127] The system is equipped with an independent current regulator for each excitation coil, achieving current control with an accuracy of 0.01A within a ±10A range based on PWM technology and an H-bridge circuit. The current control algorithm employs PID control combined with fuzzy logic, enabling rapid response to control commands while maintaining stable output characteristics. The system also includes current limiting protection and temperature monitoring functions to ensure the coils operate within safe limits.

[0128] The system establishes a mathematical model for magnetic field superposition, and calculates the spatial distribution and superposition effect of the magnetic fields generated by each coil based on finite element analysis and magnetic circuit theory. The model considers the nonlinear characteristics of the core material, edge effects, and mutual inductance between adjacent coils, and can accurately predict the magnetic field distribution under any combination of coil currents.

[0129] The system employs the least squares method and a constraint optimization algorithm to solve for the optimal coil current combination. The optimization objective is to make the actual magnetic field distribution as close as possible to the target distribution, while simultaneously satisfying voltage stability requirements and coil current constraints. The system uses an improved gradient descent algorithm combined with a penalty function to handle the constraints, enabling it to complete the optimization solution within milliseconds.

[0130] The system precisely controls the current in each coil through a PWM power drive circuit, achieving precise regulation of magnetic flux density. Simultaneously, a giant magnetoresistive sensor array monitors changes in magnetic flux density in real time, comparing the actual measured values ​​with the target values ​​to generate error signals. The closed-loop feedback control system continuously adjusts the coil currents based on these error signals, gradually bringing the actual magnetic field distribution closer to the target distribution, ultimately achieving precise regulation of magnetic flux density.

[0131] Example 9 details the design and implementation method of a two-way interactive artificial intelligence-assisted decision-making system. This system overcomes the limitations of traditional automatic control systems that separate from manual operation, establishing a human-machine collaborative intelligent decision-making framework. It leverages the computing power and data processing advantages of artificial intelligence systems while retaining the experience, judgment, and comprehensive decision-making capabilities of human maintenance personnel, thus achieving an intelligent upgrade of transformer core optimization control.

[0132] The multimodal human-machine interface (HMI) serves as the system's front-end entry point. The system features a multi-layered, multi-dimensional user interface to meet the needs of diverse user groups. For general maintenance personnel, an intuitive graphical interface is provided, including visualization of core status, real-time monitoring of key parameters, and operation guidance. For technical professionals, detailed data analysis tools are offered, including advanced functions such as historical data query, trend analysis, and anomaly diagnosis. For management decision-makers, comprehensive performance reports are provided, including energy efficiency analysis, economic benefit assessment, and optimization suggestions. The system supports multiple interaction methods, including traditional keyboard and mouse operation, touchscreen operation, voice command recognition, and even augmented reality (AR) interaction, enabling users to communicate with the system in the most natural way. The HMI adopts a responsive design, automatically adapting to different terminal devices, such as control room screens, workstations, tablets, and mobile phones, ensuring convenient system access for users in any scenario.

[0133] The knowledge graph and expert experience base form the system's knowledge foundation. The system constructs a comprehensive transformer core knowledge graph, encompassing equipment knowledge (such as transformer structure, material properties, and design parameters), domain knowledge (such as electromagnetic theory, loss mechanisms, and control principles), and practical knowledge (such as fault cases, optimization experiences, and maintenance techniques). The knowledge graph is constructed using an ontology approach, establishing a three-layer structure of concepts, relationships, and instances, forming a semantically rich knowledge network. The system also designs an expert experience extraction and representation mechanism. Through structured interviews, case studies, and knowledge mining, the tacit knowledge of senior experts is transformed into explicit rules and cases, stored in the experience base. The knowledge graph and experience base are not static resources but dynamically updated, active knowledge entities. The system continuously enriches and optimizes its knowledge content through machine learning and knowledge mining, forming a continuously evolving knowledge foundation.

[0134] The knowledge-based reasoning engine is the thinking center of the system. The system integrates rule-based and case-based reasoning to construct a hybrid reasoning engine. Rule-based reasoning is suitable for clear causal relationships and standard processing procedures, such as fault diagnosis and safety checks; case-based reasoning is suitable for complex and ambiguous situations, such as optimization strategy selection and handling unconventional faults. The reasoning process employs an uncertainty-based reasoning method, combining Bayesian networks and fuzzy logic to handle incomplete information and uncertainty in the real world. The system is designed with an explanation mechanism that clearly displays the reasoning path and decision-making basis, enabling users to understand the system's decision-making logic and enhancing their trust. The reasoning engine not only supports automatic reasoning but also human-computer interactive reasoning, allowing users to input professional judgments and supplementary information during the reasoning process, achieving a deep integration of human and machine intelligence.

[0135] Decision support and auxiliary control are the core functions of the system. The system provides three levels of decision support: information-level support, offering comprehensive, accurate, and timely information such as equipment status, historical data, and similar cases to help users grasp the basic information needed for decision-making; analysis-level support, providing in-depth data analysis and predictive models such as trend analysis, scenario simulation, and risk assessment to help users understand problems and anticipate consequences; and suggestion-level support, providing clear action suggestions such as parameter optimization, control strategies, and maintenance plans to directly guide user operations. The system is designed with a human-machine collaborative decision-making mechanism, dynamically adjusting the level of automation based on the complexity and risk level of the situation. For routine, low-risk decisions, the system can execute automatically; for complex, high-risk decisions, the system provides suggestions, with humans making the final decision; for decisions of medium complexity and risk, the system adopts a "suggestion-confirmation" model, where the system makes suggestions, which are then confirmed and executed by humans.

[0136] The system's key features are its continuous learning and self-improvement capabilities. It continuously learns and updates its knowledge base by recording the decision-making behaviors and feedback from human experts. The learning process combines supervised learning and reinforcement learning: supervised learning builds a basic model by imitating expert decisions; reinforcement learning adjusts and optimizes the model based on feedback from actual operations. The system incorporates a knowledge update mechanism, including new knowledge discovery, knowledge conflict detection and resolution, and knowledge validity verification, ensuring the accuracy and timeliness of the knowledge base. The system also implements anomaly analysis; when there are significant discrepancies between human decisions and system recommendations, the system analyzes the reasons and records the results as important cases for future learning. Through this continuous learning mechanism, the system can continuously accumulate experience, improve decision-making accuracy, and gradually evolve from a simple auxiliary tool into a reliable decision-making partner.

[0137] Practice has shown that the two-way interactive AI-assisted decision-making system significantly improves the intelligence level of transformer core optimization control. The system not only reduces operational complexity and alleviates the cognitive burden on maintenance personnel, but also improves decision-making quality and response speed, performing particularly well under complex and abnormal conditions. More importantly, the system enables the effective transfer and accumulation of knowledge, transforming expert experience into organizational assets and solving the "knowledge gap" problem faced by the power industry. This human-machine collaborative intelligent decision-making model represents the future development direction of industrial control systems and is expected to be applied in a wider range of fields.

[0138] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0139] It should be noted that those skilled in the art can make various modifications and variations to this invention without departing from the spirit and scope of this invention. If such modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include such modifications and variations.

[0140] This disclosure also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the steps of the intelligent low-loss optimization system and method for transformer cores described in the above-described method embodiments. The storage medium can be either volatile or non-volatile computer-readable storage.

[0141] In addition, this disclosure also provides a computer program product, which stores a computer program. When the computer program is run by a processor, it executes the steps of the intelligent low-loss optimization system and method for transformer core provided in any of the above embodiments of this disclosure. For details, please refer to the above method embodiments, which will not be repeated here.

[0142] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium, which can be a volatile or non-volatile computer-readable storage medium. In another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0143] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and apparatuses described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this disclosure, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0144] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0145] In addition, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0146] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0147] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A method for intelligent low-loss optimization of transformer cores, characterized in that, Includes the following steps: The structural characteristics and magnetic circuit distribution information of the transformer core are obtained, and a giant magnetoresistive sensor array is deployed at key locations in the core to obtain a multi-dimensional monitoring data stream that includes magnetic field distribution, temperature distribution and harmonic interference. The multidimensional monitoring data stream is subjected to pulse magnetization test to obtain the BH magnetization curve and loss characteristics of the core material in real time. A multidimensional loss model of temperature-frequency-magnetic induction intensity is constructed to obtain the real-time characteristic parameter set of the core material. Based on the real-time characteristic parameter set of the core material, the magnetic field optimization is divided into three stages: conventional optimization, critical optimization and emergency control. A preemption threshold is set and limited preemption control is implemented to obtain the magnetic field optimization control instruction set. The magnetic field optimization control instruction set is reconfigurable unit encoded, and the magnetic flux density is precisely controlled through the auxiliary excitation coil matrix coordinated control to obtain the optimal magnetic field distribution scheme; Based on the optimal magnetic field distribution scheme, a hybrid optimization strategy of genetic algorithm and particle swarm optimization is used to adaptively minimize losses. Combined with the analysis of hysteresis loop variation trend, a core aging prediction model is established to obtain comprehensive optimized control results and predictive maintenance suggestions.

2. The method according to claim 1, characterized in that, A giant magnetoresistive sensor array is deployed at key locations in the iron core to obtain a multi-dimensional monitoring data stream containing magnetic field distribution, temperature distribution, and harmonic interference, including: Based on the structural characteristics and magnetic circuit distribution information of the transformer core, a giant magnetoresistive sensor array with a measurement accuracy of 0.01T is deployed at key locations of the core to obtain high-precision real-time monitoring data of the magnetic field inside the core. Temperature distribution monitoring is performed on the high-precision real-time monitoring data. A temperature monitoring subsystem is constructed using thermocouples and infrared thermal imaging technology to obtain temperature distribution data with a temperature monitoring accuracy of ±0.5℃. Based on the temperature distribution data, the power grid voltage waveform is collected in real time by a harmonic analyzer and a voltage waveform monitoring device to analyze the harmonic content and distribution characteristics and obtain harmonic interference detection data. The harmonic interference detection data is acquired synchronously from multiple sources using time synchronization technology. The heterogeneous data is preprocessed and fused using a multi-sensor data fusion algorithm to obtain the multi-dimensional monitoring data stream containing magnetic field distribution, temperature distribution, and harmonic interference.

3. The method according to claim 1, characterized in that, The multidimensional monitoring data stream is subjected to pulse magnetization testing to construct a multidimensional loss model of temperature-frequency-magnetic induction intensity, thereby obtaining a real-time characteristic parameter set of the core material, including: Based on the multidimensional monitoring data stream, a pulse magnetization signal of a specific waveform is applied to the iron core, and magnetic field strength and magnetic induction intensity data are collected simultaneously to obtain the real-time BH magnetization curve. The real-time BH magnetization curve is decomposed into total core loss, and the hysteresis loss, eddy current loss and additional loss are calculated separately to obtain the core loss component calculation results. Based on the calculation results of the core loss components, a model is established to show the influence of temperature on permeability, hysteresis loss coefficient and resistivity, so as to realize the temperature correction of the core material properties and obtain the temperature-corrected material property parameters. The temperature-corrected material property parameters are fused to construct a multi-dimensional loss model that includes temperature, frequency, and magnetic induction intensity. The model accuracy is improved through small-sample testing and iterative optimization to obtain the real-time property parameter set of the core material.

4. The method according to claim 1, characterized in that, Based on the real-time characteristic parameter set of the core material, limited preemption control is implemented to obtain a magnetic field optimization control command set, including: Based on the real-time characteristic parameter set of the core material, and combined with the system operating status, the magnetic field optimization control is divided into three stages: conventional optimization, critical optimization and emergency control. Corresponding triggering conditions and optimization objectives are set to obtain a three-stage task division scheme. A preemption threshold is set for the three-stage task division scheme, including a loss rate threshold, a temperature threshold, and a harmonic interference threshold. The threshold is dynamically adjusted according to the system operating status to obtain dynamic preemption threshold parameters. Based on the dynamic preemption threshold parameter, when a high-priority task request is detected, it is determined whether to interrupt the currently executing low-priority task, implement limited preemption control to avoid wasting system resources, and obtain the preemption control execution instruction. A smooth switching process is designed for the preemptive control execution instructions to ensure stable system operation during the switching process between different optimization stages, avoid voltage fluctuations and transient processes caused by the switching process, and obtain the magnetic field optimization control instruction set.

5. The method according to claim 1, characterized in that, The magnetic field optimization control instruction set is reconfigured into reconfigurable units to obtain the optimal magnetic field distribution scheme, including: Based on the magnetic field optimization control instruction set, multiple independently controllable magnetic field control units are designed. Each unit includes an auxiliary excitation coil, a power drive circuit, and a local controller, resulting in a reconfigurable magnetic field control unit group. The reconfigurable magnetic field control unit group is encoded, and a unique code is assigned to each magnetic field control unit. Based on the optimization objective and the core state, the global magnetic field distribution mode is obtained by combining different unit activation modes. Based on the global magnetic field distribution pattern, an auxiliary excitation coil matrix is ​​deployed at key locations in the core. The magnitude and direction of the current in each coil are controlled according to the instructions of the magnetic field control unit, resulting in precise control of the magnetic flux density. The results of the precise control of magnetic flux density are monitored in real time. The actual magnetic field distribution is compared with the target distribution. Based on the deviation, closed-loop control is implemented to continuously optimize the magnetic field distribution and obtain the optimal magnetic field distribution scheme.

6. The method according to claim 2, characterized in that, A temperature monitoring subsystem was constructed using thermocouples and infrared thermal imaging technology to obtain temperature distribution data with an accuracy of ±0.5℃, including: Based on the high-precision real-time monitoring data, K-type thermocouple sensors are arranged in key parts of the core to measure the local temperature changes of the core and obtain contact temperature measurement data. The contact temperature measurement data is supplemented by infrared thermal imaging technology. A high-resolution infrared thermal imager is used to perform non-contact temperature monitoring on the iron core surface to obtain an image of the temperature distribution on the iron core surface. Based on the temperature distribution image of the core surface, a temperature field numerical model is established, and the internal temperature distribution of the core is calculated by the finite element analysis method to obtain three-dimensional temperature field distribution data. Temperature gradient analysis is performed on the three-dimensional temperature field distribution data to identify temperature anomaly areas and hotspot locations, a temperature change trend prediction model is established, and the temperature distribution analysis results are obtained. Based on the temperature distribution analysis results, a temperature monitoring alarm threshold is set. When the temperature exceeds the safe range, an early warning signal is triggered, enabling timely detection and handling of temperature anomalies, and obtaining temperature distribution data with a temperature monitoring accuracy of ±0.5℃.

7. The method according to claim 3, characterized in that, The real-time BH magnetization curve is decomposed into total core loss to obtain the core loss component calculation results, including: Based on the real-time BH magnetization curve, the area and shape parameters of the hysteresis loop are extracted, and the hysteresis loss coefficient is calculated. The hysteresis loss is proportional to the nth power of the magnetic induction intensity and the value of n ranges from 1.6 to 2.2, thus obtaining the hysteresis loss calculation data. Eddy current loss analysis is performed on the hysteresis loss calculation data. Based on the resistivity and thickness parameters of the silicon steel sheet, the eddy current loss is calculated, where the eddy current loss is proportional to the square of the frequency and the square of the magnetic induction intensity, and the eddy current loss calculation data is obtained. Based on the eddy current loss calculation data, the additional losses caused by harmonics and magnetic domain effects are analyzed, wherein the additional losses are proportional to the 1.5th power of the frequency and the 1.5th power of the magnetic induction intensity, and the additional loss calculation data is obtained. The additional loss calculation data is subjected to loss correction processing. Taking into account the influence of factors such as temperature, stress and aging on each loss component, a correction coefficient model is established to obtain the corrected loss component data. Based on the corrected loss component data, the hysteresis loss, eddy current loss, and additional loss are superimposed to calculate the total core loss value, thus completing the core loss component calculation result.

8. The method according to claim 4, characterized in that, A preemption threshold is set for the three-stage task partitioning scheme to obtain dynamic preemption threshold parameters, including: Based on the three-stage task division scheme, the loss rate threshold is set to 110% of the standard value, the temperature threshold is 85℃, and the harmonic interference threshold is 5% for the conventional optimization stage, thus obtaining the threshold parameters for the conventional optimization stage. The threshold parameters of the conventional optimization stage are set to the threshold parameters of the critical optimization stage. The loss rate threshold is set to 120% of the standard value, the temperature threshold is set to 95℃, and the harmonic interference threshold is set to 8%, thus obtaining the threshold parameters of the critical optimization stage. Based on the threshold parameters of the critical optimization stage, the loss rate threshold is set to 135% of the standard value, the temperature threshold is 105℃, and the harmonic interference threshold is 12% for the emergency control stage, thus obtaining the threshold parameters for the emergency control stage. A dynamic adjustment algorithm is designed for the threshold parameters in the emergency control phase. The threshold parameters are corrected in real time according to seasonal changes, load characteristics and historical operation data to obtain an adaptive threshold adjustment strategy. Based on the adaptive threshold adjustment strategy, a threshold parameter database is established to record the optimal threshold combination under different operating conditions, thereby realizing the intelligent optimization and self-learning function of the threshold parameters and obtaining the dynamic preemption threshold parameters.

9. The method according to claim 5, characterized in that, By deploying an auxiliary excitation coil matrix at key locations in the iron core, precise control of magnetic flux density is achieved, including: Based on the global magnetic field distribution pattern, multi-turn auxiliary excitation coils are arranged at key positions such as the core column, yoke, and clamps to form a three-dimensional spatially distributed coil matrix network, thus obtaining the spatial layout scheme of the auxiliary excitation coils. A current control algorithm is designed for the spatial layout scheme of the auxiliary excitation coil. An independent current regulator is configured for each coil to achieve precise current control within the range of ±10A, and the coil current control parameters are obtained. Based on the coil current control parameters, a magnetic field superposition mathematical model is established to calculate the spatial distribution and superposition effect of the magnetic field generated by each coil, and the magnetic field distribution calculation results are obtained. The magnetic field distribution calculation results are optimized by using the least squares method and constraint optimization algorithm to solve for the optimal coil current combination under the condition of satisfying voltage stability requirements, and obtain the optimal excitation current scheme. Based on the optimal excitation current scheme, the current of each coil is precisely controlled by the PWM power drive circuit, the change in magnetic flux density is monitored in real time and closed-loop feedback adjustment is performed to obtain the precise control result of magnetic flux density.

10. A transformer core intelligent low-loss optimization device, characterized in that, include: The multi-level dynamic monitoring module is used to deploy a giant magnetoresistive sensor array at key locations in the iron core to acquire a multi-dimensional monitoring data stream that includes magnetic field distribution, temperature distribution, and harmonic interference. The material property identification module is used to perform pulse magnetization testing on the multidimensional monitoring data stream, construct a multidimensional loss model of temperature-frequency-magnetic induction intensity, and output a real-time characteristic parameter set of the core material. The three-stage magnetic field optimization control module is used to implement a threshold-driven three-stage magnetic field optimization control strategy based on the real-time characteristic parameter set of the core material and output a magnetic field optimization control instruction set. The reconfigurable magnetic field control module is used to encode the magnetic field optimization control instruction set into reconfigurable units, and outputs the optimal magnetic field distribution scheme through the coordinated control of the auxiliary excitation coil matrix. The adaptive optimization and maintenance module is used to minimize adaptive losses based on the optimal magnetic field distribution scheme using a hybrid optimization strategy, establish a core aging prediction model, and output comprehensive optimization control results and predictive maintenance suggestions.

Images