A high frequency voltage frequency control method and system

Abstract

The application discloses a high-frequency voltage frequency control method and system, and relates to the technical field of high-frequency voltage control.The technical points are as follows: by constructing a time superposition residual influence field of historical multi-batch adjustment operations, the residual adjustment effect hidden in the transformer and unable to be captured by conventional detection means is converted into a quantifiable comprehensive field strength index, and a layer-by-layer reverse correction mechanism is introduced, which can effectively offset the nonlinear cumulative error generated by the superposition of multiple adjustments, and avoid the frequency oscillation and output instability problems easily occurring in the traditional control mode; at the same time, the basic adjustment period is determined based on the change characteristics of different operating parameters, and all periods are dynamically scaled in combination with the strength of the current residual influence to generate differentiated frequency adjustment periods, which can adapt to the difference in change rate of different parameters, maintain a high response speed for parameters changing quickly, and appropriately reduce the acquisition and adjustment frequency for parameters changing slowly.

Description

Technical Field

[0001] This invention relates to the field of high-frequency transformer control technology, and in particular to a high-frequency transformer frequency control method and system. Background Technology

[0002] High-frequency transformers, as core energy conversion components of power electronic equipment, are widely used in industrial and civilian fields such as switching power supplies, new energy power generation converters, and electric vehicle charging devices. The stability and accuracy of their frequency control directly determine the power conversion efficiency, output voltage ripple index, and lifespan of the entire equipment. The conventional implementation process of existing high-frequency transformer frequency control technology is as follows: The system collects basic operating parameters such as input voltage and output current at a fixed sampling period, compares the collected real-time parameter values ​​with preset target values, calculates the frequency adjustment amount using conventional control algorithms such as proportional-integral-derivative, and then immediately sends a control signal to the switching device to execute the frequency adjustment. The entire control process revolves entirely around the current operating state, and the adjustment period remains fixed once set, only being manually modified according to the operating conditions when the equipment is shut down for maintenance.

[0003] However, during the implementation of the above technical solution, at least the following technical problems were discovered: Firstly, the thermal inertia of the transformer windings, the hysteresis characteristics of the magnetic core, and the cumulative loss characteristics of the switching devices mean that the state changes caused by a single frequency adjustment will persist for tens of seconds and gradually decay. When the equipment is in a condition of frequent load fluctuations, the residual effects of multiple adjustments will superimpose on each other in the time dimension, forming a nonlinear cumulative error that cannot be predicted by conventional algorithms, ultimately causing problems such as frequency oscillation and increased output voltage ripple. Secondly, the input voltage and output current change rapidly, while the winding temperature and switching losses change relatively slowly. A fixed cycle either lags in responding to rapidly changing parameters or causes unnecessary waste of computational resources for slowly changing parameters. At the same time, it is impossible to dynamically adjust the control rhythm based on the strength of the residual effects of historical regulation. Therefore, we propose a high-frequency transformer frequency control method and system. Summary of the Invention

[0004] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a high-frequency transformer frequency control method and system, which solves the technical problem of nonlinear cumulative error caused by the superposition of historical adjustment residues in existing technologies.

[0005] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A high-frequency transformer frequency control method, the steps of which are as follows: Collect multi-dimensional operating status data during a single operation of a high-frequency transformer. The multi-dimensional operating status data includes input voltage, output current, winding temperature, and switching losses. Generate differentiated frequency adjustment cycles that adapt to different parameter changes based on multi-dimensional operational status data; Data is collected in batches according to the differentiated frequency adjustment cycle, and then integrated to obtain multiple batches of state sample sets; A dual analysis of structure and content is performed on multiple batches of state sample sets to calculate the frequency control quality assessment value, and an initial frequency control judgment value is generated based on the frequency control quality assessment value. Construct a four-dimensional time-stacked residual influence field corresponding to multiple historical frequency adjustment operations, and calculate the comprehensive field strength value of the time-stacked residual influence field at the current moment; Multi-order correction coefficients are generated based on the comprehensive field strength value, and the initial frequency control judgment value is corrected layer by layer in reverse to obtain the frequency control execution value; The frequency control execution value is compared with the preset threshold. If it is greater than the preset threshold, the current frequency adjustment strategy is executed; otherwise, the adjustment parameters are adjusted and the data acquisition and analysis process is restarted.

[0006] Preferably, the steps for collecting multi-dimensional operating status data during a single operation of the high-frequency transformer are as follows: Locate all accessible data acquisition nodes in the high-frequency transformer control system and acquire raw operating data through standardized interfaces; The raw operating data is subjected to feature classification and time synchronization processing to identify input voltage data, output current data, winding temperature data and switching loss data, forming a synchronized multi-dimensional operating status dataset.

[0007] Preferably, the step of generating a differentiated frequency adjustment period based on multi-dimensional operating status data to adapt to different parameter change characteristics is as follows: The corresponding basic adjustment cycle is determined based on the fluctuation characteristics of the input voltage, the load variation characteristics of the output current, the protection response characteristics of the winding temperature, and the frequency correlation characteristics of the switching loss. Obtain the current comprehensive field strength value, and dynamically scale all basic adjustment cycles proportionally according to the magnitude of the comprehensive field strength value; The scaled-up basic adjustment cycles are integrated to generate the final differentiated frequency adjustment cycle.

[0008] Preferably, the steps for performing batch data collection according to a differentiated frequency adjustment cycle and integrating it to obtain multiple batches of state sample sets are as follows: The data acquisition task is divided into multiple independent data acquisition tasks based on the differentiated frequency adjustment cycle, and corresponding acquisition parameters are configured for each task. Based on the collected parameters, locate the original data source corresponding to the multi-dimensional operating status data, and determine the collection sequence for each data source; Data collection and preliminary verification are performed according to the collection sequence. The four types of data within the same time window are correlated and integrated to form multiple batches of state sample sets.

[0009] Preferably, the step of performing a dual analysis of the structure and content of multiple batches of state sample sets to calculate the frequency control quality assessment value includes: The information continuity assessment value is calculated based on the data structure type of multiple batches of state sample sets. Quantitative evaluation of the content of multiple batches of state sample sets was performed to obtain content evaluation results for each dimension; The initial weight coefficients of each dimension are determined by the analytic hierarchy process, and the initial weights are nonlinearly corrected by combining the comprehensive field strength value at the current moment. The content information evaluation value is calculated based on the corrected weighting coefficients; By integrating the information continuity assessment value and the content information assessment value, the final frequency control quality assessment value is obtained.

[0010] Preferably, the steps for constructing the four-dimensional time-stacked residual influence field corresponding to multiple historical frequency adjustment operations and calculating the comprehensive field strength value of the influence field at the current moment are as follows: Extract a preset number of complete data sequences corresponding to historical frequency adjustment operations, with each sequence covering multi-dimensional operational status data for a preset duration before and after the adjustment execution; Perform temporal differencing on each complete data sequence to obtain a sequence of differences in the parameters of each dimension before and after the adjustment operation. Each change difference sequence is assigned a time weight that decays exponentially with time, where the time weight decreases the further away the adjustment operation is from the current time. Align all weighted change difference sequences along the time axis and perform four-dimensional vector superposition to construct a continuous time-stacked residual influence field. The residual influence field of time superposition is divided into grids according to preset rules to obtain several independent field strength calculation units; Calculate the local field strength value of each field strength calculation unit, and then sum the weighted values ​​of all local field strength values ​​to obtain the comprehensive field strength value at the current moment.

[0011] Preferably, the step of generating multi-order correction coefficients based on the comprehensive field strength value and performing layer-by-layer reverse correction is as follows: The overall correction amplitude is controlled by generating a first-order global correction coefficient based on the comprehensive field strength value. Within the constraints of the first-order global correction coefficient, based on the average field strength values ​​of each dimension in the four-dimensional time superposition residual influence field, corresponding second-order dimension correction coefficients are generated to compensate for the differences in residual influence in different dimensions. Within the constraints of each second-order dimension correction coefficient, a corresponding third-order time window correction coefficient is generated based on the rate of change of field strength in that dimension within a continuous preset time window to capture the dynamic changes of residual influence. The initial frequency control judgment value is multiplied sequentially by the third-order time window correction coefficient, the second-order dimension correction coefficient, and the first-order global correction coefficient to perform a layer-by-layer reverse correction, thereby obtaining the frequency control execution value.

[0012] A high-frequency transformer frequency control system, the control system comprising: The data acquisition module collects multi-dimensional operating status data during a single operation of the high-frequency transformer. This multi-dimensional operating status data includes input voltage, output current, winding temperature, and switching losses. The cycle generation module generates differentiated frequency adjustment cycles that adapt to different parameter change characteristics based on multi-dimensional operating status data. The sample construction module performs batch data collection according to the differentiated frequency adjustment cycle and integrates them to obtain multiple batches of state sample sets; The quality assessment module performs a dual analysis of the structure and content of multiple batches of state sample sets to calculate the frequency control quality assessment value. The initial judgment module generates initial frequency control judgment values ​​based on the frequency control quality assessment values. The residual influence field construction module constructs a four-dimensional time-stacked residual influence field corresponding to multiple batches of historical frequency adjustment operations, and calculates the comprehensive field strength value of the time-stacked residual influence field at the current moment. The correction execution module is used to generate multi-order correction coefficients based on the comprehensive field strength value, and to perform layer-by-layer reverse correction on the initial frequency control judgment value to obtain the frequency control execution value. The execution judgment module compares the frequency control execution value with a preset threshold. If the value is greater than the preset threshold, the current frequency adjustment strategy is executed; otherwise, the adjustment parameters are adjusted and the data acquisition and analysis process is restarted.

[0013] Preferably, the residual influence field construction module includes: The sequence extraction unit extracts a preset number of complete data sequences corresponding to historical frequency adjustment operations. Each sequence covers multi-dimensional operational status data for a preset duration before and after the adjustment execution. The difference processing unit performs time-series difference processing on each complete data sequence to obtain a sequence of differences in the parameters of each dimension before and after the adjustment operation. The time-weighted unit assigns a time weight that decays exponentially with time to each change difference sequence; the further away the adjustment operation is from the current time, the smaller the time weight. The vector overlay unit aligns all weighted change difference sequences along the time axis and performs four-dimensional vector overlay to construct a continuous time-overlay residual influence field. The time-stacked residual influence field is divided into several independent field strength calculation units according to a preset rule. The field strength calculation unit calculates the local field strength value of each field strength calculation unit, and performs a weighted summation of all local field strength values ​​to obtain the comprehensive field strength value at the current moment.

[0014] Preferably, the correction execution module includes: The first-order coefficient generation unit generates first-order global correction coefficients based on the comprehensive field strength value to control the overall correction amplitude. The second-order coefficient generation unit, within the constraint of the first-order global correction coefficient, generates corresponding second-order dimension correction coefficients to compensate for the differences in residual effects in different dimensions based on the average field strength values ​​of each dimension in the four-dimensional time superposition residual influence field. The third-order coefficient generation unit, within the constraint range of each second-order dimension correction coefficient, generates the corresponding third-order time window correction coefficient based on the field strength change rate of that dimension within a continuous preset time window to capture the dynamic changes of residual influence. The layer-by-layer correction unit is used to multiply the initial frequency control judgment value sequentially with the third-order time window correction coefficient, the second-order dimension correction coefficient, and the first-order global correction coefficient to perform layer-by-layer reverse correction and obtain the frequency control execution value.

[0015] (III) Beneficial Effects By constructing a time-superimposed residual influence field of historical batches of adjustment operations, the residual effects of adjustment, which were originally hidden inside the transformer and could not be captured by conventional detection methods, are transformed into a quantifiable comprehensive field strength index. Furthermore, a layer-by-layer reverse correction mechanism is introduced during the generation of the final control signal, effectively offsetting the nonlinear cumulative error generated by multiple adjustment superpositions. This ensures that the frequency adjustment always closely matches the actual operating state of the transformer, avoiding frequency oscillations and output instability problems that are prone to occur in traditional control modes. Simultaneously, based on the changing characteristics of different operating parameters, a basic adjustment cycle is determined, and all cycles are dynamically scaled according to the intensity of the current residual influence, generating differentiated frequency adjustment cycles. This adapts to the different rates of change of different parameters, maintaining a high response speed for rapidly changing parameters and appropriately reducing the acquisition and adjustment frequency for slowly changing parameters, thus reasonably optimizing the use of computational resources while ensuring control accuracy.

[0016] Secondly, the entire control process is completed through data processing logic, without requiring any modification to the existing high-frequency transformer hardware circuit. It can be directly deployed in various existing control systems, with low modification costs and strong compatibility. The historical data of residual influence fields accumulated during system operation can also provide data support for equipment status monitoring and daily maintenance, helping maintenance personnel to promptly detect abnormal changes in equipment operation and reduce the risk of sudden equipment failure. Attached Figure Description

[0017] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart of the overall method in this invention; Figure 2 This is a diagram showing the system's interconnections in this invention. Figure 3 This is a real-time trend graph of input voltage and output current in this invention; Figure 4 This is a real-time trend graph of winding temperature and switching loss in this invention. Detailed Implementation

[0019] This application provides a high-frequency transformer frequency control method and system to solve the technical problem of nonlinear cumulative error caused by the superposition of historical adjustment residues in the prior art.

[0020] Example 1: Taking a 10kW industrial-grade high-frequency switching power supply transformer as an example, a data acquisition unit is constructed using an analog-to-digital converter, a temperature sensor, an A current sensor, and a precision resistor voltage divider sampling circuit. A modular design combining front-end and back-end is adopted. The front-end is responsible for real-time data acquisition and control output, while the back-end is responsible for data processing and logical operations. Each module interacts with data through a shared memory buffer, avoiding data loss and delay during transmission. This also avoids common-mode interference (up to 2V) generated by inverter start-up / stop and motor commutation in industrial settings during actual debugging. Therefore, all analog signal transmission lines use twisted-pair shielded cables with single-end grounding of the shielding layer. A second-order RC low-pass filter circuit with a cutoff frequency of 10kHz is added to each input of the analog-to-digital converter to effectively filter out high-frequency noise superimposed on the signal. Isolated RS485 and Ethernet interfaces are also provided for communication with a host computer, enabling remote monitoring and parameter configuration. This allows maintenance personnel to view the transformer's operating status and adjust control parameters in real time from the central control room. All remote parameter modifications require authorization verification to prevent accidental operation. Figure 1 As shown, the details are as follows: First, the hardware initialization process is executed, including configuring the system clock to an external crystal oscillator clock, initializing the advanced timer TIM8 for time synchronization and PWM output, performing self-calibration and gain calibration on the ADS1256 analog-to-digital converter, automatically measuring and compensating for the ADC offset error and gain error during calibration to ensure sampling accuracy reaches 24 bits, and then performing self-tests on all sensors, including checking the connection status of the temperature sensor, whether the output of the current sensor is within the normal range, and whether the voltage sampling circuit is open. After the self-test passes, the main loop is entered; if the self-test fails, the corresponding hardware fault alarm is triggered and the operation stops.

[0021] After the main cycle starts, it first scans and locates all accessible data acquisition nodes in the high-frequency transformer control system, including the input DC bus voltage sampling circuit, the output AC current sampling circuit, the PT100 temperature sensor attached to the high-voltage winding surface, and the IGBT switching transistor loss calculation unit. Raw operating data is acquired through standardized SPI and ADC interfaces, with the sampling frequency uniformly set to 1kHz, meaning data is acquired every 1 millisecond to ensure the data's time resolution meets control requirements. After acquiring the raw operating data, it first performs feature classification and time synchronization processing, automatically identifying the data type based on the data's source port and format characteristics. Data from the voltage sampling circuit, characterized by a 100Hz DC ripple, is identified as input voltage data because the DC voltage after power frequency rectification will have a 100Hz ripple. Data from the current sampling circuit, dynamically varying with the load, is identified as output current data. Analog data from the PT100 temperature sensor, converted via AD, is identified as winding temperature data. Data calculated from the IGBT switch's on-state voltage drop and on-time is identified as switching loss data. Note that switching loss data cannot be directly measured by sensors; the collector-emitter on-state voltage drop V0 of the IGBT switch is acquired in real-time. ce and conduction time T on Combined with the current collector current I c And the switching frequency f, within the main control chip, each switching cycle is determined by the formula P=V ce ×I c ×T onThe calculation is performed once, where P is the conduction loss power of the IGBT switch. This calculation process only takes about 2 microseconds of CPU time and will not affect real-time performance. After identification, each data point is given a precise timestamp. Time synchronization uses the advanced timer TIM8 inside the microcontroller. Each count of the timer corresponds to 1 microsecond. Each time data is collected, the current timer count value is read as the timestamp to ensure that the four types of data collected at the same time can be accurately matched, forming a synchronized multi-dimensional operating status dataset. This provides a reliable data foundation for subsequent analysis and control. These four types of parameters correspond to the transformer's input energy, output load, thermal state, and device losses, which are the four core factors affecting the frequency control effect. Existing technologies often only collect the first two types of parameters, ignoring the impact of thermal inertia and loss accumulation on control, resulting in poor control performance under continuous load fluctuations.

[0022] After data acquisition and synchronization, based on the collected multi-dimensional operating status data, the corresponding basic adjustment cycle was determined according to the changing characteristics of four core parameters. Input voltage fluctuations are rapid and directly affect output voltage stability; the average fluctuation period is 95ms at 30% rated load and 105ms at 100% rated load, therefore the basic adjustment cycle is set to 100ms. Output current has a moderate load change rate; the current change period is approximately 180ms during production line equipment start-up and shutdown, and approximately 220ms during normal operation, therefore the basic adjustment cycle is set to 200ms. Winding temperature has a large thermal inertia, with a time constant of approximately 30 seconds; even with a sudden load change, the temperature takes tens of seconds to reach a new equilibrium, therefore the basic adjustment cycle is set to 500ms. Switching losses are mainly determined by frequency and temperature, and the change rate is slow, therefore the basic adjustment cycle is set to 1s. These basic adjustment cycles were determined after conducting a 7-day continuous operation test on this transformer model, statistically analyzing the average change cycle of each parameter under 30%-100% rated load conditions. In actual operation, the current... The comprehensive field strength value from the previous moment is dynamically scaled proportionally across all basic adjustment cycles according to the magnitude of the field strength. The scaling ratio is calculated using the formula S = 1 + 0.5 × E, where S is the scaling ratio of the adjustment cycle and E is the comprehensive field strength value at the current moment. The upper limit of the scaling ratio is 2.0, and the lower limit is 0.5, to prevent over-scaling from causing control failure. That is, the larger the comprehensive field strength value, the stronger the residual influence of historical adjustments, requiring an appropriate extension of the adjustment cycle to avoid frequent adjustments that could further accumulate residual influences. Conversely, the adjustment cycle should be shortened to improve response speed. For example, when the comprehensive field strength value... When the scaling factor is 0.6, the scaling ratio is 1 + 0.5 × 0.6 = 1.3. All basic adjustment cycles are multiplied by 1.3, resulting in scaled adjustment cycles of 130ms, 260ms, 650ms, and 1.3s. Finally, the four scaled adjustment cycles are integrated to generate the final differentiated frequency adjustment cycle, which is then sent to the data acquisition module. This dynamic scaling mechanism is completely absent in existing technologies. The adjustment cycle of existing technologies is fixed once set and cannot be adaptively adjusted according to historical operating conditions, leading to over-adjustment when residual effects are strong.

[0023] Upon receiving the differentiated frequency adjustment cycle information, it is broken down into four independent data acquisition tasks. Each task is configured with corresponding acquisition parameters, including acquisition range, accuracy, and triggering method. For example, the input voltage acquisition task has an acquisition range of 0-400V, an accuracy of 0.1V, and uses a timer triggering method with a triggering period of 130ms. The winding temperature acquisition task has an acquisition range of -40℃-150℃, an accuracy of 0.1℃, and also uses a timer triggering method with a triggering period of 650ms. Based on the configured parameters, the original data sources corresponding to the four core parameters are located, and the acquisition sequence for each data source is determined. The acquisition order is fixed as input voltage, output current, winding temperature, and switching loss. This is because the first two types of parameters change rapidly, and prioritizing their acquisition ensures data timeliness. The latter two types of parameters change slowly, and delaying their acquisition will not affect data accuracy. The acquisition is performed once per adjustment cycle, with the acquisition interval consistent with the corresponding adjustment cycle.

[0024] Data acquisition is performed according to a preset acquisition sequence. The acquired raw data undergoes preliminary verification, and outliers exceeding reasonable ranges are removed. The criteria for judging outliers include: input voltage exceeding the 0-400V range, output current exceeding the 0-20A range, winding temperature exceeding the -40℃-150℃ range, switching loss exceeding the 0-1000W range, and the jump amplitude between two consecutive data acquisitions exceeding a preset threshold. For example, an input voltage jump exceeding 50V or an output current jump exceeding 5A. Data meeting the above conditions is marked as outliers and discarded. The number of outliers is recorded. When three consecutive outliers are acquired, a sensor fault alarm is triggered. It then sends alarm information to the host computer; subsequently, it integrates the four types of data within the same time window to form multiple batches of status sample sets. Each sample set contains data for 10 consecutive adjustment cycles. The size of the time window can be adjusted by the host computer according to actual needs, generally set to 10-20 adjustment cycles. The sample set is stored in the format of a structure array. Each structure contains five fields: timestamp, input voltage, output current, winding temperature, and switching loss. After the sample set is constructed, it will be stored in the system's internal Flash. The Flash can store up to 1000 sample sets. Once the limit is reached, the oldest sample will be automatically overwritten to achieve cyclic storage.

[0025] A dual analysis of structure and content was performed on the constructed batches of state sample sets to calculate the frequency control quality assessment value. First, the data structure type of the sample set was analyzed; the input voltage, output current, and switching loss data were structured data, and the trends within 60 seconds were as follows: Figure 3 As shown, it has fixed fields and format; the winding temperature data is semi-structured data, and the trend within 60 seconds is as follows. Figure 4As shown, the dataset contains two fields: timestamp and temperature value. Based on the completeness and continuity of the data structure, an information continuity assessment value is calculated. If 10 consecutive samples are without missing or anomalies, the information continuity assessment value is 1.0. If missing or anomalies exist, the assessment value is reduced accordingly based on the proportion of missing samples. For example, one missing sample results in an assessment value of 0.9, two missing samples result in 0.8, and so on. Then, the sample set content is quantitatively evaluated from six preset dimensions: information completeness, accuracy, timeliness, clarity, relevance, and continuity. The assessment value for each dimension ranges from 0 to 1, with 1 indicating the best quality for that dimension. Information completeness is calculated based on the proportion of missing fields in the samples, and accuracy is calculated by comparing the results with a standard calibration source. Timeliness is calculated based on the deviation between data acquisition time and adjustment cycle; clarity is calculated based on the noise level in the data; relevance is calculated based on the correlation between the data and frequency control; and continuity is calculated based on the continuity of changes in adjacent data. Initial weight coefficients for the six dimensions were determined using the analytic hierarchy process (AHP): information completeness 0.2, accuracy 0.2, timeliness 0.2, clarity 0.1, relevance 0.1, and continuity 0.2. These initial weight coefficients were calculated by inviting five senior engineers in the field of power electronics to score the data, constructing a judgment matrix, and passing a consistency check. The consistency ratio was 0.08, less than 0.1, meeting the consistency requirements. The initial weights were then nonlinearly corrected based on the current comprehensive field strength value. The revised rule is that for every 0.1 increase in the overall field strength, the weights of the winding temperature and switching loss dimensions decrease by 0.05, while the weights of the input voltage and output current dimensions increase by 0.05. This allows for greater reliance on more stable electrical parameters when residual effects are strong, avoiding misleading control decisions by thermal and loss parameters that are heavily influenced by residual effects. For example, when the overall field strength is 0.6, the weights of the winding temperature and switching loss dimensions decrease to 0.15 and 0.15 respectively, while the weights of the input voltage and output current dimensions increase to 0.25 and 0.25 respectively. The weights of the clarity and relevance dimensions remain unchanged. Based on the revised weight coefficients, the content evaluation results of each dimension are weighted and summed to obtain the content information. The evaluation values ​​are then combined with the information continuity evaluation value and the content information evaluation value in a 1:1 ratio to obtain the final frequency control quality evaluation value. Based on the calculated frequency control quality evaluation value, and combined with the preset improved proportional-integral-derivative algorithm as the basic adjustment algorithm, an initial frequency control judgment value is generated. The frequency control quality evaluation value serves as the gain adjustment coefficient of the algorithm. When the frequency control quality evaluation value is high, it indicates that the current data quality is good, and the gain can be appropriately increased to improve the adjustment speed. When the evaluation value is low, it indicates that the data quality is poor, and the gain should be appropriately reduced to avoid adjustment oscillations. The range of the initial frequency control judgment value is 0-1. The larger the value, the greater the frequency adjustment amplitude required. For example, when the frequency control quality evaluation value is 0...At 92, the algorithm calculates an initial frequency control judgment value of 0.78, indicating that the current switching frequency needs to be adjusted upwards by a maximum adjustment margin of 78%.

[0026] Complete data sequences corresponding to the most recent 7 frequency adjustment operations were extracted. This number was determined during the commissioning phase based on the transformer's thermal inertia time constant, typically ranging from 5 to 10 times. Too few sequences would lead to incomplete calculations of residual effects, while too many would increase unnecessary computational load. Each data sequence covers 15 seconds before and 60 seconds after the adjustment. The first 15 seconds are used to calculate the average baseline state before adjustment, eliminating the influence of random fluctuations. The average baseline value is calculated using the arithmetic mean method, i.e., the average of all data points within the first 15 seconds. The last 60 seconds cover the complete decay period of the residual effects of a single adjustment. This duration was determined through numerous step response experiments. We conducted these experiments on this type of transformer. A frequency step signal was applied, and the changes in winding temperature and switching losses were monitored. It was found that after 60 seconds, the changes in temperature and losses decayed to less than 5% of their initial values. Therefore, 60 seconds was determined to be a complete decay period. Different power ratings of transformers will show slight differences; higher power transformers have stronger thermal inertia, and the decay period of residual effects will be longer. Time-series differential processing was performed on each complete data sequence, calculating the difference between the parameter value at each moment and the average reference value within the 15 seconds before adjustment. This yielded a sequence of parameter changes before and after the adjustment operation. Each sequence of changes was assigned a time weight that decays exponentially with time. The time weight calculation formula is w=e (-t / τ) Where w is the time weight of historical adjustment operations, t is the time interval between the adjustment operation and the current time, and τ is the time constant, generally set to τ = 30 seconds. This formula indicates that the further back in time the adjustment operation is from the current time, the smaller its residual effect and the smaller its time weight. For example, the time weight of an adjustment operation 30 seconds ago is e. (-1) ≈0.368, the weight of the adjustment operation time 60 seconds ago is e. (-2) ≈0.135, the weight of the adjustment operation time 90 seconds ago is e. (-3) ≈0.05, which is negligible; align all weighted variation difference sequences along the time axis and perform four-dimensional vector superposition to construct a continuous time superposition residual influence field. The four dimensions correspond to the four parameters of input voltage, output current, winding temperature and switching loss, respectively. The weighted variation difference of each dimension is used as a component of the vector. Each time point corresponds to a four-dimensional vector. The magnitude of the vector represents the intensity of the residual influence at that moment, and the direction of the vector represents the trend of the residual influence.

[0027] The constructed time-stacked residual influence field is divided into grids according to preset rules, based on a 1-second time interval and a four-dimensional parameter space. Each grid is an independent field strength calculation unit. The parameter space division is determined according to the value range of each dimension parameter. For example, the input voltage dimension is divided into four ranges: 0-100V, 100-200V, 200-300V, and 300-400V; the output current dimension is divided into four ranges: 0-5A, 5-10A, 10-15A, and 15-20A; and the winding temperature dimension is divided into... The temperature is divided into four ranges: -40℃ to 0℃, 0℃ to 50℃, 50℃ to 100℃, and 100℃ to 150℃. Switching losses are further divided into four ranges: 0-250W, 250W-500W, 500W-750W, and 750W-1000W. This gridded division allows for more precise calculation of residual effects over different time periods and parameter ranges, rather than a simple overall average. The average of all four-dimensional vector magnitudes within each field strength calculation unit is used as the local field strength value for that unit. The formula for calculating the vector magnitude is... V1, V2, V3, and V4 are the weighted variation differences in the input voltage dimension, output current dimension, winding temperature dimension, and switching loss dimension, respectively. The weighted sum of all local field strength values ​​is used to obtain the comprehensive field strength value at the current moment. The weighting coefficient is determined according to the time distance between the grid and the current moment. The closer the time, the larger the weighting coefficient. For example, the weighting coefficient of the grid at the current moment is 1.0, the weighting coefficient of the grid 1 second ago is 0.9, the weighting coefficient of the grid 2 seconds ago is 0.8, and so on. The comprehensive field strength value ranges from 0 to 1. The larger the value, the stronger the residual influence of historical regulation.

[0028] Based on the calculated comprehensive field strength value, multi-order correction coefficients are generated, and the initial frequency control judgment value is subjected to layer-by-layer reverse correction. The correction process adopts the order from the inner layer to the outer layer, first performing dynamic correction at the time window level, then performing difference correction at the dimension level, and finally performing global amplitude correction. This order can ensure that the rapid changes of residual effects are captured first, then the differences in different dimensions are compensated, and finally the overall correction amplitude is controlled, avoiding over-correction or under-correction. If global correction is performed first, local differences will be masked, and the correction accuracy will be reduced.

[0029] Existing technologies at most employ a single global correction coefficient, which cannot account for the differences in residual effects across different dimensions and time periods. First, a first-order global correction coefficient is generated based on the overall field strength value. This first-order global correction coefficient is inversely proportional to the overall field strength value, calculated using the formula K1 = 1 - 0.5 × E, where K1 is the first-order global correction coefficient and E is the overall field strength value, used to control the overall correction amplitude. This coefficient was determined experimentally; when the overall field strength value is 1.0, the correction coefficient is 0.5, meaning the maximum correction amplitude is 50%, effectively avoiding overcorrection. For example, when the overall field strength value is 0.6, the first-order global correction coefficient is 1 - 0.5 × 0.6 = 0.7. Within the constraints of the first-order global correction coefficient, corresponding second-order dimensional correction coefficients are generated based on the average field strength values ​​of each dimension in the four-dimensional time-stacked residual influence field. The formula for calculating the second-order dimensional correction coefficient is K. 2i =1-0.3×E i , where K 2i Let Ei be the second-order correction coefficient for the i-th dimension, and Ei be the average field strength value for the i-th dimension. This coefficient is used to compensate for the differences in residual effects across different dimensions. The coefficient is determined based on the varying sensitivities of different dimension parameters to frequency regulation. Winding temperature and switching losses are more sensitive, hence their larger coefficients; input voltage and output current are less sensitive, hence their smaller coefficients. For example, if the average field strength value for the input voltage dimension is 0.4, the correction coefficient is 1 - 0.3 × 0.4 = 0.88; and the average field strength value for the output current dimension is 0.5. The correction coefficient is 1 - 0.3 × 0.5 = 0.85; the average electric field strength value for the winding temperature dimension is 0.7, and the correction coefficient is 1 - 0.3 × 0.7 = 0.79; the average electric field strength value for the switching loss dimension is 0.6, and the correction coefficient is 1 - 0.3 × 0.6 = 0.82; within the constraint range of the correction coefficient for each second-order dimension, the corresponding third-order time window correction coefficient is generated based on the rate of change of electric field strength in the three most recent consecutive time windows of that dimension. The length of each time window is 1 second, and the formula for calculating the rate of change of electric field strength is ΔE. i =(E i_current -E i_previous ) / E i_previous , where ΔE i Let E be the rate of change of the field strength in the i-th dimension. i_current E represents the average field strength value of the current time window in the i-th dimension. i_previous Let K be the average field strength value of the previous time window in the i-th dimension. The formula for calculating the third-order time window correction coefficient is K. 3i =1-0.2×|ΔE i |, where K 3iThe third-order correction coefficient for the i-th dimension is used to capture the dynamic changes of residual effects. A larger rate of change indicates a faster change in the residual influence of that dimension, and a smaller correction coefficient is required. For example, if the field strength values ​​for the most recent three time windows of a certain dimension are 0.5, 0.6, and 0.7, with an average rate of change of 0.2, then the third-order correction coefficient is 1 - 0.2 × 0.2 = 0.96. The initial frequency control judgment value is then multiplied sequentially by the third-order time window correction coefficient, the second-order dimension correction coefficient, and the first-order global correction coefficient to complete the layer-by-layer reverse correction, yielding the final frequency control execution value. For example, if the initial frequency control judgment value is 0.78, the average value of the third-order time window correction coefficient is 0. 95. The average value of the second-order dimension correction coefficient is (0.88+0.85+0.79+0.82) / 4=0.835, and the first-order global correction coefficient is 0.7. Therefore, the final frequency control execution value is 0.78×0.95×0.835×0.7≈0.43. It is worth noting that in order to prevent the adjustment from failing due to the correction coefficient being too small, a lower limit for the correction coefficient is set. The first-order global correction coefficient is not lower than 0.5, the second-order dimension correction coefficient is not lower than 0.6, and the third-order time window correction coefficient is not lower than 0.7. These lower limits are determined through experiments. When the correction coefficient is lower than the lower limit, the adjustment effect will be significantly worse, and even reverse adjustment may occur.

[0030] The final frequency control execution value is compared with a preset threshold of 0.5. This threshold was determined through extensive experimentation. When the execution value is greater than 0.5, the benefit of adjustment outweighs the loss of adjustment itself, making it worthwhile to execute. When the execution value is less than 0.5, the benefit of adjustment outweighs the loss, making it not worthwhile to execute. If the final frequency control execution value is greater than the preset threshold, it indicates that the required frequency adjustment range is large, and the adjustment amount is reasonable after correction. In this case, the current frequency adjustment strategy is executed, and a PWM control signal is sent to the IGBT drive circuit to adjust the switching frequency to the target value. For example, if the current switching frequency is 100kHz, the maximum adjustment range is 50kHz, and the final frequency control execution value is 0.52, then the adjusted frequency is 100 + 50 × 0.52 = 126kHz. The frequency adjustment adopts a step-by-step adjustment method, with a maximum step size of 5kHz for each adjustment to avoid voltage surges caused by frequency abrupt changes. If the final frequency control execution value is less than the preset threshold, it indicates that the required frequency adjustment range is small, or the previous adjustment was too small. If the residual effects of historical regulation are strong and the regulation effect is poor, the regulation parameters are adjusted, including increasing the regulation dead zone from 1kHz to 2kHz and extending the regulation period by 20%. The data acquisition and analysis process is restarted, and the next regulation period is waited for a judgment. When restarting the process, the previous sample set and historical regulation data are not cleared, but calculations are continued on the basis of the original data to ensure data continuity. The overall software adopts a modular design. The various functional modules interact with each other through a shared memory buffer. The data acquisition module undertakes the core task of front-end signal acquisition and preprocessing. The period generation module dynamically adjusts the control rhythm according to the running status. The sample construction module integrates the scattered raw data into a sample set that can be analyzed. The quality assessment module quantifies and scores the data quality. The initial judgment module generates preliminary regulation instructions based on the data quality and is responsible for quantifying the superposition effect of historical regulation. The correction execution module performs multi-layer correction on the preliminary regulation instructions. Finally, the execution judgment module decides whether to execute the regulation operation.

[0031] During the construction of the time-superimposed residual influence field, the changing trend of the comprehensive field strength value is monitored in real time. When the change in the comprehensive field strength value exceeds the preset abnormal threshold of 0.3 within three consecutive adjustment cycles, and normal frequency adjustment cannot restore it to the normal range of 0-0.8, a potential fault is determined to exist inside the transformer. This abnormal threshold is determined by statistical analysis of fault data from multiple transformers of the same model. When the change in the comprehensive field strength value exceeds 0.3, the probability of transformer failure exceeds 80%. At this time, the four-dimensional residual influence field distribution characteristics within the abnormal time period are extracted and compared with the preset fault feature library. The fault feature library is established by analyzing the operating data of a large number of faulty transformers and includes features of various fault types such as winding aging, decreased magnetic core hysteresis characteristics, and performance degradation of switching devices. Each fault corresponds to a unique field strength distribution pattern. For example, winding aging will lead to... The electric field strength in the winding temperature dimension increases continuously and changes slowly. A decrease in the magnetic core hysteresis characteristic leads to simultaneous fluctuations in the electric field strength in the input voltage and output current dimensions. Performance degradation of the switching devices causes a sharp increase in the electric field strength in the switching loss dimension. Feature matching uses a cosine similarity algorithm to calculate the similarity between the current electric field strength distribution feature and each feature in the fault feature library. A similarity greater than 0.8 indicates a corresponding fault. For example, if the electric field strength in the winding temperature dimension increases abnormally and its trend matches the feature of decreased magnetic core hysteresis characteristic (similarity 0.85), it is determined to be a magnetic core aging fault. If the electric field strength in the switching loss dimension increases abnormally and its trend matches the feature of increased switching device on-resistance (similarity 0.9), it is determined to be a switching device performance degradation fault. If the electric field strength in the output current dimension fluctuates abnormally and its trend matches the feature of inter-turn short circuit in the winding (similarity 0.88), it is determined to be a winding fault.

[0032] After identifying the fault type, the system issues a fault warning signal through audible and visual alarms and cloud push notifications, and provides corresponding maintenance suggestions. For example, it suggests replacing the magnetic core within 14 days if the magnetic core is aging, and replacing the switching transistor within 7 days if the switching device performance is degraded. The system records the time and characteristics of each fault warning and continuously updates the fault feature database to improve the accuracy of fault identification. This function can detect potential transformer faults 7-14 days in advance, providing sufficient time for equipment maintenance and reducing the risk of sudden equipment failures.

[0033] Example 2: Based on Example 1, such as Figure 2 As shown, a high-frequency transformer frequency control system is disclosed, the control system comprising: The data acquisition module collects multi-dimensional operating status data during a single operation of the high-frequency transformer. This multi-dimensional operating status data includes input voltage, output current, winding temperature, and switching losses. The cycle generation module generates differentiated frequency adjustment cycles that adapt to different parameter change characteristics based on multi-dimensional operating status data. The sample construction module performs batch data collection according to the differentiated frequency adjustment cycle and integrates them to obtain multiple batches of state sample sets; The quality assessment module performs a dual analysis of the structure and content of multiple batches of state sample sets to calculate the frequency control quality assessment value. The initial judgment module generates initial frequency control judgment values ​​based on the frequency control quality assessment values. The residual influence field construction module constructs a four-dimensional time-stacked residual influence field corresponding to multiple historical frequency adjustment operations, and calculates the comprehensive field strength value of the time-stacked residual influence field at the current moment. This module includes: a sequence extraction unit, which extracts a preset number of complete data sequences corresponding to historical frequency adjustment operations, each sequence covering multi-dimensional operational status data for a preset duration before and after the adjustment; a difference processing unit, which performs time-series difference processing on each complete data sequence to obtain a sequence of changes in parameters of each dimension before and after the adjustment operation; a time weighting unit, which assigns a time weight that decays exponentially with time to each change difference sequence, with the time weight decreasing the further away the adjustment operation is from the current moment; a vector overlay unit, which aligns all weighted change difference sequences along the time axis and performs four-dimensional vector overlay to construct a continuous time-stacked residual influence field; a grid partitioning unit, which divides the time-stacked residual influence field into several independent field strength calculation units according to preset rules; and a field strength calculation unit, which calculates the local field strength value of each field strength calculation unit and performs a weighted summation of all local field strength values ​​to obtain the comprehensive field strength value at the current moment. The correction execution module is used to generate multi-order correction coefficients based on the comprehensive field strength value, and to perform layer-by-layer reverse correction on the initial frequency control judgment value to obtain the frequency control execution value. The correction execution module includes: a first-order coefficient generation unit, which generates a first-order global correction coefficient based on the comprehensive field strength value to control the overall correction amplitude; a second-order coefficient generation unit, within the constraint of the first-order global correction coefficient, generates corresponding second-order dimension correction coefficients to compensate for the differences in residual effects in different dimensions based on the average field strength value of each dimension in the four-dimensional time-superimposed residual influence field; a third-order coefficient generation unit, within the constraint of each second-order dimension correction coefficient, generates corresponding third-order time window correction coefficients to capture the dynamic changes in residual effects based on the field strength change rate of that dimension within a continuous preset time window; and a layer-by-layer correction unit, which sequentially multiplies the initial frequency control judgment value by the third-order time window correction coefficient, the second-order dimension correction coefficient, and the first-order global correction coefficient to perform layer-by-layer reverse correction and obtain the frequency control execution value. The execution judgment module compares the frequency control execution value with a preset threshold. If the value is greater than the preset threshold, the current frequency adjustment strategy is executed; otherwise, the adjustment parameters are adjusted and the data acquisition and analysis process is restarted.

[0034] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A high-frequency transformer frequency control method, characterized in that, The control method involves the following steps: Collect multi-dimensional operating status data during a single operation of a high-frequency transformer. The multi-dimensional operating status data includes input voltage, output current, winding temperature, and switching losses. Generate differentiated frequency adjustment cycles that adapt to different parameter changes based on multi-dimensional operational status data; Data is collected in batches according to a differentiated frequency adjustment cycle, and then integrated to obtain multiple batches of state sample sets; A dual analysis of structure and content is performed on multiple batches of state sample sets to calculate the frequency control quality assessment value, and an initial frequency control judgment value is generated based on the frequency control quality assessment value. Construct a four-dimensional time-stacked residual influence field corresponding to multiple historical frequency adjustment operations, and calculate the comprehensive field strength value of the time-stacked residual influence field at the current moment; Multi-order correction coefficients are generated based on the comprehensive field strength value, and the initial frequency control judgment value is corrected layer by layer in reverse to obtain the frequency control execution value; The frequency control execution value is compared with the preset threshold. If it is greater than the preset threshold, the current frequency adjustment strategy is executed; otherwise, the adjustment parameters are adjusted and the data acquisition and analysis process is restarted.

2. The high-frequency transformer frequency control method according to claim 1, characterized in that, The steps for collecting multi-dimensional operational status data during a single operation of a high-frequency transformer are as follows: Locate all accessible data acquisition nodes in the high-frequency transformer control system and acquire raw operating data through standardized interfaces; The raw operating data is subjected to feature classification and time synchronization processing to identify input voltage data, output current data, winding temperature data and switching loss data, forming a synchronized multi-dimensional operating status dataset.

3. The high-frequency transformer frequency control method according to claim 1, characterized in that, The steps for generating a differentiated frequency adjustment period adapted to different parameter variation characteristics based on multi-dimensional operating status data are as follows: The corresponding basic adjustment cycle is determined based on the fluctuation characteristics of the input voltage, the load variation characteristics of the output current, the protection response characteristics of the winding temperature, and the frequency correlation characteristics of the switching loss. Obtain the current comprehensive field strength value, and dynamically scale all basic adjustment cycles proportionally according to the magnitude of the comprehensive field strength value; The scaled-up basic adjustment cycles are integrated to generate the final differentiated frequency adjustment cycle.

4. The high-frequency transformer frequency control method according to claim 1, characterized in that, The steps for performing batch data collection according to the differentiated frequency adjustment cycle and integrating it to obtain multiple batches of state sample sets are as follows: The data acquisition tasks are divided into multiple independent tasks based on the differentiated frequency adjustment cycle, and corresponding acquisition parameters are configured for each task. Based on the collected parameters, locate the original data source corresponding to the multi-dimensional operating status data, and determine the collection sequence for each data source; Data collection and preliminary verification are performed according to the collection sequence. The four types of data within the same time window are correlated and integrated to form multiple batches of state sample sets.

5. The high-frequency transformer frequency control method according to claim 1, characterized in that, The step of performing a dual analysis of the structure and content of multiple batches of state sample sets to calculate the frequency control quality assessment value includes: The information continuity assessment value is calculated based on the data structure type of multiple batches of state sample sets. Quantitative evaluation of the content of multiple batches of state sample sets was performed to obtain content evaluation results for each dimension; The initial weight coefficients of each dimension are determined by the analytic hierarchy process, and the initial weights are nonlinearly corrected by combining the comprehensive field strength value at the current moment. The content information evaluation value is calculated based on the corrected weighting coefficients; By integrating the information continuity assessment value and the content information assessment value, the final frequency control quality assessment value is obtained.

6. The high-frequency transformer frequency control method according to claim 1, characterized in that, The steps for constructing the four-dimensional time-stacked residual influence field corresponding to multiple historical frequency adjustment operations, and calculating the comprehensive field strength value of this influence field at the current moment are as follows: Extract a preset number of complete data sequences corresponding to historical frequency adjustment operations, with each sequence covering multi-dimensional operational status data for a preset duration before and after the adjustment execution; Perform temporal differencing on each complete data sequence to obtain a sequence of differences in the parameters of each dimension before and after the adjustment operation. Each change difference sequence is assigned a time weight that decays exponentially with time, where the time weight decreases the further away the adjustment operation is from the current time. Align all weighted change difference sequences along the time axis and perform four-dimensional vector superposition to construct a continuous time-stacked residual influence field. The residual influence field of time superposition is divided into grids according to preset rules to obtain several independent field strength calculation units; Calculate the local field strength value of each field strength calculation unit, and then sum the weighted values ​​of all local field strength values ​​to obtain the comprehensive field strength value at the current moment.

7. The high-frequency transformer frequency control method according to claim 6, characterized in that, The steps for generating multi-order correction coefficients based on the comprehensive field strength value and performing layer-by-layer reverse correction are as follows: The overall correction amplitude is controlled by generating a first-order global correction coefficient based on the comprehensive field strength value. Within the constraints of the first-order global correction coefficient, based on the average field strength values ​​of each dimension in the four-dimensional time superposition residual influence field, corresponding second-order dimension correction coefficients are generated to compensate for the differences in residual influence in different dimensions. Within the constraints of each second-order dimension correction coefficient, a corresponding third-order time window correction coefficient is generated based on the rate of change of field strength in that dimension within a continuous preset time window to capture the dynamic changes of residual influence. The initial frequency control judgment value is multiplied sequentially by the third-order time window correction coefficient, the second-order dimension correction coefficient, and the first-order global correction coefficient to perform a layer-by-layer reverse correction, thereby obtaining the frequency control execution value.

8. A high-frequency transformer frequency control system, characterized in that, The control system includes: The data acquisition module collects multi-dimensional operating status data during a single operation of the high-frequency transformer. This multi-dimensional operating status data includes input voltage, output current, winding temperature, and switching losses. The cycle generation module generates differentiated frequency adjustment cycles that adapt to different parameter change characteristics based on multi-dimensional operating status data. The sample construction module performs batch data collection according to the differentiated frequency adjustment cycle and integrates them to obtain multiple batches of state sample sets; The quality assessment module performs a dual analysis of the structure and content of multiple batches of state sample sets to calculate the frequency control quality assessment value. The initial judgment module generates initial frequency control judgment values ​​based on the frequency control quality assessment values. The residual influence field construction module constructs a four-dimensional time-stacked residual influence field corresponding to multiple batches of historical frequency adjustment operations, and calculates the comprehensive field strength value of the time-stacked residual influence field at the current moment. The correction execution module is used to generate multi-order correction coefficients based on the comprehensive field strength value, and to perform layer-by-layer reverse correction on the initial frequency control judgment value to obtain the frequency control execution value. The execution judgment module compares the frequency control execution value with a preset threshold. If the value is greater than the preset threshold, the current frequency adjustment strategy is executed; otherwise, the adjustment parameters are adjusted and the data acquisition and analysis process is restarted.

9. A high-frequency transformer frequency control system according to claim 8, characterized in that, The residual influence field construction module includes: The sequence extraction unit extracts a preset number of complete data sequences corresponding to historical frequency adjustment operations. Each sequence covers multi-dimensional operational status data for a preset duration before and after the adjustment execution. The difference processing unit performs time-series difference processing on each complete data sequence to obtain a sequence of differences in the parameters of each dimension before and after the adjustment operation. The time-weighted unit assigns a time weight that decays exponentially with time to each change difference sequence; the further away the adjustment operation is from the current time, the smaller the time weight. The vector overlay unit aligns all weighted change difference sequences along the time axis and performs four-dimensional vector overlay to construct a continuous time-overlay residual influence field. The time-stacked residual influence field is divided into several independent field strength calculation units according to a preset rule. The field strength calculation unit calculates the local field strength value of each field strength calculation unit, and performs a weighted summation of all local field strength values ​​to obtain the comprehensive field strength value at the current moment.

10. A high-frequency transformer frequency control system according to claim 8, characterized in that, The correction execution module includes: The first-order coefficient generation unit generates first-order global correction coefficients based on the comprehensive field strength value to control the overall correction amplitude. The second-order coefficient generation unit, within the constraint of the first-order global correction coefficient, generates corresponding second-order dimension correction coefficients to compensate for the differences in residual effects in different dimensions based on the average field strength values ​​of each dimension in the four-dimensional time superposition residual influence field. The third-order coefficient generation unit, within the constraint range of each second-order dimension correction coefficient, generates the corresponding third-order time window correction coefficient based on the field strength change rate of that dimension within a continuous preset time window to capture the dynamic changes of residual influence. The layer-by-layer correction unit is used to multiply the initial frequency control judgment value sequentially with the third-order time window correction coefficient, the second-order dimension correction coefficient, and the first-order global correction coefficient to perform layer-by-layer reverse correction and obtain the frequency control execution value.

Images