Power supply performance optimization method and system based on high-voltage pulse technology

Abstract

The application relates to the technical field of power supply performance optimization, and discloses a power supply performance optimization method and system based on high-voltage pulse technology. The method comprises the following steps: obtaining an original electrical parameter set of a power supply device to be optimized, extracting high-voltage pulse response characteristics and constructing a power supply dynamic response model; comparing the model with a standard power supply reference model, generating a performance difference characteristic set to screen key optimization indexes, monitoring real-time operation parameters of the device, and calculating a correlation degree matrix of the key optimization indexes and the real-time operation parameters; determining an optimization priority sequence based on the correlation degree matrix, positioning a core optimization node, and performing a high-voltage pulse test to generate a node test data set; extracting an optimization characteristic vector from the data set to formulate a high-voltage pulse adjustment strategy, analyzing an adjustment parameter combination in the strategy, and matching a corresponding standard adjustment specification. The method can accurately capture power supply dynamic performance characteristics, and improves optimization pertinence and standardization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power supply performance optimization, in particular to a power supply performance optimization method and system based on high-voltage pulse technology. BACKGROUND

[0002] In the process of electronic devices developing towards high precision and high power density, as the core of energy supply, the performance stability and reliability of power supply devices directly affect the running state of the entire electronic system. At present, the requirements for power supply output precision, dynamic response speed, anti-interference ability and other indicators in the fields of industrial control, new energy power generation, aerospace and other fields continue to improve, and the traditional power supply performance optimization method gradually exposes many limitations.

[0003] Traditional power supply performance optimization relies on static parameter detection and empirical adjustment, for example, through a multimeter, an oscilloscope and other tools to collect voltage, current, ripple and other static parameters, and then adjust the internal resistance, capacitance and other component parameters of the power supply according to the experience of technical personnel. This method can only reflect the performance state of the power supply under a specific steady-state working condition, and cannot capture the response characteristics of the power supply under dynamic scenarios such as load mutation and voltage fluctuation. However, in actual application, the power supply is often in a dynamic running environment, and insufficient dynamic response performance can easily lead to problems such as freezing and data loss of electronic devices.

[0004] Some optimization schemes try to introduce dynamic testing methods, but mostly use low-frequency signal excitation, which is difficult to simulate the working state of the power supply under high-voltage pulse scenarios. With the development of power electronics technology, high-voltage pulse power is increasingly widely used in electrostatic precipitation, pulse electroplating, medical devices and other fields. Such power supplies will generate high-voltage pulse signals during operation, and changes in signal amplitude, frequency, pulse width and other parameters will have a complex impact on the performance of the power supply. The existing optimization method lacks an effective analysis mechanism for high-voltage pulse response characteristics, and cannot accurately construct a dynamic response model of the power supply, making it difficult to accurately locate the core nodes affecting the performance of the power supply during optimization, resulting in low optimization efficiency and unstable results.

[0005] In the traditional optimization process, there is a lack of correlation analysis of key optimization indicators and real-time running parameters, and each parameter is often adjusted as an independent variable, ignoring the coupling relationship between parameters. For example, the stability of the power supply output voltage is not only related to the output filter capacitor, but also related to the pulse width, trigger frequency and other high-voltage pulse parameters. If only the capacitor parameter is adjusted, it is difficult to achieve optimal improvement of the overall performance. In addition, after the optimization strategy is formulated, there is a lack of effective matching with the standard adjustment specification, resulting in arbitrariness in the adjustment process, making it difficult to ensure the consistency and reliability of the optimization results, and failing to meet the strict requirements of industrial production on the performance stability of the power supply. SUMMARY

[0006] The application aims to provide a power supply performance optimization method and system based on high-voltage pulse technology to solve the problems in the background art.

[0007] To achieve the above-mentioned purpose, the application provides a power supply performance optimization method based on high-voltage pulse technology, which comprises the following steps:

[0008] Obtaining an original electrical parameter set of a power supply device to be optimized, extracting a high-voltage pulse response feature from the original electrical parameter set, and constructing a power supply dynamic response model based on the high-voltage pulse response feature;

[0009] Comparing the power supply dynamic response model with a standard power supply reference model to generate a performance difference feature set, screening a key optimization index from the performance difference feature set, monitoring real-time running parameters of the power supply device to be optimized, and calculating a correlation matrix of the key optimization index and the real-time running parameters;

[0010] Determining an optimization priority sequence based on the correlation matrix, positioning a core optimization node in the power supply device to be optimized according to the optimization priority sequence, performing a high-voltage pulse test on the core optimization node to generate a node test data set;

[0011] Extracting an optimization feature vector from the node test data set, formulating a high-voltage pulse adjustment strategy based on the optimization feature vector, analyzing an adjustment parameter combination in the high-voltage pulse adjustment strategy, and matching a standard adjustment specification corresponding to the adjustment parameter combination.

[0012] Preferably, the power supply dynamic response model is constructed based on the high-voltage pulse response feature, which comprises the following steps:

[0013] Identifying voltage fluctuation patterns and current transient patterns in the high-voltage pulse response feature;

[0014] Performing time-domain alignment processing on the voltage fluctuation patterns and the current transient patterns to generate a synchronous feature matrix;

[0015] Analyzing pulse response rules in the synchronous feature matrix to establish the power supply dynamic response model.

[0016] Preferably, the power supply dynamic response model is compared with the standard power supply reference model, which comprises the following steps:

[0017] Determining a feature dimension space of the power supply dynamic response model and the standard power supply reference model;

[0018] Calculating a feature distance distribution in the feature dimension space;

[0019] Performing cluster analysis on the feature distance distribution to generate the performance difference feature set.

[0020] Preferably, the calculation of the correlation matrix of the key optimization indicators and the real-time operation parameters comprises:

[0021] Extracting the environmental temperature gradient, input voltage fluctuation rate and load change curve in the real-time operation parameters;

[0022] Establishing a mapping relationship between the environmental temperature gradient, input voltage fluctuation rate and load change curve and the key optimization indicators;

[0023] Based on the mapping relationship, the correlation matrix is constructed.

[0024] Preferably, the determination of the optimization priority sequence based on the correlation matrix comprises:

[0025] Analyzing the feature weight distribution in the correlation matrix;

[0026] According to the feature weight distribution, the key optimization indicators are sorted to generate the optimization priority sequence.

[0027] Preferably, the high-voltage pulse test on the core optimization node comprises:

[0028] A multi-frequency high-voltage pulse signal is applied to the core optimization node, and a dynamic response waveform under the action of the multi-frequency high-voltage pulse signal is collected, and the node test data set is constructed based on the dynamic response waveform.

[0029] Preferably, the high-voltage pulse adjustment strategy based on the optimization feature vector comprises:

[0030] Analyzing the correlation characteristics of the optimization feature vector and the power circuit topology;

[0031] Determining an adjustment parameter space according to the correlation characteristics;

[0032] Searching for an optimal parameter combination in the adjustment parameter space to form the high-voltage pulse adjustment strategy;

[0033] The matching of the standard adjustment specification corresponding to the adjustment parameter combination comprises:

[0034] Analyzing the pulse frequency feature, pulse amplitude feature and duty cycle feature in the adjustment parameter combination;

[0035] Querying the adjustment specification in the standard parameter library that matches the pulse frequency feature, pulse amplitude feature and duty cycle feature;

[0036] Obtaining the standard adjustment specification.

[0037] Preferably, the method further comprises:

[0038] generate high-voltage pulse control instructions according to the standard adjustment specification, apply the high-voltage pulse control instructions to the power supply device to be optimized, collect a post-optimization performance data set, perform stability analysis on the post-optimization performance data set, and output a performance optimization evaluation report;

[0039] The application of the high-voltage pulse control instructions to the power supply device to be optimized comprises:

[0040] adjusting the working parameters of a pulse generator according to the high-voltage pulse control instructions;

[0041] real-time monitoring of the performance change curve of the power supply device to be optimized, and recording the post-optimization performance data set.

[0042] Preferably, the stability analysis of the post-optimization performance data set comprises:

[0043] extracting the voltage stability index, the current response speed index, and the power conversion efficiency index in the post-optimization performance data set;

[0044] establishing a stability evaluation matrix of the voltage stability index, the current response speed index, and the power conversion efficiency index;

[0045] generating the performance optimization evaluation report based on the stability evaluation matrix.

[0046] Preferably, the application further comprises a power supply performance optimization system based on high-voltage pulse technology, which comprises a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor, when executing the computer program, implements the steps of the power supply performance optimization method based on high-voltage pulse technology as described above.

[0047] Compared with the prior art, the application has the following beneficial effects:

[0048] By obtaining an original electrical parameter set of the power supply device to be optimized and extracting high-voltage pulse response characteristics, a power supply dynamic response model is constructed, breaking through the limitation of traditional optimization methods that only rely on static parameter detection. The traditional method cannot capture the performance change of the power supply in a dynamic scene, while the method focuses on high-voltage pulse response characteristics and can fully reflect the dynamic working state of the power supply under the action of a high-voltage pulse signal, making the constructed dynamic response model more consistent with the actual operation of the power supply and providing a more accurate basis for subsequent optimization work.

[0049] In the feature comparison and correlation analysis link, the power dynamic response model is compared with the standard power reference model to generate a performance difference feature set, and after screening the key optimization indicators, the correlation matrix of the key optimization indicators and the real-time operating parameters is further calculated. This process effectively solves the problem of insufficient parameter correlation analysis in traditional optimization. Through the correlation matrix, the coupling relationship between parameters can be clearly presented, avoiding the blindness of adjusting parameters as independent variables in the past. With the help of the correlation matrix to determine the optimization priority sequence, the importance and adjustment sequence of each optimization indicator can be clearly defined, making the optimization work more targeted and avoiding wasting too much effort on non-key indicators, significantly improving optimization efficiency.

[0050] By optimizing the priority sequence to locate the core optimization node and performing high-voltage pulse testing, a node test data set is generated, which can accurately lock the key position affecting the performance of the power supply. The traditional optimization method relies on experience to locate the core node, which is prone to positioning deviation, resulting in incorrect optimization direction. However, the method based on the scientific priority sequence locates the core optimization node, and then obtains the node test data through high-voltage pulse testing, which can accurately grasp the performance of the core node in the high-voltage pulse scenario, providing direct data support for subsequent optimization strategy, avoiding resource waste and performance risks caused by blind adjustment.

[0051] In terms of optimization strategy formulation and specification matching, the optimization feature vector is extracted from the node test data set and a high-voltage pulse adjustment strategy is formulated, while the adjustment parameter combination is analyzed and matched with the standard adjustment specification, effectively improving the scientificity of the optimization strategy and the standardization of the adjustment process. Traditional optimization strategies rely heavily on experience, and the adjustment process lacks standard basis, resulting in poor consistency of optimization results. However, the adjustment strategy based on the optimization feature vector can accurately adjust the specific performance defects of the core optimization node, and through matching with the standard adjustment specification, it ensures that the adjustment parameter combination meets industry standards and technical requirements, making the optimization process more controllable and the optimized power supply performance more stable and reliable.

[0052] This method is applied throughout the application of high-voltage pulse technology and is suitable for optimizing high-voltage pulse power supply performance. Compared with traditional low-frequency signal excitation optimization schemes, it is more suitable for the performance improvement needs of high-voltage pulse power supplies. In the context of the increasing application of high-voltage pulse power supplies, this method can meet the strict requirements of electrostatic precipitation, medical equipment and other fields for high-voltage pulse power supply performance, widening the application scope of power supply performance optimization methods and providing protection for the stable operation of electronic equipment in related fields, while promoting the development of power supply performance optimization technology towards more practical application scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 A working principle diagram of the power supply performance optimization method based on high-voltage pulse technology.

[0054] Figure 2 Flow chart for constructing power supply dynamic response model;

[0055] Figure 3 Flow chart for formulating and matching high-voltage pulse regulation strategy. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0057] Please refer to Figure 1 The present application provides a power supply performance optimization method based on high-voltage pulse technology, which comprises the following steps: obtaining an original electrical parameter set of a power supply device to be optimized, the set containing basic parameters such as voltage, current and frequency; extracting high-voltage pulse response features from the original electrical parameter set, the features being obtained by applying a high-voltage pulse signal and measuring the response waveform; constructing a power supply dynamic response model based on the high-voltage pulse response features, the model being used to simulate the dynamic behavior of the power supply under the action of the high-voltage pulse; comparing the power supply dynamic response model with a standard power supply reference model, the standard power supply reference model being established based on ideal power supply performance data; generating a performance difference feature set by comparison, the set highlighting the deviation of the power supply to be optimized from the standard performance; screening key optimization indicators from the performance difference feature set, the indicators including voltage stability, response time, etc.; simultaneously monitoring real-time running parameters of the power supply device to be optimized, the real-time running parameters including temperature, load change, etc.; calculating a correlation degree matrix of the key optimization indicators and the real-time running parameters, the matrix quantifying the correlation between the indicators and the parameters. Determining an optimization priority sequence based on the correlation degree matrix, the sequence identifying which indicators need to be processed first; locating a core optimization node in the power supply device to be optimized according to the optimization priority sequence, the core optimization node possibly being a specific component in the circuit such as a switch tube or a filter; performing a high-voltage pulse test on the core optimization node, the high-voltage pulse test involving applying a controllable pulse signal and recording the response; generating a node test data set, the data set containing voltage and current waveforms during the test. Extracting an optimization feature vector from the node test data set, the optimization feature vector representing the key performance attributes of the node; formulating a high-voltage pulse regulation strategy based on the optimization feature vector, the strategy defining how to adjust the pulse parameters to optimize the performance; analyzing a regulation parameter combination in the high-voltage pulse regulation strategy, the regulation parameter combination including pulse frequency, amplitude, etc.; matching a standard regulation specification corresponding to the regulation parameter combination, the standard regulation specification being derived from industry standards or experimental verification data.

[0058] In the process of constructing the power dynamic response model, identifying the voltage fluctuation pattern and current transient pattern in the high-voltage pulse response characteristics is a fundamental work. The voltage fluctuation pattern reflects the periodic or non-periodic variation law of the output voltage of the power under the action of the high-voltage pulse, and is usually manifested as amplitude modulation, frequency drift or damped oscillation. The current transient pattern captures the rapid change process of the current under pulse excitation, including peak current, decay response or step characteristics. The identification of these patterns depends on high-precision measuring instruments and signal processing techniques. For example, after capturing the real-time waveform by a digital oscilloscope, a pattern recognition algorithm such as clustering analysis or feature extraction method is used for classification and identification, so as to convert the original data into structured feature description. Time domain alignment processing of the voltage fluctuation pattern and the current transient pattern is a key step to ensure data consistency. Time domain alignment is achieved by synchronizing the sampling clock and signal trigger mechanism, eliminating the time offset caused by measurement device delay or signal transmission difference. In specific operation, interpolation algorithm or dynamic time warping technology can be used to match the time axis of voltage and current signals, so that the data points of the two are corresponding at the same time stamp. When generating a synchronous feature matrix, the aligned voltage and current data are organized into a two-dimensional array according to the time sequence. The rows of the matrix represent uniformly distributed time points, and the columns store the voltage instantaneous value and the current instantaneous value respectively. If necessary, derived dimensions such as differential features or integral features can be added to enhance information density. The matrix serves as the standardized data basis for subsequent analysis.

[0059] Analyzing the pulse response law in the synchronous feature matrix requires the application of various mathematical and engineering analysis methods. For example, time domain analysis can be used to observe the rise time, fall time and overshoot amplitude of the waveform, or frequency domain analysis such as fast Fourier transform can be used to identify the resonant frequency and harmonic components. Statistical methods can also be used to calculate the distribution parameters of characteristic values such as mean, variance or correlation coefficient. When establishing the power dynamic response model, appropriate modeling framework is selected based on the above analysis results. For example, a state space model can be used to describe the dynamic behavior of the system, where the state variables can include capacitor voltage and inductor current. Model parameters are fitted from the synchronous feature matrix by least squares method or system identification technology, and the final model can simulate the dynamic characteristics of the power under high-voltage pulse excitation, providing a prediction tool for performance optimization.

[0060] The entire implementation process relies on strict experimental control and data processing. The acquisition of high-voltage pulse response characteristics needs to be carried out in a shielded environment to avoid external interference. The measuring equipment such as high-voltage probes and current sensors needs to be calibrated to ensure accuracy. The data acquisition frequency should be higher than twice the highest frequency of the pulse to meet the Nyquist sampling theorem. In the time domain alignment process, the time synchronization accuracy needs to reach the microsecond level to avoid phase errors. The construction of the synchronization feature matrix is usually completed with the help of computer software, such as using the array processing library in MATLAB or Python to realize data reorganization and formatting. The analysis of pulse response rules combines automatic algorithms and manual verification, such as identifying pattern types through machine learning classifiers or adjusting parameter boundaries based on engineers' experience. The final output of the model construction is a mathematical representation or a software module that can receive high-voltage pulse parameters as input and output the predicted power response waveform. This model can be integrated into a larger optimization system for real-time testing or offline analysis. In the implementation process, attention should be paid to avoid overfitting or simplification, and the model complexity should be balanced with actual needs. For example, in scenarios where computing resources are limited, a simplified transfer function model can be used, while in scenarios where high precision is required, a nonlinear dynamic model can be used. All processing steps need to record parameter settings and intermediate results to trace and verify the reliability of the model.

[0061] Example 2: see Figure 2 In the feature comparison stage, the feature dimension space of the power dynamic response model and the standard power reference model needs to establish a unified feature coordinate system. The construction of the feature dimension space is based on the performance indicators common to both models, such as transient response time, steady-state error, overshoot amplitude, etc. Each parameter corresponds to an orthogonal dimension in the space, forming a multi-dimensional framework that can be used for mathematical comparison. When calculating the feature distance distribution in the feature dimension space, the Euclidean distance is used to measure the difference value of the two models in each dimension. By traversing all feature dimensions, a distance scalar sequence is obtained, which reflects the dispersion degree of the models in each performance indicator. The larger the distance value, the more significant the performance difference in that dimension. When performing cluster analysis on the feature distance distribution, the density clustering algorithm is applied to identify the clustering area of the distance value in the numerical space, and the feature dimensions with similar difference degrees are classified to form several difference level clusters. The representative feature vectors extracted from the cluster centers constitute the performance difference feature set, which systematically describes the performance gap between the power to be optimized and the standard reference model.

[0062] In the calculation of the correlation matrix stage, the environmental temperature gradient in the real-time operating parameters needs to monitor the temperature change rate around the power supply equipment over time, collect data at a fixed sampling frequency through a temperature sensor and calculate the temperature change amount per unit time; the input voltage fluctuation rate is measured by a high-precision voltage probe to measure the voltage fluctuation at the input end, and the ratio of the standard deviation to the mean value is taken as the fluctuation rate quantitative index; the load change curve is recorded by a current sensor to record the change trajectory of the output load current over time, forming a time-current relationship map. When establishing the mapping relationship between the environmental temperature gradient, the input voltage fluctuation rate and the load change curve and the key optimization index, a multiple regression analysis method is adopted, taking each operating parameter as the independent variable and the key optimization index as the dependent variable, and the least squares method is used to fit the mathematical relationship between the parameters and the index, and the influence coefficient of each operating parameter on the performance index is quantified. When constructing the correlation matrix based on the mapping relationship, the correlation coefficients obtained by regression analysis are arranged in a specific order to form a square matrix, the rows of the matrix correspond to the key optimization index, the columns correspond to the real-time operating parameters, the element values represent the correlation strength between the index and the parameter, and the positive and negative signs indicate the correlation direction. This matrix provides a data basis for the subsequent determination of optimization priorities.

[0063] The entire implementation process requires the cooperative work of precision measuring instruments and computing tools. The determination of the feature dimension space depends on the model parameters obtained in the early modeling stage, and the feature definitions and dimensions of the two models must be consistent to construct the space. When calculating the feature distance, attention should be paid to data standardization to eliminate the calculation deviation caused by the magnitude difference between different dimensions. The parameter setting of the clustering analysis algorithm needs to be adjusted according to the actual data distribution, such as the determination of the clustering radius which requires multiple iterations to obtain the best clustering effect, and the generation of the performance difference feature set which needs to be verified by engineering to ensure that the contained features can truly reflect the performance defects of the power supply. The collection of real-time operating parameters needs to be carried out in the actual working environment of the power supply, and the monitoring time should cover the typical working period of the power supply to obtain statistically significant parameter data; the establishment of the mapping relationship requires a sufficient number of sample points to ensure the reliability of the regression analysis, and the construction of the correlation matrix needs to follow the matrix operation rules and be symmetrically processed if necessary to meet the mathematical requirements.

[0064] Quality control measures during implementation include: regular calibration of measuring equipment to ensure data collection accuracy, environmental temperature monitoring to avoid local heat sources affecting the sensor, input voltage measurement considering power grid interference and taking filtering measures, load change record covering various working conditions from no load to full load. In the data analysis stage, multiple methods are used for cross-validation, such as using Pearson correlation coefficient and Spearman rank correlation coefficient to evaluate correlation strength, clustering results are evaluated by silhouette coefficient to ensure that the output performance difference feature set and correlation matrix have engineering practical value. All intermediate data and processing parameters need to be recorded in detail to form a complete process document for subsequent tracing and scheme optimization.

[0065] Taking the power system optimization of a certain type of medical imaging equipment as an example, the output ripple of the equipment abnormally increased during operation. In the feature comparison stage, when determining the feature dimension space of the power dynamic response model and the standard power reference model, five key feature dimensions are selected: transient response time, steady-state voltage accuracy, ripple coefficient, load regulation rate, and temperature drift coefficient. Each dimension uses the same quantization standard, the transient response time is in microseconds, the steady-state voltage accuracy is in percentage, the ripple coefficient is in millivolts, the load regulation rate is in percentage per ampere, and the temperature drift coefficient is in percentage per degree Celsius. When calculating the feature distance distribution in the feature dimension space, the Euclidean distance formula is used to calculate the difference value in each dimension. The measured transient response time difference between the medical power supply and the standard model is 15 μs, the steady-state voltage accuracy difference is 0.8%, the ripple coefficient difference is 12 mV, the load regulation rate difference is 0.5% / A, and the temperature drift coefficient difference is 0.03% / ℃. The feature distance distribution composed of these distance values shows obvious hierarchical characteristics, and the distance value of the ripple coefficient is significantly higher than that of other dimensions. When performing clustering analysis on the feature distance distribution, the k-means algorithm is applied to divide the five feature dimensions into two clustering clusters.

[0066] The first cluster contains two dimensions of ripple factor and load regulation, and the cluster center distance value is 8.6; the second cluster contains the remaining three dimensions, and the cluster center distance value is 2.1. The feature vector extracted from the center point of the first cluster constitutes the performance difference feature set, which clearly indicates that the ripple characteristics and load response characteristics are the main performance gap of the medical power supply from the standard model. In the correlation matrix calculation stage, it is monitored that the environmental temperature gradient presents a periodic change during the operation of the device, with a change amplitude of ±3°C / hour; the input voltage fluctuation rate is recorded as ±2.5%; the load change curve shows that there is a pulse load with a period of 200ms during the operation of the device, and the peak current reaches 150% of the rated value. When establishing the mapping relationship between these parameters and the key optimization indicators, it is found through multiple regression analysis that the correlation coefficient of the ripple factor with the environmental temperature gradient is 0.72, the correlation coefficient with the input voltage fluctuation rate is 0.85, and the correlation coefficient with the load change amplitude is 0.91.

[0067] When constructing the correlation matrix based on the mapping relationship, the ripple factor is taken as the row vector, and the three operating parameters are taken as the column vector, forming a 3x3 square matrix. The matrix element values are: temperature gradient correlation degree 0.72, voltage fluctuation rate correlation degree 0.85, and load change correlation degree 0.91. The matrix shows that the ripple factor has the strongest correlation with the load change characteristics, providing a clear direction indication for subsequent optimization. During the entire implementation process, the selection of feature dimensions refers to the industry standard IEC60601-1 of medical device power supply, ensuring the authority of the comparison benchmark. Standardization is adopted in distance calculation to eliminate the calculation deviation caused by different dimensions. The clustering analysis is verified through multiple iterations, and when the cluster number is 2, the silhouette coefficient reaches the maximum value of 0.68, indicating that the clustering result has good consistency. Real-time parameter monitoring uses a high-precision data acquisition system, temperature monitoring uses a PT100 sensor, voltage fluctuation measurement uses a 16-bit ADC chip, and load current is collected by a Hall effect sensor. The sampling frequency of all sensors is unified to 10kHz, ensuring the time synchronization of the data. In the data analysis stage, the cross-validation method is used, and the 120 groups of collected operating parameters are randomly divided into training set and test set, respectively, to construct the correlation matrix and compare the similarity of the results. The final correlation matrix is subjected to significance test, and the p value of all correlation coefficients is less than 0.01, indicating that the statistical result has significant meaning. All intermediate data and analysis results generated during the implementation process are recorded in the project document, including the original monitoring data, feature distance calculation process, clustering analysis parameter setting, and the construction details of the correlation matrix.

[0068] In the process of determining the optimization priority sequence based on the correlation matrix, the eigenvalue decomposition method is used to analyze the feature weight distribution in the correlation matrix. This method reveals the internal structure by solving the eigenvectors and eigenvalues of the matrix, and the size of the eigenvalue directly reflects the importance of the dimension represented by the corresponding eigenvector. When sorting the key optimization indicators according to the feature weight distribution, the eigenvalues are arranged in descending order and mapped back to the original indicator space, so that each key optimization indicator obtains a quantitative weight score, which represents the relative importance of the indicator in the overall performance optimization. When generating the optimization priority sequence, the key optimization indicators are arranged in order according to their weight scores from high to low to form an ordered list, and the higher the weight score, the greater the impact of the indicator on system performance, which needs to be processed first in the subsequent optimization process. This sequence provides clear guidance for the allocation of optimization resources.

[0069] When performing high-voltage pulse testing on the core optimization node, a programmable pulse generator is used to apply a multi-frequency high-voltage pulse signal to the core optimization node. This device can generate pulse waveforms with a frequency range from several hertz to several megahertz, and the pulse amplitude can be adjusted between several hundred volts and several thousand volts according to testing requirements. The signal application method needs to follow electrical safety specifications and use isolation measures to protect the testing equipment. When collecting dynamic response waveforms under the action of a multi-frequency high-voltage pulse signal, a high-bandwidth oscilloscope and a current probe are used to monitor the voltage and current changes at the node simultaneously, and the sampling rate needs to meet the Nyquist criterion to accurately capture the transient characteristics. Waveform data is transmitted in real time to the data processing system through a digital interface. Based on the dynamic response waveform, the original waveform data collected is preprocessed, including denoising, normalization, and time alignment, and then feature parameters such as rise time, fall time, peak voltage, and energy loss are extracted, and finally organized into a structured dataset for subsequent analysis.

[0070] Technical details in the implementation process need special attention. The calculation of the feature weight distribution uses the following formula:

[0071]

[0072] Where: wi represents the weight coefficient of the ith feature, where λi represents the ith eigenvalue of the correlation matrix, and n represents the total number of eigenvalues. This formula ensures that the sum of all weight coefficients is 1, facilitating the comparison of the relative importance of different features. In multi-band high-voltage pulse testing, the selection of frequency bands needs to cover the main frequency bands of the power supply, usually including low frequency bands (10 Hz-1 kHz), medium frequency bands (1 kHz-100 kHz), and high frequency bands (100 kHz-5 MHz). The number of test points in each frequency band should be determined according to the actual situation, generally not less than 10 frequency points to obtain sufficient frequency response characteristics. The collection of dynamic response waveforms needs to pay attention to trigger settings and storage depth to ensure that the complete transient process can be captured, especially for nanosecond-level rapid changes that require high sampling rate mode. The data processing process needs strict quality control, the original waveform data is first corrected for baseline to eliminate DC bias, then a digital filter is applied to suppress high-frequency noise, and the sliding window method is used to calculate local eigenvalues for feature extraction to avoid information loss. The construction of the node test data set adopts a hierarchical structure, the first layer stores the original waveform data, the second layer stores the extracted feature parameters, and the third layer stores metadata such as test conditions, timestamps, and device information. This structure facilitates subsequent data retrieval and analysis. The entire implementation process needs to record all parameter settings and operation steps, including pulse generator output parameters, measurement device configuration information, and data processing algorithm parameter selection, to ensure the repeatability of the experiment and the reliability of the results. The generation of the optimization priority sequence needs to consider actual engineering constraints, in addition to the weight score, it also needs to be evaluated comprehensively in combination with the optimizability and cost factors of the indicators, for example, some indicators with high weights may be difficult to optimize due to hardware limitations, at this time the priority order needs to be adjusted. Safety protection measures need to be taken in high-voltage pulse testing, including the use of isolation transformers, the setting of overvoltage and overcurrent protection circuits, and the implementation of remote operation mechanisms to prevent high-voltage pulses from causing damage to personnel and equipment. Environmental conditions such as temperature and humidity need to be monitored during testing, as these factors may affect the accuracy of the test results, if necessary, testing should be carried out in a constant temperature and humidity environment to obtain reliable data.

[0073] Taking the optimization of a server power module in a data center as an example, the power module has output voltage oscillation phenomenon when the load suddenly changes. In the process of determining the optimization priority sequence based on the correlation matrix, the principal component analysis method is used to analyze the characteristic weight distribution in the correlation matrix. From the correlation matrix containing six dimensions such as ripple coefficient, transient response time, efficiency index, etc., the main feature vector is extracted. The matrix eigenvalue calculation shows that the contribution rate of the first principal component is 68%, the weight coefficient of the transient response time in the corresponding feature vector is 0.83, the weight coefficient of the ripple coefficient is 0.76, and the weight coefficient of the efficiency index is 0.42. When sorting the key optimization indicators according to the characteristic weight distribution, the priority score of the transient response time is 0.39, the priority score of the ripple coefficient is 0.36, and the priority score of the efficiency index is 0.19 after normalization of the weight coefficients. The scores of the other three indicators are all less than 0.1. When generating the optimization priority sequence, the scores are arranged from high to low as follows: transient response time, ripple coefficient, efficiency index. This sequence shows that the dynamic response characteristics of the power supply should be improved first, then the output ripple problem should be handled, and finally the efficiency optimization should be considered. When performing high-voltage pulse testing on the core optimization nodes, the power MOSFET switch tube and the output filter capacitor are determined as the key nodes.

[0074] When applying multi-band high-voltage pulse signals to these nodes, a high-voltage pulse generator is used to generate test signals with frequencies from 50 kHz to 500 kHz, pulse widths set between 100 ns and 500 ns, and voltage amplitudes gradually increased from 50 V to 300 V. Each frequency point maintains a test duration of 10 ms, and a total of 20 different frequency bands of response data are collected. When collecting dynamic response waveforms, differential probes and current probes are used for synchronous measurement, the sampling rate is set to 2.5 GS / s, the storage depth reaches 250 M points, and the nanosecond-level transient details can be captured. At a test frequency of 500 kHz, obvious ringing phenomenon is observed, and at a frequency of 200 kHz, voltage overshoot is found in the switch node, and below 100 kHz, it shows relatively stable performance. When constructing the node test dataset based on the dynamic response waveforms, the key feature parameters are extracted after noise reduction processing of the original waveform data. For the switch tube node, 12 parameters such as conduction delay time, turn-off tail current, and voltage peak value are extracted; for the filter capacitor node, 8 parameters such as equivalent series resistance, resonance frequency, and dielectric loss angle are extracted. All parameters are stored according to the test frequency, forming a structured dataset containing 240 data points.

[0075] Special attention was paid to test safety during the implementation process. High-voltage pulses were applied in progressively increasing increments, and a thermal scan was performed immediately after each voltage level test to prevent overheating damage to components. The data acquisition system employed fiber optic isolation technology to avoid ground loop interference affecting measurement accuracy. The feature parameter extraction algorithm used a sliding window averaging method, effectively suppressing noise interference while maintaining feature accuracy. The optimization priority sequence generation considered actual maintenance conditions; although the efficiency index had a lower weight score, it was kept as the third priority due to the difficulty of optimizing it involving the replacement of magnetic components. The high-voltage pulse test covered the main frequency band of the power supply operation, with a particularly increased test density near the resonant frequency. Five test points were added near 350kHz to accurately locate the resonant characteristics. The dataset construction adopted a hierarchical storage structure, with raw waveform data saved in binary format, feature parameters stored in a structured database, and metadata information for all test conditions retained. The entire implementation process lasted approximately 6 hours, including 2 hours of warm-up and stabilization time, 3 hours of testing time, and 1 hour of data processing time. All operations strictly adhered to electrostatic discharge (ESD) protection specifications, with the test environment temperature controlled at 25±2℃ and humidity maintained at 45%±5%. The generated test data totaled approximately 35GB, including the original waveform files, processed feature parameter files, and complete test log records.

[0076] Example 4: See Figure 3 In formulating a high-voltage pulse regulation strategy based on optimized eigenvectors, analyzing the correlation between the optimized eigenvectors and the power circuit topology requires examining the physical meaning of each component in the eigenvector and its manifestation in the circuit. For example, one eigenvector component might correspond to the conduction loss of a switching device, while another component might reflect the saturation characteristics of a magnetic component. This correlation analysis requires a systematic mapping based on the circuit schematic and component parameters. When determining the regulation parameter space based on the correlation characteristics, adjustable pulse parameters such as frequency, amplitude, and rise time are used as dimensions to construct a multi-dimensional space. The value range of each dimension is determined by hardware limitations and safety standards, forming a bounded continuous parameter space for subsequent searches. When searching for the optimal parameter combination in the regulation parameter space, a multi-objective optimization algorithm is used to simultaneously consider the performance improvement and operational stability. Iterative calculations are used to find the parameter points that optimize the objective function. The final high-voltage pulse regulation strategy includes specific parameter setting schemes and application condition descriptions.

[0077] In the stage of matching the standard regulation specification corresponding to the combination of adjustment parameters, the pulse frequency characteristic in the combination of adjustment parameters needs to extract the distribution information of the fundamental frequency and its harmonic components from the optimization results, the pulse amplitude characteristic needs to distinguish the specification requirements of peak voltage and effective value voltage, and the duty cycle characteristic needs to determine the range of the ratio of on-time to period and its temperature-dependent characteristics. When querying the standard parameter library that matches the pulse frequency characteristic, the pulse amplitude characteristic and the duty cycle characteristic, it is necessary to establish the mapping relationship between the characteristic value and the standard parameter, for example, comparing the measured frequency value with the standard frequency tolerance table to confirm whether it meets the industry general specification. After obtaining the standard regulation specification, it needs to be consistent with the parameter combination obtained by optimization to ensure that the optimization result meets the performance requirements and conforms to the safety standards and industry specifications.

[0078] Taking the optimization of a certain type of switching power supply as an example, it is found that the voltage regulation rate characteristic component is strongly associated with the inductance parameter of the main power loop, and the efficiency characteristic component is closely related to the driving timing of the switching device. The determination of this correlation characteristic provides a clear direction for subsequent parameter adjustment. The construction of the adjustment parameter space considers the actual working conditions of the power supply, the pulse frequency range is set to 50 kHz to 200 kHz, the pulse amplitude is limited to 300 V to 500 V, and the duty cycle adjustable range is 10% to 90%. This parameter space not only ensures the flexibility of adjustment but also ensures the safety of the device. In the optimal parameter search process, the simulated annealing algorithm is used, and after multiple iterations, a parameter combination that balances various performance indicators is obtained. The specific values are recorded in the following table, see Table 1.

[0079] Table 1: High-voltage pulse adjustment parameter optimization results table

[0080] Parameter Type Pre-optimization Value Post-optimization Value Allowed Range Standard Specification Reference Pulse Frequency 100 kHz 152 kHz 50-200 kHz IEC 62040-3 Pulse Amplitude 400V 450V 300-500V UL 1012 Duty Cycle 50% 65% 10-90% IEEE 181 Rise Time 100 ns 75 ns 50-150 ns MIL-STD-704 Fall Time 120 ns 80 ns 50-150 ns MIL-STD-704

[0081] In matching the standard regulation specification, the 152 kHz pulse frequency obtained by optimization is compared with the frequency tolerance requirement in IEC62040-3 standard, confirming that it falls within the standard specified ±5% tolerance range, and the pulse amplitude of 450 V also meets the requirement of UL1012 standard for insulation strength. The verified duty cycle parameter 65% meets the limitation condition of IEEE181 standard for pulse waveform duty cycle, and the optimization results of rise time and fall time are within the allowed range specified by MIL-STD-704 standard, which indicates that the optimization parameter combination fully meets the relevant standard specifications.

[0082] The mutual influence between parameters is particularly noted during implementation, for example, the change of pulse frequency will affect the heating characteristics of the magnetic element, for this reason, the temperature rise limit standard is also consulted when querying the standard specification, to ensure that the optimized parameters will not cause the equipment to overheat. The duty cycle adjustment may affect the stress level of the switching device, therefore, the voltage stress limit value in the relevant safety standard is checked, and all optimized parameters are finally determined after multiple standard checks. The entire matching process is assisted by automated tools, and the standard parameter library is constructed as a queryable database system, which quickly locates the applicable specification clauses through feature matching algorithms, greatly improving the efficiency and accuracy of consistency checking. The final standard regulation specification not only contains parameter value requirements, but also clearly specifies the operation process and safety precautions when applying these parameters.

[0083] In example 5, generating high-voltage pulse control instructions according to the standard regulation specification requires converting the parameter requirements in the specification into executable digital commands for the equipment. This conversion process is achieved through an encoding algorithm, which packages parameters such as pulse frequency and amplitude into data frames according to the equipment communication protocol. The instruction contains necessary fields such as parameter value, execution timing, and verification information. When applying the high-voltage pulse control instruction to the power supply equipment to be optimized, the working parameters of the pulse generator are adjusted according to the instruction content. This process is completed through a digital interface such as GPIB or Ethernet connection, and parameter settings include specific items such as output waveform shape, voltage level, and trigger mode. Real-time monitoring of the performance change curve of the power supply equipment to be optimized requires deploying a multi-channel data acquisition system to synchronously record key parameters such as output voltage, current, and temperature over time. The sampling interval is set according to the dynamic characteristics of the power supply, usually using microsecond-level sampling to capture fast transient processes. When recording the optimized performance data set, the collected raw data is stored together with metadata such as timestamps and environmental conditions to form a structured data set for subsequent analysis. When performing stability analysis on the optimized performance data set, the voltage stability index in the data set is extracted by calculating the standard deviation and fluctuation range of the voltage value. The current response speed index is determined by analyzing the rise time and regulation time of the current step change, and the power conversion efficiency index is evaluated by comparing the ratio change of input and output power. Establishing a stability evaluation matrix for voltage stability index, current response speed index, and power conversion efficiency index requires organizing each index into a multi-dimensional array according to the time sequence. The dimensions of the matrix correspond to different observation periods and load conditions. Through matrix operations, the mutual relationship and development trend of the indices can be analyzed. Based on the stability evaluation matrix, a performance optimization evaluation report is generated by organizing the analysis results into a standardized format, including index comparison tables, trend curve graphs, and textual descriptions. The report focuses on the performance changes and stability performance before and after optimization.

[0084] Taking the optimization application of a certain industrial power supply as an example, the generation of high-voltage pulse control instructions uses the MODBUS communication protocol to convert the parameters specified in the standard regulation, such as 152 kHz frequency and 450 V amplitude, into register write commands. The instructions contain necessary information such as starting address, data length, and CRC check code. The pulse generator uses a certain brand of high-voltage amplifier, and its working parameters are set through a special software interface. During the setting process, special attention is paid to the smooth transition of parameter changes to avoid sudden parameter jumps that may impact the power supply equipment. Performance monitoring uses an eight-channel data acquisition card to record the voltage and current waveforms at the output end of the power supply at a sampling rate of one million per second, as well as auxiliary parameters such as radiator temperature and environmental humidity.

[0085] During the thirty-minute test process, the data acquisition system recorded more than ten million data points. After preliminary filtering, these data were stored as CSV format files with a file size of about 2 GB, containing complete time series information. Voltage stability analysis showed that the fluctuation range of the optimized output voltage was reduced from ±5% to ±2%, and the current response speed index indicated that the regulation time when the load stepped changed was reduced from 200 microseconds to 120 microseconds, and the average power conversion efficiency under different load conditions was improved by about 3 percentage points. The construction of the stability evaluation matrix used a time segmentation method, dividing the thirty-minute test data into six time intervals of five minutes each. The statistical characteristics of each index were calculated in each time interval to form a 6x3 evaluation matrix for comparative analysis.

[0086] The final performance optimization evaluation report uses a combination of text and graphics. The first part summarizes the test conditions and process, the second part shows the comparison data of the main performance indicators, the third part analyzes the stability performance, and the fourth part proposes further improvement suggestions. The report pays special attention to avoiding subjective evaluation and only objectively presents test data and analysis results. All conclusions are derived based on measured data. The entire implementation process strictly follows the quality control program, all measuring equipment is within the effective calibration period, and the data acquisition and processing algorithms have been verified to ensure the reliability and accuracy of the evaluation results.

[0087] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing the performance of a power supply based on high voltage pulse technology, characterized in that, The method comprises the following steps: acquiring an original electrical parameter set of a power supply device to be optimized, extracting a high-voltage pulse response feature from the original electrical parameter set, and constructing a power supply dynamic response model based on the high-voltage pulse response feature; performing feature comparison between the power supply dynamic response model and a standard power supply reference model to generate a performance difference feature set, screening a key optimization index from the performance difference feature set, monitoring real-time operation parameters of the power supply device to be optimized, and calculating a correlation matrix of the key optimization index and the real-time operation parameters; determining an optimization priority sequence based on the correlation matrix, positioning a core optimization node in the power supply device to be optimized according to the optimization priority sequence, performing a high-voltage pulse test on the core optimization node to generate a node test data set; extracting an optimization feature vector from the node test data set, formulating a high-voltage pulse adjustment strategy based on the optimization feature vector, analyzing adjustment parameter combinations in the high-voltage pulse adjustment strategy, and matching standard adjustment specifications corresponding to the adjustment parameter combinations. The method for constructing the power supply dynamic response model based on the high-voltage pulse response feature comprises the following steps: identifying voltage fluctuation modes and current transient modes in the high-voltage pulse response feature; performing time-domain alignment processing on the voltage fluctuation modes and the current transient modes to generate a synchronous feature matrix; analyzing pulse response rules in the synchronous feature matrix to establish the power supply dynamic response model. The method for performing feature comparison between the power supply dynamic response model and the standard power supply reference model comprises the following steps: determining a feature dimension space of the power supply dynamic response model and the standard power supply reference model; calculating a feature distance distribution in the feature dimension space; performing cluster analysis on the feature distance distribution to generate the performance difference feature set.

2. The method for power supply performance optimization based on high- voltage pulsed technology according to claim 1, characterized in that, The method for calculating the correlation matrix of the key optimization index and the real-time operation parameters comprises the following steps: extracting an environmental temperature gradient, an input voltage fluctuation rate, and a load change curve in the real-time operation parameters; establishing a mapping relationship between the environmental temperature gradient, the input voltage fluctuation rate, and the load change curve and the key optimization index; constructing the correlation matrix based on the mapping relationship.

3. The method for power supply performance optimization based on high- voltage pulsed technology according to claim 1, characterized in that, The method for determining the optimization priority sequence based on the correlation matrix comprises the following steps: analyzing a feature weight distribution in the correlation matrix; sorting the key optimization index according to the feature weight distribution to generate the optimization priority sequence.

4. The method for power supply performance optimization based on high- voltage pulsed technology according to claim 1, characterized in that, The method for performing a high-voltage pulse test on the core optimization node comprises the following steps: applying a multi-frequency high-voltage pulse signal to the core optimization node, collecting a dynamic response waveform under the action of the multi-frequency high-voltage pulse signal, and constructing the node test data set based on the dynamic response waveform.

5. The method for power supply performance optimization based on high- voltage pulsed technology according to claim 1, characterized in that, The method for formulating a high-voltage pulse adjustment strategy based on the optimization feature vector comprises the following steps: analyzing an association characteristic of the optimization feature vector and a power supply circuit topology; determining an adjustment parameter space according to the association characteristic; searching for an optimal parameter combination in the adjustment parameter space to form the high-voltage pulse adjustment strategy; The method for matching the standard adjustment specifications corresponding to the adjustment parameter combinations comprises the following steps: analyzing pulse frequency characteristics, pulse amplitude characteristics and duty cycle characteristics in the adjustment parameter combination; querying an adjustment specification in a standard parameter library that matches the pulse frequency characteristics, pulse amplitude characteristics and duty cycle characteristics; obtaining the standard adjustment specification.

6. The method for power supply performance optimization based on high- voltage pulsed technology according to claim 1, characterized in that, Further comprising: generating a high-voltage pulse control instruction according to the standard adjustment specification, applying the high-voltage pulse control instruction to the power supply device to be optimized, collecting a post-optimization performance data set, performing stability analysis on the post-optimization performance data set, and outputting a performance optimization evaluation report; applying the high-voltage pulse control instruction to the power supply device to be optimized comprises: adjusting the working parameters of the pulse generator according to the high-voltage pulse control instruction; real-time monitoring the performance change curve of the power supply device to be optimized, and recording the post-optimization performance data set.

7. The method for power supply performance optimization based on high- voltage pulsed technology according to claim 6, characterized in that, the stability analysis on the post-optimization performance data set comprises: extracting voltage stability indicators, current response speed indicators and power conversion efficiency indicators in the post-optimization performance data set; establishing a stability evaluation matrix of the voltage stability indicators, current response speed indicators and power conversion efficiency indicators; generating the performance optimization evaluation report based on the stability evaluation matrix.

8. A system for optimizing the performance of a power supply based on high voltage pulse technology, comprising a memory, a processor and a computer program stored in the memory and running on the processor, characterized in that, The processor, when executing the computer program, implements the steps of the power supply performance optimization method based on high-voltage pulse technology in any one of claims 1 to 7.

Images