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

Abstract

The application relates to the technical field of power supply testing, and discloses a power supply testing method and system based on high-voltage pulse technology. The method comprises the following steps: acquiring voltage response data and current response data of a to-be-tested power supply under preset high-voltage pulse parameters, wherein the voltage response data comprises voltage values of multiple test time points, and the current response data comprises current values corresponding to the time points; arranging the two types of data in a test time sequence to form voltage and current response sequences; calculating voltage fluctuation coefficients and current fluctuation coefficients of each test time point based on the two sequences; and determining an initial performance evaluation index of the to-be-tested power supply according to the two types of fluctuation coefficients. The method can accurately simulate high-voltage pulse working conditions, realizes accurate time correspondence of voltage and current data, quantifies fluctuation degrees of each time point to construct a comprehensive evaluation index, can comprehensively reflect dynamic response characteristics of the power supply, and is suitable for power supply performance detection in a high-voltage pulse scene.

Description

TECHNICAL FIELD

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

[0002] During the operation of electronic equipment, as the core of energy supply, the performance stability of the power supply is directly related to the reliable operation of the entire electronic system. With the development of electronic technology towards high precision and high power density, the testing requirements for power supply performance are also continuously improving, especially in the fields of aerospace, industrial control, medical equipment and other fields with strict requirements for power supply reliability, it is necessary to accurately capture the performance of the power supply under complex working conditions to determine whether it meets the actual application requirements.

[0003] The current mainstream power supply testing method is mostly based on steady-state working condition design, that is, by applying a constant voltage or current load, the stability of the power supply output parameters is monitored. Although this method can realize the detection of the basic output capability of the power supply, it has obvious limitations when facing instantaneous high-voltage pulse scenes. In actual application, the power supply often encounters instantaneous high-voltage pulse impact, such as power grid fluctuation, pulse signal generated during equipment start-stop moment, etc., at this time the voltage output and current supply of the power supply will appear temporary fluctuation, and the steady-state testing method cannot simulate such dynamic working condition, it is difficult to obtain the real response data of the power supply under high-voltage pulse, leading to the evaluation of the performance of the power supply is not comprehensive enough.

[0004] Some existing technologies try to introduce dynamic testing means to simulate dynamic scenes by adjusting the load change rate, but there are deficiencies in the control of high-voltage pulse parameters. Either the key parameters such as amplitude, duration, pulse frequency of high-voltage pulse cannot be accurately set, leading to a large deviation between the test conditions and the pulse environment in actual application; or in the data acquisition process, the voltage data and current data cannot be accurately corresponded in time dimension, making it difficult to accurately associate the voltage and current change at the same time point during subsequent analysis. In addition, the existing dynamic testing method is relatively single in performance evaluation index construction, mostly only through the voltage fluctuation range or current fluctuation range to judge the performance of the power supply, ignoring the influence of the difference of fluctuation degree at different time points on the overall stability of the power supply, it is difficult to fully reflect the dynamic response characteristics of the power supply under high-voltage pulse, which may lead to misjudgment of the performance of the power supply, affecting the selection of electronic equipment and system reliability design. SUMMARY

[0005] The purpose of the present application is to provide a power supply testing method and system based on high-voltage pulse technology to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a power supply testing method based on high-voltage pulse technology, the method comprises:

[0007] obtaining voltage response data and current response data of a to-be-tested power supply under preset high-voltage pulse parameters, the voltage response data including voltage values at multiple test time points, and the current response data including current values corresponding to the test time points;

[0008] arranging the voltage response data and the current response data in a test time sequence to form a voltage response sequence and a current response sequence;

[0009] calculating a voltage fluctuation coefficient and a current fluctuation coefficient at each test time point based on the voltage response sequence and the current response sequence;

[0010] determining an initial performance evaluation index of the to-be-tested power supply according to the voltage fluctuation coefficient and the current fluctuation coefficient.

[0011] Preferably, the calculation of the voltage fluctuation coefficient at each test time point comprises:

[0012] selecting voltage values within a preset time range before and after a current test time point to form a local voltage sequence;

[0013] calculating a ratio of a standard deviation to an average value of the local voltage sequence as the voltage fluctuation coefficient at the current test time point;

[0014] comparing the voltage fluctuation coefficient with a preset voltage fluctuation threshold value, and marking a test time point exceeding the preset voltage fluctuation threshold value as an abnormal test time point.

[0015] Preferably, the determination of the initial performance evaluation index of the to-be-tested power supply comprises:

[0016] calculating a ratio of a number of all abnormal test time points to a total number of test time points as a first abnormality ratio;

[0017] calculating an average value of all voltage fluctuation coefficients as a first stability parameter;

[0018] weighting and summing the first abnormality ratio and the first stability parameter to obtain the initial performance evaluation index.

[0019] Preferably, the method further comprises:

[0020] obtaining reference voltage response data and reference current response data of the to-be-tested power supply in historical tests;

[0021] performing dynamic time warping matching on the voltage response sequence of the current test and the reference voltage response data to obtain a voltage matching degree;

[0022] performing dynamic time warping matching on the current response sequence of the current test and the reference current response data to obtain a current matching degree;

[0023] Based on the voltage matching degree and current matching degree, the initial performance evaluation index is corrected to obtain the corrected performance evaluation index.

[0024] Preferably, the correction of the initial performance evaluation index includes:

[0025] Calculate the geometric mean of the voltage matching degree and the current matching degree as the overall matching degree;

[0026] The overall matching degree is compared with a preset matching degree threshold to obtain a matching degree correction coefficient;

[0027] The initial performance evaluation index is linearly adjusted using the matching degree correction coefficient to obtain the corrected performance evaluation index.

[0028] Preferably, the method further includes:

[0029] Collect test data from multiple power supplies of the same model under the same test conditions to form a reference dataset;

[0030] Calculate the similarity between the corrected performance evaluation index of the power supply under test and the performance evaluation index of each power supply in the reference dataset;

[0031] The performance level of the power supply under test is determined based on the similarity.

[0032] Preferably, determining the performance level of the power supply under test includes:

[0033] Reference power sources with a similarity greater than a preset similarity threshold are marked as similar power sources;

[0034] The proportion of similar power supplies whose performance evaluation indicators are better than those of the power supply under test is counted and used as the performance advantage.

[0035] The performance level of the power supply under test is determined based on the preset range in which the performance advantage falls.

[0036] Preferably, the method further includes:

[0037] Based on the performance level, adjust the high-voltage pulse parameters for subsequent tests;

[0038] A secondary test was conducted using the adjusted high-voltage pulse parameters to obtain secondary test data.

[0039] The secondary test data was compared and analyzed with the primary test data to verify the stability of the test results.

[0040] Preferably, the stability of the verification test results includes:

[0041] Calculate the correlation coefficient between the voltage response sequences of the first test and the second test, and use it as the voltage stability coefficient;

[0042] Calculate the correlation coefficient between the current response sequences of the first test and the second test, and use it as the current stability coefficient;

[0043] The weighted sum of the voltage stability coefficient and the current stability coefficient is compared with a preset stability threshold to determine whether the test result is stable.

[0044] When the test results are stable, the modified performance evaluation index and performance level will be used as the final test results;

[0045] If the test results are unstable, readjust the high-voltage pulse parameters and perform a third test until stable test results are obtained.

[0046] Preferably, the present invention also includes a power supply testing system based on high-voltage pulse technology, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the power supply testing method based on high-voltage pulse technology as described above.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] By acquiring the voltage and current response data of the power supply under test under preset high-voltage pulse parameters, and ensuring that the voltage and current response data correspond to the same test time points, this method can accurately simulate the high-voltage pulse scenarios encountered by the power supply in actual applications. This makes the collected response data more closely resemble real-world operating conditions, providing a more reliable foundation for subsequent performance evaluation. Compared to traditional steady-state testing methods that cannot cover dynamic pulse conditions, this method fills the gap in power supply testing under high-voltage pulse scenarios, making the testing process more targeted and effectively avoiding performance evaluation deviations caused by a disconnect between test conditions and actual applications.

[0049] In the data processing stage, the voltage and current response data are arranged in sequence according to the test time, clearly presenting the dynamic process of voltage and current changes over time. This facilitates intuitive observation of the changing trends of power supply output parameters at different time points. This ordered processing in the time dimension solves the problem of poor time correspondence between voltage and current data in some existing dynamic testing methods, ensuring that subsequent analysis can accurately correlate voltage and current fluctuations at the same moment, providing data support for in-depth research into the dynamic response mechanism of the power supply under high-voltage pulses.

[0050] By calculating the voltage and current fluctuation coefficients at each test time point, the degree of fluctuation in power supply output parameters at different time points can be accurately quantified, overcoming the limitations of existing testing methods that rely solely on a single fluctuation range for performance judgment. The fluctuation coefficients at different time points reflect the stability differences of the power supply at different stages under the action of a high-voltage pulse, such as the severity of voltage and current fluctuations at the initial moment of the pulse, during the pulse's duration, and after the pulse ends, thus capturing the detailed characteristics of the power supply's dynamic response more precisely. This refined fluctuation quantification method comprehensively demonstrates the dynamic performance of the power supply under high-voltage pulses, avoiding performance misjudgments caused by a single evaluation indicator, and allowing for a more in-depth analysis of power supply performance.

[0051] Determining initial performance evaluation indicators based on voltage and current fluctuation coefficients allows for the construction of a comprehensive evaluation system for power supply performance from multiple dimensions, moving beyond the limitations of considering a single parameter. These initial performance evaluation indicators comprehensively reflect the power supply's voltage and current stability under high-voltage pulses, providing a more complete picture of its dynamic response capabilities. For electronic equipment selection, these evaluation indicators can more accurately determine whether a power supply is suitable for the high-voltage pulse environment in practical applications. For power supply design optimization, analyzing the fluctuation coefficients and initial performance evaluation indicators at different time points can identify weaknesses in the power supply's dynamic response, providing direction for subsequent design improvements. This ultimately enhances the overall performance and market competitiveness of power supply products, while also improving the reliability of electronic systems when encountering high-voltage pulse impacts and reducing the risk of system failures due to insufficient power supply performance. Attached Figure Description

[0052] Figure 1 This is a schematic diagram illustrating the working principle of the power supply testing method based on high-voltage pulse technology described in this invention.

[0053] Figure 2 A flowchart for calculating the voltage fluctuation coefficient and marking abnormal time points;

[0054] Figure 3 A flowchart for revising initial performance evaluation metrics by matching historical benchmark data. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Please see Figure 1This invention provides a power supply testing method based on high-voltage pulse technology. The method includes: applying preset high-voltage pulse parameters to the power supply under test using a high-voltage pulse generator, including pulse voltage amplitude, pulse width, and repetition frequency. A data acquisition device collects voltage response data and current response data of the power supply under test in real time. The voltage response data consists of voltage values ​​at multiple test time points, and the current response data consists of current values ​​corresponding to the test time points. After acquisition, the voltage response data and current response data are arranged in chronological order according to the test time to form a voltage response sequence and a current response sequence. Based on the voltage response sequence and current response sequence, the voltage fluctuation coefficient and current fluctuation coefficient at each test time point are calculated. The voltage fluctuation coefficient is obtained by the ratio of the standard deviation to the mean of the local voltage sequence, and the current fluctuation coefficient is calculated in a similar manner. Based on the voltage fluctuation coefficient and current fluctuation coefficient, the initial performance evaluation index of the power supply under test is determined by a weighted summation of statistical anomaly ratio and stability parameters.

[0057] Example 1: See Figure 2 When calculating the voltage fluctuation coefficient at each test time point, voltage values ​​within a preset time range before and after the current test time point are selected. This preset time range needs to be reasonably set according to the sampling rate of the test system and the characteristics of the power supply under test. For example, in a test system with a sampling rate of 1MHz, the preset time range can be set to a time window covering 100 sampling points. The process of constructing a local voltage sequence is a dynamic data extraction process. As the test time points move, this time window also slides on the voltage response sequence to ensure that each point is included in the calculation range. Calculating the ratio of the standard deviation to the mean of the local voltage sequence requires a numerical calculation method. The standard deviation reflects the dispersion of the voltage value within that time period, while the mean represents the voltage baseline level. The ratio of the two can effectively eliminate the influence of absolute numerical magnitude, making the fluctuation coefficient a dimensionless relative indicator. As the voltage fluctuation coefficient at the current test time point, this value can quantify the stability of the voltage at that moment. A higher fluctuation coefficient indicates that there is a significant change in the voltage near that time point. When comparing the voltage fluctuation coefficient with a preset voltage fluctuation threshold, this threshold needs to be preset according to the power supply's technical specifications and test requirements. Usually, a reasonable critical value is determined based on historical test data or industry standards. Test time points that exceed the preset voltage fluctuation threshold are marked as abnormal test time points. This marking process can be achieved by adding flag bits to the data sequence. The marking of these abnormal points provides an important basis for subsequent analysis and processing.

[0058] The calculation of the current fluctuation coefficient follows the same processing logic as the voltage fluctuation coefficient. A local current sequence is constructed by selecting current values ​​within the same preset time range before and after the current test time point. The selection of this time window needs to be consistent with the voltage analysis to ensure data consistency. The ratio of the standard deviation to the mean of the local current sequence is calculated as the current fluctuation coefficient. The processing method for current data is completely symmetrical with that for voltage data, ensuring the comparability of the analysis methods for the two parameters. The current fluctuation coefficient is compared with a preset current fluctuation threshold. The current fluctuation threshold may differ from the voltage threshold, depending on the different voltage and current characteristics required in the power supply system. Test time points exceeding the threshold are marked as abnormal test time points. Identifying these abnormal points helps to comprehensively understand the dynamic response characteristics of the power supply during operation.

[0059] Determining the preset time range requires consideration of multiple factors, including the power supply's response characteristics, the duration of the high-voltage pulse, and the accuracy requirements of the test. A time window that is too small may fail to capture meaningful fluctuation information, while a time window that is too large may smooth out important transient characteristics. The construction of the local voltage sequence is a continuous process; a new local sequence is generated at each test time point. These sequences partially overlap to ensure that no important fluctuation information is missed. The standard deviation is calculated using standard statistical methods, and the average is calculated using an arithmetic mean algorithm. These calculations require both accuracy and efficiency in numerical processing. The fluctuation coefficient calculation provides a quantitative indicator for subsequent analysis. The fluctuation coefficient at each time point constitutes a new time series, reflecting the changes in the power supply's stability throughout the test. The comparison with the threshold uses a simple numerical comparison algorithm; all time points exceeding the threshold are recorded in the outlier list. Outlier markers include not only the time point location but also the specific value and degree of exceeding the threshold. This detailed information helps in in-depth analysis of the power supply's abnormal performance. The entire processing process needs to ensure computational efficiency, especially in high-sampling-rate test scenarios where the data volume can be very large. Therefore, optimized algorithms are required to achieve real-time or near-real-time processing. The sliding window approach can be implemented using a circular buffer to avoid repetitive data copying operations. The standard deviation and mean can be calculated incrementally, utilizing the results from the previous window to simplify the calculation process for the current window and improve processing speed.

[0060] The processing flow for current data is completely parallel to that for voltage data. Both processes can use the same algorithm structure, differing only in the input data. This symmetrical design facilitates code reuse and system maintenance. During processing, it is crucial to ensure the time synchronization of voltage and current data; the voltage and current data at each test time point must strictly correspond to guarantee the accuracy of subsequent analysis. The marking information for abnormal test time points is stored in a dedicated data structure that records the location and type of all anomalies. This marking information plays a vital role in subsequent performance evaluation; for example, calculating the anomaly ratio requires counting these anomalies. The distribution pattern of anomalies can also provide valuable information; for instance, consecutive anomalies may indicate a persistent stability problem in the power supply, while isolated anomalies may simply be random disturbances.

[0061] Statistical analysis of local sequences is not limited to standard deviation and mean. In some implementations, other statistics, such as kurtosis or skewness, can be added to more comprehensively describe the distribution characteristics of voltage and current. The calculation method for fluctuation coefficients can also vary; for example, absolute deviation can be used instead of standard deviation, or moving range can be used as a fluctuation indicator. These variations can be adjusted according to specific test requirements and power supply characteristics. Flexibility in threshold setting is a key feature of this method. Preset voltage and current fluctuation thresholds can be dynamically adjusted according to different test scenarios. A more lenient threshold can be used in the initial testing phase, gradually tightening the threshold requirements as testing progresses. Thresholds can also be adaptively adjusted according to the power supply specifications; for example, a stricter threshold standard can be used for high-precision power supplies. Data processing accuracy must be guaranteed, especially when calculating standard deviation and mean. Sufficient floating-point arithmetic should be used to avoid the accumulation of rounding errors. For embedded system implementations, fixed-point arithmetic or optimized algorithms may need to be considered to balance accuracy and computational resources. The data storage format also needs careful design to ensure that the accuracy of the original data is preserved without consuming excessive storage space.

[0062] Quality control throughout the process can be achieved in various ways, such as incorporating data validity checks during processing to eliminate obviously unreasonable data points. Redundant calculations can also be employed to double-verify critical calculation steps, ensuring the reliability of the results. Post-processing of outliers goes beyond simply marking them; in some implementations, real-time response mechanisms can be triggered. For example, when multiple consecutive outliers are detected, test parameters can be automatically adjusted or the test can be paused to prevent equipment damage. This real-time feedback mechanism improves the intelligence and security of the testing system. The method's scalability is reflected in the ease with which new analytical metrics can be added, such as calculating the differential rate of change or integral effect based on the volatility coefficient. The modular design of the processing flow allows each calculation step to be independently optimized or replaced without affecting the overall architecture. This design facilitates continuous improvement and adaptability to different testing needs. All calculated volatility coefficients and outlier information are integrated into a comprehensive data structure, providing complete foundational data for subsequent performance evaluation. The time-series characteristics of the data are fully preserved, allowing for the traceability of detailed calculation processes and results at each time point. This data structure is designed to meet the needs of fast querying and batch processing, supporting various subsequent analytical operations.

[0063] Example 2: See Figure 3 In determining the initial performance evaluation indicators of the power supply under test, the ratio of the number of all abnormal test time points to the total number of test time points is calculated. This calculation requires accurate identification of each abnormal time point and elimination of possible misjudgments. The total number of test time points is determined by the product of the test duration and the sampling rate. The calculation of the first abnormality ratio reflects the proportion of time during which the power supply exhibits abnormal responses during the high-voltage pulse test. A higher ratio indicates greater concern for the power supply's stability. Calculating the average of all voltage fluctuation coefficients requires traversing every calculation point throughout the entire test cycle. The average is calculated using an arithmetic mean algorithm, retaining sufficient decimal places to ensure accuracy. The first stability parameter is reflected through this average value, which characterizes the overall fluctuation level of the power supply's voltage output. Lower values ​​typically correspond to better stability performance. When weighted summing the first abnormality ratio and the first stability parameter, the weighting coefficients need to consider the different importance of the two parameters. Usually, the weight of the first abnormality ratio is set higher because it directly reflects the frequency of abnormal situations. This weighting process for obtaining the initial performance evaluation indicators needs to ensure the uniformity of dimensions. Sometimes, it is necessary to normalize the two parameters before performing the weighted calculation.

[0064] When acquiring reference voltage and current response data from historical tests of the power supply under test (PST), this reference data typically comes from the same power supply's factory tests or early verification tests and is stored in a dedicated test database. The selection of reference data must ensure that the test conditions are as consistent as possible with the current test, including ambient temperature, load conditions, and high-pulse parameter settings. When dynamically time-warping the current test voltage response sequence with the reference voltage response data, this matching process needs to handle potential timing offsets and scaling phenomena. The dynamic time warping algorithm can find the optimal alignment path between the two sequences. The voltage matching degree is calculated based on the degree of difference between the aligned sequences; a higher matching degree indicates better consistency between the current test results and the historical reference. The current matching degree is obtained using the same dynamic time warping algorithm to process the current sequence. Matching the current data is equally important because it reflects the power supply's load response characteristics. When correcting initial performance evaluation metrics based on voltage and current matching degrees, the correction algorithm needs to comprehensively consider the impact of both matching degrees, typically using a weighted average or geometric average to incorporate matching degree information into the evaluation metrics.

[0065] The implementation of the Dynamic Time Warping (VTW) algorithm requires careful attention to optimizing computational efficiency, especially when processing long-term series data. Sliding windows or segmented processing methods can be used to reduce computational complexity. The matching degree calculation considers not only the overall sequence similarity but also the matching of key feature points, such as the correspondence between peak and valley points. Parameter adjustments during the correction process need to be flexibly set according to actual testing requirements. For example, in applications with higher voltage stability requirements, the weight of voltage matching degree can be appropriately increased. The entire correction process must maintain the reasonable range of evaluation index values ​​to avoid values ​​exceeding the expected range due to correction.

[0066] Quality management of historical benchmark data is crucial, requiring regular verification of its validity and accuracy, and the removal of potentially problematic historical data. Benchmark data selection is sometimes not limited to a single historical test; the aggregated results of multiple historical test data sets can be used as a comprehensive benchmark, improving its representativeness and reliability. During the matching process between current test data and benchmark data, systematic differences that may arise from equipment aging and environmental factors must be considered, and compensation mechanisms should be introduced to eliminate the influence of these factors when necessary. The calculation results of voltage and current matching degrees can be analyzed separately, which helps identify different performance characteristics of the power supply in terms of voltage output and current response. If the voltage matching degree is significantly lower than the current matching degree, it may indicate a problem with the power supply's voltage regulation circuit; conversely, it may reflect changes in load response characteristics. This segmented analysis provides directional information for subsequent problem diagnosis. The matching degree threshold needs to be determined based on the power supply type and usage scenario; generally, a certain matching degree is required for the test results to be considered consistent with historical performance. The design of the correction algorithm allows for the introduction of a non-linear adjustment mechanism, using a larger correction magnitude when the matching degree is below a certain critical value, and a smaller adjustment magnitude when the matching degree is high. This non-linear correction can amplify the indicative effect of anomalies and improve the sensitivity of evaluation metrics to performance changes. The correction process also needs to consider the impact of the matching confidence level; for test segments with poor data quality, the weight of their matching degree can be reduced. The entire implementation process requires establishing a robust data recording and tracking mechanism to save the raw data, intermediate calculation results, and final evaluation metrics for each test. This data accumulation provides valuable information for subsequent analysis and improvement, especially when deviations between evaluation results and actual conditions are found; backtracking analysis can then be used to optimize algorithm parameters. Data records should include all key parameter settings, such as weighting coefficients, matching degree algorithm parameters, and threshold settings, to ensure the repeatability of the testing process.

[0067] The statistical analysis of outlier time points requires rigorous counting methods to avoid errors caused by improper data boundary handling. Outliers at the edge of the time window require special handling to ensure they are not missed or double-counted due to window slippage. The calculation of the total number of test time points must accurately reflect the actual number of valid data points, excluding invalid data points caused by acquisition failures. The weighting coefficients in the weighted summation formula can be optimized using machine learning methods, training historical test data to find the weight combination that best reflects the actual performance of the power supply. This data-driven weight determination method can improve the accuracy and reliability of evaluation metrics. The weighting coefficients can also be designed as configurable parameters, allowing test engineers to adjust them according to different testing objectives. During dynamic time warping matching, the choice of distance metric affects the matching degree calculation results; commonly used Euclidean or Manhattan distances each have their applicable scenarios. In some cases, distance metrics based on derivatives or curvature can also be used to better capture the shape characteristics of the waveform. Matching degree normalization needs to be performed carefully to ensure the comparability of matching degrees between different tests. The corrected performance evaluation metrics need to be verified for correlation with other test metrics to ensure they accurately reflect the performance status of the power supply. This verification process can be achieved through comparison with actual usage performance or cross-validation with results from other testing methods. Continuous performance indicator verification helps identify and improve the evaluation method. The entire implementation process needs to be robust, capable of handling various anomalies such as data acquisition interruptions, noise interference, or equipment failures. When anomalies occur, the system should be able to provide clear error indications rather than generating misleading evaluation results. Robustness is ensured through a robust data verification and anomaly handling mechanism. The final corrected performance evaluation index is a comprehensive quantitative value that integrates real-time test data, historical benchmark data, and professional weight settings. This index can serve as an important basis for power supply performance evaluation, providing data support for power supply quality judgment and application selection. The interpretation of the index value needs to be combined with specific test conditions and power supply specifications; different power supply models may have different index value ranges.

[0068] Example 3: When calculating the geometric mean of voltage and current matching degrees, this process requires ensuring that both matching degree values ​​are standardized and within the same numerical range, typically normalized to the interval between zero and one. The overall matching degree is obtained by multiplying the voltage and current matching degree values ​​and taking the square root. This calculation method balances the influence of the two parameters, preventing a single parameter from dominating the final result. When comparing the overall matching degree with a preset matching degree threshold, this threshold is usually determined based on statistical analysis of a large amount of historical test data, reflecting the typical matching level of similar power supplies under normal conditions. The process of obtaining the matching degree correction coefficient involves the quantitative conversion of the comparison results. When the overall matching degree is higher than the threshold, the correction coefficient will appropriately increase the initial evaluation index, and vice versa. When using the matching degree correction coefficient to linearly adjust the initial performance evaluation index, the adjustment process maintains the simplicity and interpretability of mathematical operations, ensuring that the final result does not contain unreasonable extreme values.

[0069] The modified performance evaluation index can be calculated using the following formula:

[0070]

[0071] in: This represents the revised performance evaluation metric. This represents the initial performance evaluation metric. It is the calculated overall matching degree value. This is the preset matching threshold. Parameter It is the baseline adjustment factor, used to maintain the basic balance of the correction results; It adjusts the amplitude coefficient to control the strength of the correction; It is the sensitivity coefficient, which affects the steepness of the corrected response curve; hyperbolic tangent function. This is to ensure that the correction process is smooth and limited, avoiding excessive adjustments.

[0072] The calculation of the geometric mean requires both matching degree values ​​to be of good quality. Significant errors in either matching degree will greatly affect the final result. Therefore, the validity of the voltage and current matching degrees needs to be verified before actual calculation, eliminating obviously unreasonable data points. The calculation of the overall matching degree needs to maintain sufficient numerical accuracy to avoid distortion of the final result due to rounding errors. The matching degree threshold is not fixed and can be dynamically adjusted according to the power supply type, service life, and importance. For power supplies in critical equipment, a stricter threshold standard can be used, while the requirements can be appropriately relaxed for general-purpose power supplies. This flexibility allows the evaluation method to adapt to different application scenarios and quality requirements. The calculation of the matching degree correction factor uses a smooth mathematical function. This design ensures that small fluctuations in the overall matching degree around the threshold will not cause drastic changes in the correction factor. This smoothing characteristic improves the stability of the evaluation results and avoids significant changes in the final evaluation conclusion due to small variations in test data.

[0073] During linear adjustment, special attention must be paid to the reasonableness of the numerical range to ensure that the corrected evaluation index remains within a meaningful numerical range. Sometimes, upper and lower limits need to be set to prevent the corrected index value from exceeding the expected range. This protection mechanism improves the robustness of the method and enables it to handle various boundary conditions. The selection of the hyperbolic tangent function is well-considered; this function possesses excellent mathematical properties: it is an odd function, symmetric about the origin; its output value is restricted to between -1 and +1; and it exhibits an approximately linear response near the origin. These properties allow the calculation of the correction coefficient to reflect both directionality (increase or decrease) and control over the adjustment magnitude. Parameters , and Determining the optimal parameter combination requires analysis of a large amount of experimental data, typically using statistical learning methods or optimization algorithms. Once determined, these parameters remain stable over a certain period to ensure the consistency of the evaluation method. Regularly recalibrating these parameters helps maintain the timeliness of the evaluation method.

[0074] The implementation of the correction process requires the establishment of corresponding quality control mechanisms, including logging of the calculation process, saving intermediate results, and verifying the final results. These measures facilitate traceability analysis when problems occur and promote continuous improvement of the method. Quality control also includes repeatability verification of the calculation process, ensuring that the same input data always produces the same output results. The calculation of the overall matching degree relies not only on numerical calculations but also on the matching of waveform characteristics. Sometimes, although the numerical matching degree is high, there are significant differences in waveform characteristics, requiring the introduction of additional correction factors. This feature-based matching analysis can improve the accuracy of the evaluation. Setting the matching degree threshold sometimes requires a multi-level threshold system, with different levels of thresholds triggering corrections of different intensities. For example, warning thresholds and critical thresholds can be set; when the overall matching degree is below the warning threshold, a mild correction is applied, and when it is below the critical threshold, a strong correction is applied. This tiered approach can more precisely reflect the actual state of the power supply. The application of correction coefficients needs to be transparent, and the magnitude and reason for the correction should be clearly explained in the test report. This transparency helps users understand the formation process of the evaluation results, increasing their trust in the evaluation conclusions. It can also provide feedback information for subsequent parameter optimization. The computational efficiency of the entire correction process needs to be guaranteed, especially in batch testing scenarios, to ensure that the correction calculation does not become a performance bottleneck. Optimizing algorithm implementation, using efficient mathematical libraries, and appropriate parallelization are all effective ways to improve efficiency. Efficiency optimization must be carried out while ensuring computational accuracy. The final corrected performance evaluation index is a more comprehensive and accurate quantitative value, which not only considers the statistical characteristics of the current test data but also combines the degree of matching with historical benchmarks. This index provides a more reliable basis for power supply performance evaluation and supports various subsequent application decisions. The interpretation of the index value needs to be combined with the specific context; the same value may represent different performance levels in different application scenarios. The scalability of the method is reflected in the ease with which other matching factors can be introduced, such as temperature matching or power consumption matching. By simply incorporating these factors into the calculation of the comprehensive matching degree in an appropriate manner, the dimensions and depth of the evaluation can be expanded.

[0075] Example 4: When acquiring test data from multiple power supplies of the same model under identical test conditions, it is necessary to extract data records that meet the criteria from the historical test database. These data should have the same test environment settings and instrument configurations. Taking the test of a certain model of switching power supply as an example, the reference dataset contains performance evaluation indicators of twenty power supply samples of the same model under standard test conditions. These data were collected in a standard experimental environment with a temperature of 25 degrees Celsius and a humidity of 60%. When calculating the similarity between the corrected performance evaluation indicators of the power supply under test and the performance evaluation indicators of each power supply in the reference dataset, a similarity measurement method based on Euclidean distance is used to compare the indicator values ​​of the power supply under test with the indicator values ​​of each power supply in the reference dataset one by one. Based on the similarity calculation results, the performance level of the power supply under test is determined. The higher the similarity value, the closer the performance characteristics of the power supply under test are to the reference power supply.

[0076] Reference power supplies with a similarity greater than a preset similarity threshold are marked as similar power supplies. The preset threshold is typically set around 0.7, a reasonable dividing point determined through extensive experimental analysis. The proportion of similar power supplies whose performance evaluation indicators are superior to the power supply under test is statistically analyzed. This proportion calculation requires a clearly defined standard for "superior," usually referring to the reference power supply's evaluation indicator value being a certain percentage higher than the power supply under test. This performance advantage score reflects the relative position of the power supply under test within the group of similar power supplies; the lower the score, the more outstanding the performance of the power supply under test. Based on the preset range of the performance advantage score, the performance level of the power supply under test is determined. The preset range is typically divided into three levels: below 30% is excellent, between 30% and 60% is good, and above 60% is average (see Table 1).

[0077] Table 1: Power Performance Evaluation Metrics in the Reference Dataset

[0078] Power supply number Performance evaluation index Test batch Running time (hours) Ambient temperature (°C) PS-2023-001 0.87 Batch 1 1200 25 PS-2023-002 0.92 Batch 1 1150 25 PS-2023-003 0.85 Batch 1 1250 25 PS-2023-004 0.89 Batch 2 980 25 PS-2023-005 0.91 Batch 2 1050 25 PS-2023-006 0.88 Batch 2 1100 25 PS-2023-007 0.84 Batch 3 850 25 PS-2023-008 0.90 Batch 3 900 25 PS-2023-009 0.86 Batch 3 950 25 PS-2023-010 0.93 Batch 3 920 25

[0079] In practice, the construction of the reference dataset must ensure data consistency and comparability. All reference power supplies should be tested under the same test procedures. Test conditions include, but are not limited to: identical load configurations, consistent ambient temperature and humidity, uniform instrument accuracy levels, and identical test durations. Detailed test metadata needs to be recorded during data acquisition, including auxiliary information such as the power supply's production batch, cumulative operating time, and test date. This information is helpful for subsequent in-depth analysis. Similarity calculation employs standardization, first normalizing all performance evaluation metrics to the range of zero to one to eliminate the influence of dimensions before calculating the distance metric. The Euclidean distance calculation result is converted into a similarity score using an exponential function, making the final similarity value more interpretable. The selection of the similarity threshold needs statistical validation, typically using receiver operating characteristic curve analysis to determine the optimal threshold point to balance classification accuracy and recall. The identification of similar power supplies is a dynamic screening process; as the reference dataset expands, the number and composition of similar power supplies may change. Therefore, a data update mechanism is needed to regularly incorporate new test data into the reference dataset, maintaining its timeliness and representativeness. At the same time, a data quality audit mechanism should be established to ensure that newly added data meets quality requirements.

[0080] Calculating performance superiority requires a clearly defined benchmark, typically using the evaluation index of the power supply under test as a reference point. The percentage of similar power supplies whose index values ​​are superior to this reference point is then calculated. Weighting factors can also be introduced into the calculation, assigning higher weights to reference power supplies with more recent production dates and more similar operating conditions, thus making the comparison results more valuable. Performance level classification needs to be tailored to the specific application requirements. For power supplies used in high-reliability applications, the classification standards should be more stringent; while for general-purpose power supplies, the standards can be more relaxed. The determination of the level classification boundary should be based on the distribution characteristics of a large amount of historical data, typically using the percentile method to determine a reasonable boundary.

[0081] The entire evaluation process requires a robust data recording and tracking mechanism to preserve all intermediate calculation results and final judgments. These records are not only used for quality control in this evaluation but also provide data support for subsequent method optimization. The records should include key information such as the version information of the reference dataset, similarity calculation parameters, threshold settings, and grading standards. The evaluation results need to be presented clearly and concisely. In addition to the final performance grade, detailed supporting data should be provided, including the number of similar power sources, the specific numerical value of performance advantage, and the distribution of power sources at each grade. This supplementary information helps users fully understand the formation process of the evaluation results and assess their reliability.

[0082] The method's flexibility lies in its adaptability to reference datasets of varying sizes, producing meaningful evaluation results regardless of whether there are only a few reference samples or a large amount of historical data. For datasets with limited data, methods such as nearest neighbor algorithms or kernel density estimation can be used to improve the stability of the evaluation. As reference data accumulates, the reliability and accuracy of the evaluation results gradually improve. Quality control measures include three stages: data validation, computational verification, and result review. Data validation ensures the completeness and rationality of the input data; computational verification checks the correctness of the calculation process; and result review involves professional personnel providing final confirmation of the evaluation results. These measures collectively guarantee the quality and reliability of the evaluation process. The final performance level evaluation results provide an important basis for power supply quality control and can be used in various scenarios such as factory inspection, periodic inspections, and fault analysis. The evaluation results can also be combined with other test indicators to form a more comprehensive power supply health status evaluation report, providing data support for equipment maintenance and management.

[0083] Example 5: When adjusting the high-voltage pulse parameters for subsequent tests based on performance levels, a targeted parameter adjustment strategy needs to be developed based on the obtained performance evaluation results. For example, if the power supply under test exhibits excellent performance in the initial test, the pulse amplitude and pulse time can be appropriately increased to further verify its performance limits. When using the adjusted high-voltage pulse parameters for secondary testing, the test engineer needs to reconfigure the output parameters of the high-voltage pulse generator, including setting new voltage peak value, pulse width, and repetition frequency. The adjustment range of these parameters is usually controlled within the range of 10% to 30% to avoid irreversible damage to the equipment. The process of acquiring secondary test data needs to maintain completely consistent acquisition conditions and environmental factors with the initial test. The same model of data acquisition equipment should be used to record voltage and current response data at the same sampling rate to ensure the comparability of the two test data. When comparing and analyzing the secondary test data with the initial test data, a strict data alignment mechanism needs to be established, using timestamp matching and sequence interpolation to eliminate possible minor differences in acquisition time points. Verifying the stability of the test results requires examination from multiple dimensions, including the consistency of waveform morphology, the repeatability of eigenvalues, and the similarity of statistical characteristics. When calculating the correlation coefficient of the voltage response sequences of the first and second tests, a standard statistical method based on covariance and standard deviation is used. This calculation process needs to consider the overall trend and local characteristics of the sequences. The correlation coefficient result serves as the voltage stability coefficient, reflecting the degree of consistency between the voltage waveforms of the two tests. The current stability coefficient is calculated using the same methodology, focusing on the repeatability and consistency characteristics of the current response.

[0084] When comparing the weighted sum of voltage stability coefficients and current stability coefficients with a preset stability threshold, the weighting needs to be adjusted according to the testing objective. If the testing focus is on voltage stability, the voltage coefficient should be given a higher weight; conversely, the weighting of the current coefficient should be increased. The decision-making process for determining whether the test results are stable requires clear judgment criteria. For example, a weighted sum exceeding 0.9 indicates good repeatability, while a weighted sum below 0.8 indicates a significant difference between the two tests. When the test results are stable, the revised performance evaluation indicators and performance levels are output as the final test results. These results need to be presented in a standardized report format, including all key data and supporting information. When the test results are unstable, readjusting the high-voltage pulse parameters requires a systematic analysis. First, potential influencing factors, including changes in environmental conditions, equipment connection status, and power supply preheating levels, should be investigated. Then, a new parameter adjustment plan should be developed based on the data analysis results of the first two tests. During the third test, a more conservative strategy is typically adopted for parameter adjustment, with the pulse parameter variation controlled within a small range of 5% to 15%, to gradually approach the optimal testing conditions. The test loop continues until stable test results are obtained. Each iteration requires complete recording of adjusted parameters, test data, and stability coefficients to form a complete test log.

[0085] In practice, parameter adjustment should follow a gradual approach. For example, a power supply model initially used a 1000V pulse amplitude and a 100-microsecond pulse width in its first test. When the test result was excellent, the pulse amplitude was increased to 1200V and the pulse width extended to 120 microseconds in the second test. Data from the second test showed a significant difference in the voltage response waveform compared to the first test, and the stability coefficient calculation showed a weighted sum of 0.76, lower than the set threshold of 0.85. In this case, the pulse parameters were adjusted back to 1100V and 110 microseconds in the third test, based on regression analysis of the data from the first two tests. Comparative analysis of the test data requires professional data processing tools to visualize the voltage and current sequences from both tests on the same coordinate system, observing the overlap and deviation distribution of the waveforms. The calculation of the stability coefficient must consider not only overall correlation but also the consistency of key characteristic parameters such as peak voltage, rise time, and overshoot amplitude. Any discovered instability requires in-depth analysis to determine the cause, which may be due to the power supply's thermal characteristics, component aging, or minor changes in the test system. The final test results require multiple verifications. Once stable results are obtained, repeatability testing is necessary—that is, conducting the test again under identical conditions to confirm the reliability of the results. All test data must be properly preserved and a complete traceability record established, including raw data, processing procedures, calculation results, and decision-making basis. Test report generation must follow a standardized template, clearly presenting the test conditions, methods, results, and conclusions, providing authoritative evidence for power supply performance evaluation. The entire implementation process reflects the scientific rigor and precision of the testing methodology. Through multiple iterative tests and parameter optimizations, the accuracy and reliability of the evaluation results are ensured. This method is particularly suitable for critical applications with stringent power supply performance requirements, providing reliable data support for power supply selection and use. The data and experience accumulated during the testing process also provide valuable resources for subsequent improvement and optimization of the testing methodology.

[0086] Although embodiments of the invention 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 to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A power supply testing method based on high-voltage pulse technology, characterized in that, The method includes: Acquire voltage response data and current response data of the power supply under test under preset high voltage pulse parameters. The voltage response data includes voltage values ​​at multiple test time points, and the current response data includes current values ​​corresponding to the test time points. The voltage response data and current response data are arranged in chronological order according to the test time to form a voltage response sequence and a current response sequence; Based on the voltage response sequence and current response sequence, calculate the voltage fluctuation coefficient and current fluctuation coefficient at each test time point; Based on the voltage fluctuation coefficient and current fluctuation coefficient, the initial performance evaluation index of the power supply under test is determined; The calculation of the voltage fluctuation coefficient and current fluctuation coefficient at each test time point includes: Select the voltage and current values ​​within a preset time range before and after the current test time point to form a local voltage sequence and a local current sequence, respectively; The ratio of the standard deviation to the mean of the local voltage sequence is calculated and used as the voltage fluctuation coefficient at the current test time point; The ratio of the standard deviation to the mean of the local current sequence is calculated and used as the current fluctuation coefficient at the current test time point; The voltage fluctuation coefficient is compared with a preset voltage fluctuation threshold, and the test time point that exceeds the preset voltage fluctuation threshold is marked as an abnormal test time point; The current fluctuation coefficient is compared with a preset current fluctuation threshold, and the test time point that exceeds the preset current fluctuation threshold is marked as an abnormal test time point. The determination of the initial performance evaluation indicators of the power supply under test includes: The ratio of the number of all abnormal test time points to the total number of test time points is used as the first abnormality ratio; Calculate the average value of all voltage fluctuation coefficients as the first stability parameter; The initial performance evaluation index is obtained by weighted summation of the first anomaly ratio and the first stability parameter.

2. The power supply testing method based on high-voltage pulse technology according to claim 1, characterized in that, The method further includes: Acquire the reference voltage response data and reference current response data of the power supply under test in historical tests; The voltage response sequence currently being tested is dynamically time-warped and matched with the reference voltage response data to obtain the voltage matching degree. The current response sequence being tested is dynamically time-warped and matched with the reference current response data to obtain the current matching degree. Based on the voltage matching degree and current matching degree, the initial performance evaluation index is corrected to obtain the corrected performance evaluation index.

3. The power supply testing method based on high-voltage pulse technology according to claim 2, characterized in that, The correction of the initial performance evaluation metric includes: Calculate the geometric mean of the voltage matching degree and the current matching degree as the overall matching degree; The overall matching degree is compared with a preset matching degree threshold to obtain a matching degree correction coefficient; The initial performance evaluation index is linearly adjusted using the matching degree correction coefficient to obtain the corrected performance evaluation index.

4. The power supply testing method based on high-voltage pulse technology according to claim 3, characterized in that, The method further includes: Collect test data from multiple power supplies of the same model under the same test conditions to form a reference dataset; Calculate the similarity between the corrected performance evaluation index of the power supply under test and the performance evaluation index of each power supply in the reference dataset; The performance level of the power supply under test is determined based on the similarity.

5. The power supply testing method based on high-voltage pulse technology according to claim 4, characterized in that, Determining the performance level of the power supply under test includes: Reference power sources with a similarity greater than a preset similarity threshold are marked as similar power sources; The proportion of similar power supplies whose performance evaluation indicators are better than those of the power supply under test is counted and used as the performance advantage. The performance level of the power supply under test is determined based on the preset range in which the performance advantage falls.

6. The power supply testing method based on high-voltage pulse technology according to claim 5, characterized in that, The method further includes: Based on the performance level, adjust the high-voltage pulse parameters for subsequent tests; A secondary test was conducted using the adjusted high-voltage pulse parameters to obtain secondary test data. The secondary test data was compared and analyzed with the primary test data to verify the stability of the test results.

7. The power supply testing method based on high-voltage pulse technology according to claim 6, characterized in that, The stability of the verification test results includes: Calculate the correlation coefficient between the voltage response sequences of the first test and the second test, and use it as the voltage stability coefficient; Calculate the correlation coefficient between the current response sequences of the first test and the second test, and use it as the current stability coefficient; The weighted sum of the voltage stability coefficient and the current stability coefficient is compared with a preset stability threshold to determine whether the test result is stable. When the test results are stable, the modified performance evaluation index and performance level will be used as the final test results; If the test results are unstable, readjust the high-voltage pulse parameters and perform a third test until stable test results are obtained.

8. A power supply testing system 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, When the processor executes the computer program, it implements the steps of the power supply testing method based on high voltage pulse technology as described in any one of claims 1 to 7.

Images