A method and system for automatically testing dynamic response characteristics of a special power supply

Abstract

The present application relates to the technical field of automation testing, in particular to a kind of special power dynamic response characteristic automatic testing method and system, comprising: obtaining target test point containing multiple working condition parameters, the target test point is any test point constructed by different working condition parameters, and the test result of target test point is obtained using digital signal processing algorithm;Using Gaussian process regression model, the predicted mean and predicted uncertainty of any target test point are obtained based on the target test point and its test result;The fluctuation degree and abnormal degree of target test point are calculated, and the global worst test point is obtained using improved bayesian optimization algorithm.The present application solves the problems of existing bayesian optimization algorithm in test efficiency, worst condition coverage ability and stability.

Description

Technical Field

[0001] This invention relates to the field of automated testing technology. More specifically, this invention relates to an automated testing method and system for the dynamic response characteristics of special power supplies. Background Technology

[0002] With the rapid development of aerospace, rail transportation, shipbuilding equipment, and high-end industrial control, special power supplies, as key basic energy equipment, are used in application scenarios with extremely high requirements for power supply reliability, dynamic response capability, and safety stability. Compared with general power supplies, special power supplies typically need to maintain stable operation under complex operating conditions such as large fluctuations in input voltage, drastic changes in load current, and rapid load transitions. Their dynamic response characteristics are directly related to the safety, functional integrity, and overall system reliability of downstream equipment.

[0003] However, the dynamic response behavior of special power supplies often exhibits significant nonlinearity, strong coupling, and operating condition sensitivity. Under different combinations of input voltage, load current, and load switching rate, the power supply's transient overshoot, voltage sag, and recovery time vary significantly, and the most unfavorable operating conditions are often distributed in local regions of a high-dimensional operating condition space. Traditional dynamic response testing methods typically employ manual experience-based point selection or rule-based scanning to perform point-by-point combination tests on operating parameters. This approach is not only inefficient and time-consuming but also prone to missing extremely unfavorable test points hidden in complex operating conditions, making it difficult to comprehensively reflect the power supply's worst-case dynamic performance under limited testing resources.

[0004] To improve testing efficiency, recent research has attempted to introduce optimization methods based on surrogate models, utilizing Bayesian optimization algorithms to adaptively select test points, aiming to quickly locate extreme response conditions of power supplies with fewer tests. However, existing Bayesian optimization algorithms in engineering applications generally employ fixed exploration coefficients or exploration strategies that only change with the number of iterations. These exploration coefficients remain constant or exhibit a singular change pattern throughout the testing process, failing to fully consider the differences in dynamic response risk, model uncertainty, and operating condition sensitivity among different test points. In special power supply dynamic response testing scenarios, due to the smooth response in some operating condition regions and the sudden and drastic changes in others, fixed exploration coefficients can easily lead to over-exploration in stable response regions, wasting test resources, or insufficient exploration near high-risk, high-uncertain operating conditions, resulting in premature convergence to local worst-case scenarios and difficulty in timely detection of the true global worst-case conditions. This leads to shortcomings in testing efficiency, worst-case coverage, and stability of existing Bayesian optimization algorithms. Summary of the Invention

[0005] To address the shortcomings of existing Bayesian optimization algorithms in terms of testing efficiency, worst-case coverage, and stability as mentioned in the background section, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides an automated testing method for the dynamic response characteristics of a special power supply, comprising: acquiring a target test point comprising multiple operating condition parameters, wherein the target test point is any test point constructed from different operating condition parameters, and acquiring the test results of the target test point using a digital signal processing algorithm; acquiring the predicted mean and predicted uncertainty of any target test point based on the target test point and its test results using a Gaussian process regression model; calculating the fluctuation degree of the target test point; calculating the anomaly degree of the target test point, wherein the anomaly degree is positively correlated with the fluctuation degree and predicted mean of the target test point, and negatively correlated with the mean of the fluctuation degree of all test points; and acquiring the global worst test point using an improved Bayesian optimization algorithm; wherein the improved Bayesian optimization algorithm includes an adaptive exploration coefficient, wherein the adaptive exploration coefficient is positively correlated with the initial value and the anomaly degree.

[0007] The above technical solution can continuously guide test points to the potentially most unfavorable operating condition area under limited test counts and time constraints, significantly reducing blind testing and repeated sampling, improving the efficiency and reliability of finding the worst global test point, and enhancing the stability and engineering practical value of the automated testing process in high-dimensional, strongly nonlinear dynamic response scenarios, providing more comprehensive and reliable technical support for the dynamic performance evaluation and reliability design of special power supplies.

[0008] Furthermore, target test points The degree of fluctuation for: , For predicting the mean function For the target test point The Middle First-order partial derivatives of each operating parameter For predicting the mean function For the target test point The Middle The second-order partial derivatives of each working condition parameter, For the target test point The total number of parameters under medium operating conditions.

[0009] The above technical solution introduces a comprehensive measure of the local variation characteristics of the prediction model at the target test point, so that the degree of fluctuation not only reflects the sensitivity of the prediction result to small changes in the operating parameters, but also can characterize the nonlinear bending and potential abrupt change trend of the prediction response in this region, thereby achieving a refined characterization of the instability of the dynamic response.

[0010] Furthermore, target test points abnormality for: , For the target test point The degree of fluctuation, This represents the average degree of fluctuation across all test points. To preset hyperparameters, For the natural constant An exponential function with base 0. For the target test point The predicted mean, This is a preset threshold.

[0011] The aforementioned technical solution couples the local fluctuation characteristics of the target test point with the predicted response level in its modeling, enabling the degree of anomaly to simultaneously reflect the intensity of dynamic changes near the operating condition and the risk level of the response amplitude. Furthermore, it avoids interference from extreme individual test points in anomaly assessment by introducing a global scale reference and a smoothing factor. The exponential mapping mechanism causes a nonlinear amplification effect when the predicted response approaches or exceeds the safety boundary, thereby significantly enhancing the discriminative power of high-risk operating conditions in anomaly identification. The normalized fluctuation factor highlights relatively unstable regions rather than regions with excessively large absolute values.

[0012] Furthermore, adaptive exploration coefficients for: , As the initial value, It is the hyperbolic tangent function. For the target test point The degree of abnormality.

[0013] The above technical solution dynamically correlates the exploration intensity with the anomaly degree of the target test point, enabling the Bayesian optimization algorithm to adaptively adjust the search strategy under different risk levels of operating conditions. When the anomaly degree of the test point is low, the exploration intensity remains moderate to avoid invalid or redundant tests in stable regions. As the anomaly degree increases, the exploration coefficient naturally amplifies, prioritizing the algorithm to sample more deeply in potentially high-risk or high-uncertain regions. At the same time, the smooth nonlinear mapping function avoids oscillations or instability caused by sudden changes in exploration intensity. This significantly improves the efficiency and reliability of discovering the most unfavorable global operating conditions, reduces the number of blind tests, accelerates the convergence of the testing process, and enhances the robustness and engineering applicability of the dynamic response automated testing method in multidimensional, highly nonlinear operating condition spaces.

[0014] Furthermore, to obtain the worst global test point, specifically: the improved Bayesian optimization algorithm is used to obtain the test point with the largest adaptive acquisition function, and the test point with the largest adaptive acquisition function is taken as the next optimal test point. All test points are iterated until a preset stopping condition is met to obtain the worst global test point.

[0015] The aforementioned technical solution, by introducing an adaptive acquisition strategy within a Bayesian optimization framework, achieves continuous exploration and dynamic updating of potential extreme operating conditions. This ensures that each test prioritizes key test points with both high predictive response and high uncertainty, thus avoiding the problems of traditional methods easily getting stuck in local worst-case scenarios or wasting test resources in low-risk areas. By iteratively updating the prediction model and continuously adjusting the selection of the next test point, this method can gradually approach and lock the global worst-case test point under limited test runs or time constraints. This effectively improves the efficiency and accuracy of identifying the most unfavorable operating conditions, while enhancing the coverage and robustness of the dynamic response testing process across complex, multi-dimensional operating conditions. This provides a reliable engineering basis for the dynamic performance evaluation and safety margin analysis of special power supplies.

[0016] Furthermore, the preset stopping condition is reaching the maximum number of tests or the maximum test time.

[0017] Furthermore, the adaptive acquisition function for: , , These are the target test points. The predicted mean and prediction uncertainty, For adaptive exploration coefficients.

[0018] Furthermore, the multiple operating parameters include input voltage, load current, and load switching rate.

[0019] Furthermore, the test result is the overshoot voltage.

[0020] In a second aspect, the present invention provides an automated testing system for the dynamic response characteristics of a special power supply, comprising a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an automated testing method for the dynamic response characteristics of a special power supply as described above is implemented.

[0021] The beneficial effects of this invention are as follows:

[0022] This invention deeply integrates digital signal processing, predictive modeling, and an improved Bayesian optimization method to achieve automated and intelligent testing of the dynamic response characteristics of special power supplies under multidimensional and complex operating conditions. It can efficiently acquire key test data with limited testing resources, thereby prioritizing the identification of potentially worst-case operating conditions. An adaptive exploration strategy enables the testing process to dynamically adjust the search focus based on the risk of the operating condition, balancing the utilization of high-predictability response regions with the exploration of high-uncertainty regions, significantly improving the efficiency and reliability of finding the global worst-case test point. Simultaneously, through multidimensional operating condition parameter modeling, acquisition of dynamic response indicators such as overshoot voltage, and Gaussian process regression prediction, it achieves forward-looking perception of complex nonlinear response characteristics and hidden high-risk regions. Overall, it enhances the stability, accuracy, and engineering practicality of the testing system under high-dimensional, multi-parameter, and strongly nonlinear operating conditions, providing comprehensive and reliable technical support for the reliability assessment, safety margin analysis, and subsequent optimization design of special power supplies. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an automated testing method for the dynamic response characteristics of a special power supply according to an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram illustrating the Gaussian process regression model predicting the mean of an automated testing method for the dynamic response characteristics of a special power supply according to an embodiment of the present invention.

[0025] Figure 3 This is a schematic block diagram illustrating the structure of an automated testing system for the dynamic response characteristics of a special power supply according to an embodiment of the present invention. Detailed Implementation

[0026] An embodiment of an automated testing method for the dynamic response characteristics of a special power supply.

[0027] like Figure 1 As shown, a flowchart of an automated testing method for the dynamic response characteristics of a special power supply according to an embodiment of the present invention includes the following steps:

[0028] S1: Obtain the target test point containing multiple operating parameters.

[0029] In a preferred embodiment, the target test point is any test point constructed based on different combinations of operating condition parameters. The operating condition parameters are used to comprehensively characterize the electrical stress characteristics and load response characteristics of the tested object during dynamic operation. These multiple operating condition parameters include at least input voltage, load current, and load transition rate. Specifically, the input voltage characterizes changes in power supply conditions on the power supply side, the load current reflects the steady-state or quasi-steady-state operating state of the tested object under different load levels, and the load transition rate describes the intensity of dynamic changes in load as it changes from one operating state to another. Furthermore, the operating condition parameters are normalized.

[0030] Furthermore, when constructing target test points, a full factorial design or adaptive sampling strategy can be adopted to combine input voltage, load current, and load switching rate to form a set of test points covering the boundary and intermediate areas of the operating space, thereby avoiding modeling deviations caused by test points being concentrated in local areas.

[0031] After acquiring the target test point, a digital signal processing algorithm is used to process the voltage response signal corresponding to the target test point to obtain the test result of the target test point. The digital signal processing algorithm includes, but is not limited to, signal denoising, sampling synchronization, transient feature extraction, and peak detection algorithms. By filtering and time-domain analysis of the original voltage waveform, the transient features of the voltage response when the load changes are accurately extracted, and the maximum deviation value in the transient response is taken as the test result of the target test point; the test result is the overshoot voltage.

[0032] After obtaining the target test point and its corresponding overshoot voltage test results, a Gaussian process regression model is used to model the mapping relationship between the target test point and the test results.

[0033] like Figure 2 The diagram shows the Gaussian process regression model predicting the mean of an automated testing method for the dynamic response characteristics of a special power supply according to an embodiment of the present invention.

[0034] Specifically, the input voltage, load current, and load switching rate contained in the target test point are used as model input variables, and the overshoot voltage is used as the model output variable. By training a Gaussian process regression model, the predicted mean and prediction uncertainty of any target test point are obtained.

[0035] The predicted mean is used to characterize the expected response level of the overshoot voltage under given operating parameters, and the predicted uncertainty is used to quantify the reliability of the model or the potential error range at the target test point.

[0036] Through the above steps, an overshoot voltage prediction model that balances prediction accuracy and uncertainty description capability can be constructed under limited test sample conditions, thereby effectively reducing the need for a large number of repetitive physical tests, lowering test costs, and improving test efficiency.

[0037] S2: Calculate the degree of fluctuation and anomaly of the target test point.

[0038] In a preferred embodiment, the target test point The degree of fluctuation for: , For predicting the mean function For the target test point The Middle First-order partial derivatives of each operating parameter For predicting the mean function For the target test point The Middle The second-order partial derivatives of each working condition parameter, For the target test point The total number of parameters under medium operating conditions. The predicted mean function. The posterior predictive mean function is obtained by training the completed test points using a Gaussian process regression model. The Gaussian process regression model takes the operating parameters of the target test points as input and the corresponding dynamic response test results as output, and uses a preset kernel function to characterize the changing trend of dynamic response characteristics under different combinations of operating parameters.

[0039] By quantifying the local variation characteristics of the prediction model at the target test point, the sensitivity and nonlinear variation trend of the model output to small disturbances in the operating parameters are uniformly characterized. The first-order variation term reflects the instantaneous response strength of the prediction result to changes in the operating parameters, characterizing the system's sensitivity near the test point. The second-order variation term characterizes the curvature and nonlinear abrupt change trend of the prediction function in this region, effectively identifying potential severe fluctuations or unstable intervals. By integrating the two, a unified measure of the overall fluctuation characteristics of the target test point can be formed within a multi-dimensional operating space. This not only accurately distinguishes between safe operating conditions with gently changing prediction results and high-risk operating conditions highly sensitive to parameter disturbances, but also reveals implicit unstable regions that are difficult to identify traditionally based solely on the magnitude of predicted values. This provides a more reliable and forward-looking decision-making basis for test point selection, key operating condition identification, and subsequent control strategy optimization, significantly improving the practicality and engineering adaptability of model-driven testing and control methods under complex operating conditions.

[0040] Target test point abnormality for: , For the target test point The degree of fluctuation, This represents the average degree of fluctuation across all test points. To preset hyperparameters, For the natural constant An exponential function with base 0. For the target test point The predicted mean, A preset threshold is used. Further, the degree of anomaly is normalized.

[0041] The anomaly discrimination of target test points, driven by a single prediction result, is expanded into a composite measurement mechanism that simultaneously integrates local fluctuation characteristics and predicted amplitude levels. Specifically, by normalizing the fluctuation degree and introducing a smoothing term, different test points are made comparable on an overall scale, thus highlighting regions highly sensitive to disturbances in the operating space. Simultaneously, the prediction result is mapped exponentially, causing a nonlinear amplification effect on the anomaly degree when the predicted value approaches or exceeds a safety threshold, thereby enhancing the discriminative power of high-risk operating conditions. This coupling ensures that the anomaly degree reflects both the severity of the predicted response itself and the drastic changes in the model near the operating condition, avoiding misjudgments that ignore potential instability based solely on prediction magnitude.

[0042] S3: Use an improved Bayesian optimization algorithm to obtain the worst global test point.

[0043] In a preferred embodiment, the improved Bayesian optimization algorithm includes adaptive exploration coefficients, wherein the adaptive exploration coefficients... for: , As the initial value, It is the hyperbolic tangent function. For the target test point The degree of abnormality.

[0044] The trade-off mechanism between exploration and exploitation in Bayesian optimization is transformed from fixed-coefficient control to a dynamic adjustment mechanism driven by the adaptive degree of anomaly. By introducing a nonlinear mapping function with smooth saturation characteristics, the exploration intensity can be continuously adjusted according to changes in the degree of anomaly. When the degree of anomaly is low, a relatively restrained exploration intensity is maintained, thus avoiding invalid or redundant sampling in the stable operating condition region. When the degree of anomaly increases significantly, the exploration coefficient is rapidly amplified to enhance the search capability in high-risk or high-uncertainty regions. This avoids optimization oscillations caused by abrupt changes in exploration parameters and ensures the stability of the algorithm through natural upper and lower bound constraints.

[0045] To obtain the worst-case test point globally, the following steps are taken: A modified Bayesian optimization algorithm is used to find the test point with the maximum adaptive acquisition function. This test point is then used as the next optimal test point. The process is iterated over all test points until a preset stopping condition is met, thus obtaining the worst-case test point globally. The preset stopping condition is reaching the maximum number of tests or the maximum test time. The adaptive acquisition function... for: , , These are the target test points. The predicted mean and prediction uncertainty, For adaptive exploration coefficients.

[0046] By introducing an adaptive acquisition strategy, the utilization of high-risk prediction results and the exploration of high-uncertainty regions are dynamically balanced during iterative testing. This allows test point selection to move beyond currently known extreme value locations and continuously expand the search range to potentially worse operating conditions. Driven by the adaptive exploration mechanism, the acquisition function prioritizes test points with both large predictive responses and high uncertainty, thus avoiding getting trapped in local minima due to early model biases. By repeatedly updating the model and iteratively selecting the next test point under preset test counts or time constraints, the system can gradually approach and ultimately lock onto the global worst-case test point under limited testing resources, achieving efficient identification of extremely unfavorable operating conditions.

[0047] This invention organically combines multi-dimensional operating condition parameters, digital signal processing, predictive modeling, and adaptive optimization methods to automate and intelligently test the dynamic response characteristics of special power supplies. It can efficiently identify the worst-case operating condition in a high-dimensional, nonlinear, and multi-condition space. By using predictive models to obtain the mean and uncertainty of test points, and combining this with fluctuation and anomaly levels to comprehensively quantify risk, it can not only identify potential high-risk operating conditions in advance, but also, based on this, dynamically balance the utilization of known extreme responses with the exploration of unknown high-risk areas through adaptive adjustment of the exploration strategy. This significantly improves the efficiency and accuracy of identifying the worst-case test point. Accurate evaluation of key dynamic response indicators can be achieved within a limited number of tests or time constraints, effectively reducing blind or redundant testing, ensuring the reliability and repeatability of test results, and providing efficient and reliable technical support for the safety assessment, design optimization, and reliability verification of special power supplies.

[0048] An example of an automated testing system for the dynamic response characteristics of a special power supply:

[0049] like Figure 3As shown in the diagram, an automated testing system for the dynamic response characteristics of a special power supply according to an embodiment of the present invention includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the automated testing method for the dynamic response characteristics of a special power supply according to the present invention.

[0050] The automated testing system for the dynamic response characteristics of special power supplies also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0051] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0052] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0053] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. An automated testing method for the dynamic response characteristics of a special power supply, characterized in that, include: A target test point containing multiple operating condition parameters is obtained. The target test point is any test point constructed with different operating condition parameters. The test result of the target test point is obtained by using a digital signal processing algorithm. The test result is the overshoot voltage. The prediction mean and prediction uncertainty of any target test point are obtained by using a Gaussian process regression model based on the target test point and its test result. Calculate the degree of fluctuation at the target test point , , For predicting the mean function For the target test point The Middle First-order partial derivatives of each operating parameter For predicting the mean function For the target test point The Middle The second-order partial derivatives of each working condition parameter, For the target test point The total number of parameters under medium operating conditions; calculate the degree of anomaly at the target test point, wherein the degree of anomaly is positively correlated with the fluctuation degree and predicted mean of the target test point, and negatively correlated with the mean of the fluctuation degree of all test points; The global worst test point is obtained using an improved Bayesian optimization algorithm: The test point with the largest adaptive acquisition function is obtained using an improved Bayesian optimization algorithm, and the test point with the largest adaptive acquisition function is taken as the next optimal test point. The process is repeated for all test points until a preset stopping condition is met to obtain the global worst test point. The improved Bayesian optimization algorithm includes an adaptive exploration coefficient, which is positively correlated with the initial value and the degree of anomaly.

2. The automated testing method for the dynamic response characteristics of a special power supply according to claim 1, characterized in that, Target test point abnormality for: , For the target test point The degree of fluctuation, This represents the average degree of fluctuation across all test points. To preset hyperparameters, For the natural constant An exponential function with base 0. For the target test point The predicted mean, This is a preset threshold.

3. The automated testing method for the dynamic response characteristics of a special power supply according to claim 1, characterized in that, Adaptive exploration coefficient for: , As the initial value, It is the hyperbolic tangent function. For the target test point The degree of abnormality.

4. The automated testing method for the dynamic response characteristics of a special power supply according to claim 1, characterized in that, The preset stopping condition is reaching the maximum number of tests or the maximum test time.

5. The automated testing method for the dynamic response characteristics of a special power supply according to claim 1, characterized in that, The adaptive acquisition function for: , , These are the target test points. The predicted mean and prediction uncertainty, For adaptive exploration coefficients.

6. The automated testing method for the dynamic response characteristics of a special power supply according to claim 1, characterized in that, The multiple operating parameters include input voltage, load current, and load switching rate.

7. An automated testing system for the dynamic response characteristics of a special power supply, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an automated testing method for the dynamic response characteristics of a special power supply as described in any one of claims 1 to 6 is implemented.

Images