A method and system for evaluating performance of a direct current power supply

Abstract

The application discloses a kind of performance evaluation method and system of direct-current power supply, specifically relates to power supply test technical field, and by reasonably dividing power supply voltage interval, and according to load demand and power output ability corresponding parameter is set, ensure that the output stability of power supply under different load working conditions;Combining dynamic response performance acquisition and high-precision data synchronization technology, real-time monitoring of dynamic response data of power supply, ensure that in the process of load change and voltage regulation, power supply can maintain stable output;Through nonlinear regression analysis, accurately model the working efficiency of power supply in different voltage intervals, power loss and temperature fluctuation, identify potential failure risk in advance;Combined with the maximum electrical stress and thermal stress of power supply components, set abnormal determination threshold, timely discover and early warning possible failure point of power supply, the method can comprehensively evaluate the working performance of power supply and provide optimization suggestion, solve the problem of insufficient early warning of potential failure in traditional power supply test.

Description

Technical Field

[0001] This invention relates to the field of power supply testing technology, and more specifically, to a method and system for evaluating the performance of a DC power supply. Background Technology

[0002] In the field of modern electronic equipment testing and R&D, DC power supplies are widely used to provide power to various devices under test. Since the operating voltage ranges of the devices under test typically vary, DC power supplies must be able to provide a wide voltage range to meet the needs of different applications.

[0003] Traditional power supply performance testing methods fail to account for the nonlinear variations in power supply performance across different voltage ranges. DC power supplies consist of multiple electronic components, and their efficiency, output quality, and thermal characteristics change significantly with voltage variations. At low voltage outputs, certain power supply losses may increase significantly; while at high voltage outputs, critical components (such as power transistors and transformers) experience intensified electrical and thermal stresses, impacting stability and reliability. Existing point-to-point testing methods cannot comprehensively and systematically capture performance fluctuations and critical states across the entire voltage range. This means that in practical applications, especially at untested voltage points, power supplies may experience sudden performance drops, output instability, or even malfunctions, affecting the accuracy of test results and potentially damaging high-value equipment under test. Furthermore, fault diagnosis is difficult and risks are unpredictable, increasing subsequent maintenance costs and reducing production testing efficiency.

[0004] Therefore, existing technologies suffer from the inability to comprehensively and systematically evaluate the overall performance stability of DC power supplies over a wide voltage range. Solving this problem requires a new method that uses continuous, dynamic performance testing and early warning mechanisms to assess the power supply's operating status under different voltage conditions, ensuring stable and reliable operation across the entire voltage range. Summary of the Invention

[0005] To overcome the aforementioned deficiencies in the prior art, this invention provides a DC power supply performance evaluation method and system. By rationally dividing the voltage range, collecting dynamic response data in real time, and combining it with nonlinear regression analysis, the system accurately evaluates the power supply's operating efficiency, power loss, and fault risk. In conjunction with the physical limitations of the power supply components, it sets anomaly judgment threshold and provides fault warnings, thereby solving the problem that traditional power supply performance testing cannot accurately predict power supply fault risks and stability.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for evaluating the performance of a DC power supply, comprising the following steps: Step S1: Voltage range segmentation and power index setting Based on the rated output range of the target DC power supply, the rated output range is divided into multiple voltage sub-ranges, and corresponding power output indicators are set for each sub-range to ensure coverage of key operating states under different voltage ranges; the voltage range of each sub-range should ensure that the corresponding output power does not exceed the design carrying capacity of the power supply. Step S2: Dynamic Response Performance Acquisition Within each voltage sub-range, dynamic voltage regulation is implemented to gradually adjust the power supply output voltage to the set range, and dynamic response data is recorded in real time, including multiple parameters such as power supply output current, voltage, and temperature. Special attention is paid to the impact of instantaneous voltage fluctuations and load changes on output stability during power supply operation. Step S3: Nonlinear performance evaluation and fault early warning modeling Based on the collected dynamic response data, nonlinear regression analysis is used to model the power supply's operating efficiency, power loss, and temperature changes of key components in different voltage ranges; by establishing a mathematical model, the critical performance boundaries near each voltage point are calculated, and potential failure risks are predicted based on historical data and trends. Step S4: Critical State Analysis and Anomaly Determination By analyzing the obtained mathematical model, we can identify potential performance degradation points of the power supply near the critical voltage range and establish anomaly judgment thresholds. By combining the maximum electrical and thermal stress of key components, we can set the tolerance range for performance fluctuations and mark fault risk points (voltage points where unstable output or faults may occur, i.e., voltage points where the risk probability exceeds the preset requirements). Step S5: Performance Reporting and Feedback Mechanism Based on the test results across various voltage ranges, a performance evaluation report is generated. This report should include information such as the power supply's performance, dynamic response characteristics, power loss, temperature fluctuations, and potential fault risks across each voltage range, serving as a comprehensive basis for power supply performance verification. A report feedback mechanism enables real-time monitoring and optimization of subsequent power supply design and testing processes.

[0007] Preferably, in step S1, the division of the voltage sub-intervals adopts a non-equal division method, specifically: based on the efficiency-voltage characteristic curve of the power supply, the boundary of the voltage segment in the curve where the rate of change of efficiency with voltage exceeds a preset first threshold is taken as the division point of the sub-interval. In one possible embodiment, the range of the rate of change of efficiency with voltage is 2% to 5% / V.

[0008] Preferably, the efficiency-voltage characteristic curve of the power supply is obtained by performing a pre-scan test with a constant current load and a fixed sampling interval within the rated output range and calculating the efficiency value in real time.

[0009] Preferably, in step S2, the dynamic voltage regulation adopts a predefined voltage-time waveform sequence, which includes at least one voltage step change and a voltage ramp change in each voltage sub-interval. In one possible embodiment, the step amplitude is set to 10% to 20% of the voltage span of the sub-interval, and the ramp change rate is set to 0.5 to 2V / ms.

[0010] Preferably, in step S3, the nonlinear regression analysis adopts the support vector regression (SVR) algorithm, with voltage value, output current value and load change rate as input features, and instantaneous power supply efficiency and temperature rise of key components as output targets, to establish a predictive model of power supply performance, and use the predictive model to perform boundary search to obtain the critical performance boundary point where the efficiency decreases by more than 5% or the temperature rises by more than 15% of the rated value.

[0011] Preferably, the support vector regression algorithm uses a radial basis function as the kernel function, with the kernel width parameter γ ranging from 0.01 to 0.1, and the tolerance ε of the insensitive loss function is set to 0.5%, while the penalty coefficient C ranges from 1 to 100. The hyperparameters required for training the prediction model are optimized and determined by a combination of grid search and cross-validation. When using the prediction model to perform boundary search for critical performance boundary points, the voltage is kept constant within each voltage sub-interval, the output current value is increased incrementally, and the model is used to predict the corresponding instantaneous efficiency and temperature rise of key components. The (voltage, current) coordinate points corresponding to the first time the instantaneous efficiency decreases by more than 5% or the first time the temperature rise exceeds 15% of the component's rated maximum allowable temperature rise are marked as critical performance boundary points.

[0012] Preferably, in step S4, the setting of the anomaly determination threshold is dynamic. The specific method is as follows: based on the predicted temperature rise ΔT(V) of the component corresponding to the current output voltage V output by the prediction model, combined with the real-time junction-to-case thermal impedance Zth(jc) and temperature coefficient k obtained from the component datasheet, the dynamic current limit value I_lim(V) = P_max / (V×(1+k×ΔT(V))) that changes with the output voltage is calculated, where P_max is the maximum allowable power consumption of the component; the value of the real-time junction-to-case thermal impedance Zth(jc) corresponds to the instantaneous thermal impedance value jointly determined by the power loss P_loss(V) and temperature rise ΔT(V) of the component under the current operating state predicted by the prediction model.

[0013] Preferably, in step S2, the amplitude of the voltage step change is adaptively adjusted; the specific adjustment method is as follows: based on the prediction results of the prediction model for the instantaneous efficiency and temperature rise of the current voltage point and the target step voltage point, the optimization target is to make the predicted efficiency decrease value after the step within the range of 3% to 8% or the predicted temperature rise value within the range of 10% to 20% of the rated temperature rise of the component, and the final step amplitude is dynamically determined by iterative calculation.

[0014] Preferably, in step S5, the performance evaluation report automatically generates a safe operating area map of the power supply across the entire voltage range. This map uses voltage as the horizontal axis and current as the vertical axis, and marks the prohibited operating area, derating operating area, and full-power operating area enclosed by each fault risk point determined in step S4 with different colors or patterns. By monitoring environmental factors (such as temperature and humidity) in real time, the safe operating area map is dynamically adjusted to optimize the maximum allowable current and power output of the power supply in real time according to environmental changes.

[0015] Preferably, in step S3, after obtaining the critical performance boundary point by using the prediction model to perform boundary search, the dynamic credibility verification of the boundary point is also performed: the voltage stress concentration index and the thermal accumulation steep change index are calculated; these two indices are input into the predefined credibility evaluation model to calculate the boundary credibility coefficient; when the boundary credibility coefficient is lower than the preset second threshold, the hyperparameters of the support vector regression algorithm are readjusted and the modeling process is iterated. The calculation process of the voltage stress concentration index is as follows: within the target voltage sub-interval, calculate the spectral entropy value of the voltage fluctuation rate, that is, perform a fast Fourier transform on the instantaneous voltage fluctuation sequence collected in the dynamic test to extract the distribution dispersion of the main frequency components; calculate the correlation coefficient between the load change rate and the voltage recovery time within the sub-interval; multiply the spectral entropy value and the correlation coefficient, and perform normalization to obtain the voltage stress concentration index; The calculation process of the thermal accumulation steep change index is as follows: extract the temperature rise curve of the key component in dynamic testing, calculate the coefficient of variation of the temperature rise rate, that is, the ratio of the standard deviation of the temperature rise rate to the mean; calculate the phase lag angle of the temperature rise relative to the voltage change rate, and determine the delay characteristics of the temperature response relative to the voltage change through cross-correlation analysis; and obtain the thermal accumulation steep change index by weighted summation of the coefficient of variation and the sine value of the phase lag angle. The predefined credibility assessment model is: Boundary credibility coefficient = 1 / (1+exp(-k1×voltage stress concentration index+k2×thermal accumulation steep change index)), where k1 and k2 are model calibration parameters, which are obtained through machine learning training of historical test data.

[0016] To achieve the above objectives, the present invention provides the following technical solution: a DC power supply performance evaluation system, comprising: Range configuration module: Based on the rated output range of the target DC power supply, the rated output range is divided into multiple voltage sub-ranges, and corresponding power output indicators are set for each sub-range to ensure coverage of key operating states under different voltage ranges; Dynamic testing module: Within each voltage sub-range, dynamic voltage regulation is implemented to gradually adjust the power supply output voltage to the set range, and the dynamic response data of the power supply output is recorded in real time, including the current, voltage, temperature, instantaneous voltage fluctuations during the operation of the power supply, as well as the impact of load changes on output stability. Intelligent modeling module: Based on the collected dynamic response data, it uses nonlinear regression analysis to model the power supply's operating efficiency, power loss, and temperature changes of key components in different voltage ranges; by establishing mathematical models, it calculates the critical performance boundaries near each voltage point and predicts potential failure risks based on historical data and trends. Risk Analysis Module: By analyzing the obtained mathematical model, it identifies the performance degradation points of the power supply near the critical voltage range and establishes anomaly judgment thresholds; combined with the maximum electrical stress and thermal stress of key components, it sets the tolerance range for performance fluctuations and marks the fault risk points. Report generation module: Combines test results from various voltage ranges to generate a performance evaluation report. The performance evaluation report includes the power supply's performance, dynamic response characteristics, power loss, temperature fluctuations, and fault risk information in each voltage range, serving as a global basis for power supply performance verification.

[0017] The voltage sub-range division scheme and power index output by the range configuration module are transmitted to the dynamic test module as test instructions; the real-time dynamic response data collected by the dynamic test module are transmitted to the intelligent modeling module as raw input; the mathematical model and predicted critical performance boundary constructed by the intelligent modeling module are input to the risk analysis module as the basis for analysis; the abnormal thresholds and fault risk point information identified by the risk analysis module are finally summarized to the report generation module to generate a comprehensive evaluation report that includes the performance across the entire voltage range.

[0018] The technical effects and advantages of this invention are as follows: (1) This invention ensures the stability and reliability of the power supply under various operating conditions by dividing the voltage range and setting the power output capability based on the DC power supply, combined with precise load simulation and real-time data acquisition technology. Reasonable voltage range segmentation and power setting avoid the risk of failure caused by exceeding the design capacity during actual use. High-precision measurement equipment and data synchronization technology are used to monitor key parameters such as current, voltage, and temperature in real time, ensuring the dynamic response stability of the power supply under different load fluctuations. This method can ensure that the power supply output performance is not affected under conditions of large load fluctuations, effectively avoiding the problem of power supply performance instability caused by load fluctuations.

[0019] (2) This invention uses a nonlinear regression model and multivariate analysis to accurately identify the nonlinear characteristics of the power supply in different voltage ranges and predict the failure risk of the power supply under specific operating conditions. By combining the maximum electrical stress and temperature rise limits of the power supply components and setting reasonable anomaly judgment thresholds, potential fault points of the power supply can be detected in a timely manner and warnings can be issued. By integrating and comparing historical data, the model can accurately simulate the working behavior of the power supply and generate performance prediction reports, providing data support for power supply design optimization. This solution effectively solves the problem of failing to detect potential risks in a timely manner in traditional power supply performance evaluation through a precise fault warning mechanism, thereby improving the working safety and service life of the power supply. Attached Figure Description

[0020] Figure 1 A flowchart of a DC power supply performance evaluation method provided by the present invention.

[0021] Figure 2 The present invention provides a structural block diagram of a DC power supply performance evaluation system. Detailed Implementation

[0022] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0023] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0024] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the scope of this application and its application or use.

[0025] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0026] Example 1, see Figure 1 A flowchart of a DC power supply performance evaluation method is provided in this invention. Figure 1 The method for evaluating the performance of a DC power supply, as shown, includes the following steps: Step S1: Voltage range segmentation and power index setting Based on the rated output range of the target DC power supply, the rated output range is divided into multiple voltage sub-ranges, and corresponding power output indicators are set for each sub-range to ensure coverage of key operating states under different voltage ranges; the voltage range of each sub-range should ensure that the corresponding output power does not exceed the design carrying capacity of the power supply. Specifically, the process of step S1 includes the following operations: First, based on the rated output range of the target DC power supply and the power requirements of the device under test, the power supply's output voltage range is rationally divided into multiple sub-ranges, each matching its voltage range and power output capability. This division process considers potential load fluctuations to ensure the power supply does not exceed its design capacity during testing. The voltage range of each sub-range should accurately reflect the power supply's performance under different loads, ensuring the power supply's output power does not exceed its maximum design capacity. Within each sub-range, appropriate load simulation parameters are set based on the relationship between required power and current to ensure the power supply's output stability under different load conditions.

[0027] Step S2: Dynamic Response Performance Acquisition Within each voltage sub-range, dynamic voltage regulation is implemented to gradually adjust the power supply output voltage to the set range, and dynamic response data is recorded in real time, including multiple parameters such as power supply output current, voltage, and temperature. Special attention is paid to the impact of instantaneous voltage fluctuations and load changes on output stability during power supply operation.

[0028] Step S3: Nonlinear performance evaluation and fault early warning modeling Based on the collected dynamic response data, nonlinear regression analysis is used to model the power supply's operating efficiency, power loss, and temperature changes of key components in different voltage ranges; by establishing a mathematical model, the critical performance boundaries near each voltage point are calculated, and potential failure risks are predicted based on historical data and trends. Specifically, the process of step S3 includes the following operations: First, the collected data is preprocessed to remove noise and outliers, ensuring data accuracy. Then, based on this data, a nonlinear regression model of the power supply across different voltage ranges is constructed. Parameters such as voltage, current, power loss, and temperature change are used as independent variables, while power supply efficiency and power loss are used as dependent variables. Least squares regression methods are employed for fitting the model. The core of the model is to identify the performance variation patterns of the power supply across different voltage ranges. For example, power efficiency decreases at low voltage output; while at high voltage output, key components of the power supply (such as power transistors and transformers) experience greater electrical and thermal stress, leading to a sharp increase in power loss and temperature rise. Through multivariate analysis, the nonlinear relationship between power supply performance and voltage is further explored, establishing an accurate regression model to predict the critical operating state of the power supply and identify potential fault risks. The output of the regression model can be used to generate power supply performance prediction reports, assess the stability of the power supply under different voltage ranges and load conditions, and generate fault warning indicators. Threshold judgments automatically identify potential fault points, providing early warnings to operators and ensuring that the power supply does not fail.

[0029] In this invention, key components primarily refer to the core components of a DC power supply that significantly impact its performance and stability. The operating state of these components directly affects the power supply's output quality and failure risk. Specifically, this includes, but is not limited to, the following aspects: Power conversion components: such as power switches, rectifier diodes, and power transistors (e.g., MOSFETs, IGBTs). These components are responsible for voltage conversion and regulation of the power supply and bear a large amount of electrical and thermal stress during load fluctuations and voltage regulation. Filtering and voltage regulation components: such as capacitors, inductors and voltage regulators. These components are responsible for filtering out high-frequency noise in the power supply and maintaining a stable voltage output. They are greatly affected by temperature and load fluctuations. Thermal management systems, including heat sinks, fans, and heat-conducting materials, are used to control the temperature rise of various power supply components, preventing performance degradation or malfunctions caused by overheating. The performance of thermal management components directly affects the reliability of the power supply under high loads or extreme operating environments. Protection circuits: such as overcurrent protection, overvoltage protection, and temperature monitoring circuits. These circuits are responsible for monitoring the power supply's operating status in real time to prevent component damage due to overload. Power supply control and regulation circuits: including control modules for controlling voltage output, current regulation, and load sensing. The stability of these circuits determines the accuracy and response speed of the power supply output.

[0030] By identifying the operating characteristics of these key components under different voltage ranges and load conditions, the performance fluctuations of the power supply under various operating states can be accurately determined, thereby providing necessary data support for anomaly detection and fault early warning.

[0031] Step S4: Critical State Analysis and Anomaly Determination By analyzing the obtained mathematical model, we can identify potential performance degradation points of the power supply near the critical voltage range and establish anomaly judgment thresholds. By combining the maximum electrical and thermal stress of key components, we can set the tolerance range for performance fluctuations and mark fault risk points (voltage points where unstable output or faults may occur, i.e., voltage points where the risk probability exceeds the preset requirements). Specifically, the process of step S4 includes the following operations: First, performance fluctuation tolerance ranges and anomaly detection thresholds are set for each voltage range. Based on the maximum rated electrical stress and temperature rise limits of power supply components, anomaly detection thresholds for parameters such as voltage, power loss, and temperature changes are determined. When the operating parameters of the power supply exceed the set tolerance range in a certain voltage range, a fault warning is triggered, and that voltage point is automatically identified as a potential fault risk point. Specifically, the setting of the anomaly detection thresholds depends on the maximum rated operating parameters of each power supply component and the power supply design tolerances. The warning signal can be promptly notified to operators via interface or communication methods, allowing them to take appropriate measures to prevent equipment damage. By integrating historical data and regression model prediction results, risk assessment is optimized, fault prediction accuracy is improved, and a basis is provided for subsequent power supply maintenance and optimization.

[0032] Step S5: Performance Reporting and Feedback Mechanism Based on the test results across various voltage ranges, a performance evaluation report is generated. This report should include information such as the power supply's performance, dynamic response characteristics, power loss, temperature fluctuations, and potential fault risks across each voltage range, serving as a comprehensive basis for power supply performance verification. A report feedback mechanism enables real-time monitoring and optimization of subsequent power supply design and testing processes.

[0033] Furthermore, in step S1, the voltage sub-intervals are divided in a non-equal manner, specifically: based on the power supply efficiency-voltage characteristic curve, the voltage segment boundary where the rate of change of efficiency with voltage exceeds a preset first threshold is taken as the sub-interval division point. The first threshold is set to a specific numerical range. Through analysis of test data from multiple typical DC power supplies, the first threshold for the rate of change of efficiency with voltage is preferably set between 2% / V and 5% / V; for example, it can be set to 3% / V. Those skilled in the art can select or fine-tune within this range based on the typical performance of the power supply to achieve the technical effects of this invention.

[0034] The purpose of this setup is to address the problem that traditional equal-division voltage range testing methods cannot accurately capture the actual performance variation range of a power supply. Because the internal topology of a DC power supply (such as the Buck / Boost circuit of a switching power supply) exhibits significant nonlinearity in its operating modes and efficiency characteristics at different voltage points, the traditional equal-division method wastes testing resources in areas of smooth efficiency change, while failing to sample sufficiently near critical voltage points where efficiency changes drastically. This solution, by analyzing the rate of change of the efficiency-voltage curve in real time, precisely concentrates testing resources on the voltage range boundaries corresponding to efficiency inflection points (such as circuit mode switching points or component stress abrupt changes). This allows for comprehensive coverage of all potential critical performance regions of the power supply with the fewest possible test sub-intervals, achieving accurate location and evaluation of nonlinear performance degradation points over a wide voltage range.

[0035] Furthermore, the efficiency-voltage characteristic curve of the power supply is obtained by performing a pre-scan test with a constant current load and a fixed sampling interval within the rated output range and calculating the efficiency value in real time.

[0036] Furthermore, in step S2, the dynamic voltage regulation adopts a predefined voltage-time waveform sequence, which includes at least one voltage step change and a voltage ramp change in each voltage sub-interval. In one possible embodiment, the step amplitude is set to 10% to 20% of the voltage span of the sub-interval, and the ramp change rate is set to 0.5 to 2V / ms.

[0037] The purpose of this design is to address the technical problem that traditional steady-state testing or single-rate scanning cannot effectively elicit and capture dynamic performance defects in power supplies. When a DC power supply operates over a wide voltage range, the response of its feedback control loop, power device switching characteristics, and output filter network exhibits significant nonlinearity with changes in the operating point. Slow ramp testing alone masks the stability issues of the loop during voltage abrupt changes, while step testing alone misses the gradual performance changes during continuous voltage variations. This solution combines step changes of a specific amplitude (for testing load transient response and loop stability margin) and ramp changes of a specific rate (for evaluating linearity and thermal accumulation effects during continuous voltage regulation) within each critical voltage sub-range. By actively constructing dynamic stresses close to actual operating conditions using a preset waveform sequence, it systematically exposes hidden fault modes such as oscillations, overshoot, and slow regulation during voltage switching, achieving a complete evaluation of the power supply's dynamic performance across the entire voltage range.

[0038] The numerical range is set to effectively cover typical operating conditions for DC power supply dynamic performance testing while avoiding irreversible damage to the equipment. The step amplitude (10%–20% of the sub-interval voltage span) generates a sufficiently large dynamic disturbance to trigger transient responses (such as overshoot and oscillation) in the power supply control loop. This amplitude is both less than the typical operating threshold of the power supply protection circuit (typically >30%) and significantly greater than the system's inherent noise (typically <2%), ensuring the acquisition of effective dynamic characteristic data. The ramp rate of change (0.5–2V / ms) is set to match the settling time constant (0.1–1ms) corresponding to the bandwidth of the industrial-grade switching power supply control loop (typically 1–10kHz). This rate range exposes stability defects in the loop during quasi-static regulation without causing the common digital power supply to enter a fault protection state due to excessively rapid changes.

[0039] Furthermore, in step S3, the nonlinear regression analysis employs the support vector regression (SVR) algorithm, using voltage, output current, and load change rate as input features, and instantaneous power supply efficiency and temperature rise of key components as output targets to establish a predictive model for power supply performance. The predictive model is then used to perform boundary search to obtain critical performance boundary points where efficiency drops by more than 5% or temperature rise exceeds the rated value by more than 15%. Boundary search refers to increasing the output current value by a preset step size under a given voltage value, and calculating the corresponding predicted efficiency and temperature rise values ​​in real time. The (voltage, current) coordinate point corresponding to the first time the instantaneous efficiency predicted value falls below the rated value by 5% or the first time the temperature rise predicted value exceeds the rated value by 15% is marked as the critical performance boundary point.

[0040] The purpose of this design is to address the technical problem that traditional linear models or empirical threshold methods cannot accurately predict the nonlinear performance boundaries of power supplies under complex operating conditions. Because the efficiency changes and temperature rise characteristics of DC power supplies are influenced by multiple factors such as voltage, current, and the rate of change of dynamic load, they exhibit high nonlinearity, making it difficult to establish an accurate global model using conventional methods. This solution employs the Support Vector Regression (SVR) algorithm, leveraging its advantage in handling small-sample, high-dimensional nonlinear problems. Through three key dynamic features—voltage value, output current value, and load change rate—it accurately learns the complex mapping relationship between these features and the instantaneous efficiency of the power supply and the temperature rise of key components. The critical performance boundary points obtained through boundary search based on this model (efficiency decrease > 5% or temperature rise > 15% of rated value) can more scientifically and reliably identify potential fault areas that are easily missed by traditional testing methods, thereby achieving accurate early warning of vulnerable operating points of the power supply across the entire voltage range.

[0041] The support vector regression algorithm in this scheme is applied as follows: During the model building phase, the input features are the feature vector composed of the voltage value (unit: V), output current value (unit: A), and load change rate (unit: A / s) collected in step S2; the output targets are the corresponding instantaneous power supply efficiency (unit: %) and key component temperature rise (unit: °C). The algorithm maps the input features to a high-dimensional feature space through a nonlinear kernel function (specifically, a radial basis function, with the kernel width parameter γ ranging from 0.01 to 0.1), and searches for a regression hyperplane in this space that ensures the deviation between the predicted values ​​and the true values ​​of most training samples does not exceed a preset tolerance ε (ε is set to 0.5% in this scheme). The hyperparameters necessary for model training, including the penalty coefficient C (ranging from 1 to 100) and the aforementioned ε and γ, are optimized and determined through a combination of grid search and cross-validation. Based on this description, those skilled in the art can use well-known tools such as Python's scikit-learn library or MATLAB's Statistics and Machine Learning Toolbox to call the SVR module and configure the above parameters to complete the model construction and training, thereby establishing an accurate nonlinear mapping relationship from dynamic electrical operating conditions to the thermal and electrical performance of the power supply.

[0042] Furthermore, the prediction model in this scheme refers to the mathematical function trained by the aforementioned SVR algorithm, which characterizes the complex relationship between power supply performance and operating conditions. The model's input is any given combination of voltage, current, and load change rate, and its output is the predicted instantaneous efficiency and temperature rise of key components. When applying this model to perform boundary search for critical performance boundary points, the specific steps are as follows: Within a specified voltage range, keeping the voltage constant, the model predicts the instantaneous efficiency and temperature rise under different output currents (current values ​​increasing in small steps). When the model prediction shows that, compared to the rated efficiency value at that voltage point, the instantaneous efficiency decreases by more than 5% for the first time, or the model-predicted component temperature rise exceeds 15% of the component's rated maximum allowable temperature rise for the first time, the corresponding (voltage, current) coordinate point is marked as a critical performance boundary point. This process is completed through a systematic search on the voltage-current plane, and ultimately, the critical points identified in all sub-intervals together constitute the power supply's performance boundary across the entire voltage range. Based on these clearly defined input and output definitions, model application methods, and judgment criteria, technical personnel in the relevant field can implement this solution without creative effort, achieving accurate prediction of potential power supply fault areas.

[0043] Furthermore, the support vector regression algorithm uses a radial basis function as the kernel function, with the kernel width parameter γ ranging from 0.01 to 0.1. The tolerance ε of the insensitive loss function is set to 0.5%, and the penalty coefficient C ranges from 1 to 100. The hyperparameters required for training the prediction model are optimized and determined by combining grid search and cross-validation. When using the prediction model to search for critical performance boundary points, the voltage is kept constant within each voltage sub-interval, the output current value is increased incrementally, and the model is used to predict the corresponding instantaneous efficiency and temperature rise of key components. The (voltage, current) coordinate points corresponding to the first time the instantaneous efficiency decreases by more than 5% or the first time the temperature rise exceeds 15% of the rated maximum allowable temperature rise of the component are marked as critical performance boundary points.

[0044] Furthermore, in step S4, the setting of the anomaly judgment threshold is dynamic. The specific method is as follows: based on the predicted temperature rise ΔT(V) of the component corresponding to the current output voltage V output by the prediction model, combined with the real-time junction-to-case thermal impedance Zth(jc) and temperature coefficient k obtained from the component datasheet, the dynamic current limit value I_lim(V) = P_max / (V×(1+k×ΔT(V))) that changes with the output voltage is calculated, where P_max is the maximum allowable power consumption of the component; the value of the real-time junction-to-case thermal impedance Zth(jc) corresponds to the instantaneous thermal impedance value jointly determined by the power loss P_loss(V) and temperature rise ΔT(V) of the component under the current operating state predicted by the prediction model.

[0045] The specific implementation of the joint determination is as follows: Based on the component power loss P_loss(V) and temperature rise ΔT(V) at the current operating point output by the prediction model in real time as dynamic operating condition parameters, the transient thermal resistance curve family provided in the component datasheet with power pulse time and temperature rise as variables is queried, or it is substituted into the multi-order thermal network parameterized model of the component for convolution calculation, thereby dynamically matching and determining a real-time junction-to-shell thermal resistance Zth(jc) value that can accurately reflect the relationship between instantaneous heat flow and temperature response and is adapted to the current actual operating state. This method makes the thermal resistance value no longer a fixed parameter, but a key variable that adaptively adjusts with the actual electrothermal state of the power supply, thereby significantly improving the accuracy and reliability of subsequent dynamic current limit calculation.

[0046] In one possible embodiment, a hybrid determination method is provided for determining the real-time junction-to-shell thermal impedance Zth(jc). This method executes two calculation paths based on different principles in parallel and selects the more conservative (i.e., larger) result as the final value. This simultaneously captures the fundamental thermal accumulation effect caused by long-term average power and the short-term thermal shock effect caused by instantaneous power mutation, thereby ensuring the completeness and reliability of the safety boundary determination under dynamic operating conditions. In practice, the following two calculation paths are executed in parallel: The first path is a fast lookup method based on the datasheet. Using the current power loss P_loss(V) and temperature rise ΔT(V) output by the prediction model, combined with the calculated equivalent thermal shock duration t_eff, interpolation is performed in the transient thermal resistance curve family provided by the component manufacturer to obtain a basic impedance value that reflects the steady-state and long-term thermal state. The second path is a dynamic calculation method based on the thermal network model. Using the fitted multi-order Foster model parameters, convolution operation is performed on the power loss sequence within a short time window (e.g., 50 milliseconds) to calculate a dynamic impedance value that characterizes the most recent instantaneous thermal change in real time. Finally, the basic impedance value and the dynamic impedance value are compared, and the larger value is obtained by taking the larger value.

[0047] The purpose of this design is to address the technical problem of traditional evaluation methods that use fixed thresholds, which fail to accurately reflect the true safety boundaries of a power supply under different operating voltages. Because the heat dissipation capacity and electrical stress of key power supply components (such as power MOSFETs) change non-linearly with variations in output voltage and current, a current value safe at low voltages may cause component damage due to overheating or overstress at high voltages. This solution dynamically calculates the maximum allowable current I_lim (V) at each voltage point by combining the model-predicted real-time temperature rise ΔT (V) with the component's thermal impedance Zth and temperature coefficient k. This makes the safety threshold no longer a fixed value, but a dynamic boundary that adaptively adjusts according to the actual operating state of the power supply (voltage, predicted temperature rise). This allows for a more accurate definition of the power supply's true safe operating area across the entire voltage range, effectively preventing misjudgments or omissions caused by overly conservative or excessive static thresholds.

[0048] Furthermore, in step S2, the amplitude of the voltage step change is adaptively adjusted. The specific adjustment method is as follows: based on the prediction results of the prediction model on the instantaneous efficiency and temperature rise of the current voltage point and the target step voltage point, the optimization target is to make the predicted efficiency decrease value after the step within the range of 3% to 8% or the predicted temperature rise value within the range of 10% to 20% of the rated temperature rise of the component. The final step amplitude is dynamically determined by iterative calculation.

[0049] The purpose of this setup is to address the technical problem of mismatch between test stress and actual power supply conditions in traditional fixed-amplitude step tests. A fixed step amplitude may fail to expose potential stability defects in the low-voltage operating range due to insufficient excitation, while in the high-voltage operating range, excessive stress may trigger power supply protection or cause irreversible damage, thus interrupting the test. This solution utilizes predictive models for forward-looking simulations, optimizing two physical quantities that directly reflect the dynamic stress level of the power supply: a moderate efficiency decrease (3%–8%) and a safe and controllable temperature rise (10%–20% of rated value). It dynamically calculates and applies a customized step excitation that best matches the current power supply operating point, sufficient to expose problems while ensuring safety. This allows for the most effective and accurate evaluation of the power supply's dynamic performance across the entire voltage range, while ensuring the safety of the test equipment.

[0050] Furthermore, in step S5, the performance evaluation report automatically generates a safe operating area map of the power supply across the entire voltage range. This map uses voltage as the horizontal axis and current as the vertical axis, and marks the prohibited operating area, derating operating area, and full power operating area enclosed by each fault risk point determined in step S4 with different colors or patterns. By monitoring environmental factors (such as temperature and humidity) in real time, the safe operating area map is dynamically adjusted to optimize the maximum allowable current and power output of the power supply in real time according to environmental changes.

[0051] The purpose of this design is to address the technical problem that traditional performance evaluation reports, often presented in data tables or discrete curves, lack intuitiveness and fail to allow engineers to quickly grasp the overall performance boundaries and risk distribution of the power supply. Textual and numerical reports are difficult to directly translate into specific design or usage guidance. This solution integrates and visualizes the scattered fault risk points identified in step S4 into a safe operating area map with voltage and current as coordinate axes, clearly marking prohibited areas, derating areas, and safe areas with different colors or patterns. This directly transforms complex test data into a clear engineering operation guide, enabling power supply designers and users to intuitively and holistically understand the power supply's performance limits, avoiding inadvertently entering dangerous operating areas during subsequent design or use, thereby improving the reliability and safety of power supply applications at the system level.

[0052] Example 2: To ensure the reliability of the predicted critical performance boundary points in step S3, this example adds a dynamic credibility verification step. In step S3, after obtaining the critical performance boundary points using the prediction model, dynamic credibility verification of the boundary points is also performed. Calculate the voltage stress concentration index and the thermal accumulation steep change index; input these two indices into a predefined credibility assessment model to calculate the boundary credibility coefficient; when the boundary credibility coefficient is lower than a preset second threshold, readjust the hyperparameters of the support vector regression algorithm and iterate the modeling process; the calculation process of the voltage stress concentration index is as follows: within the target voltage sub-interval, calculate the spectral entropy value of the voltage fluctuation rate, that is, perform a fast Fourier transform on the instantaneous voltage fluctuation sequence collected in the dynamic test to extract the distribution dispersion of the main frequency components; calculate the correlation coefficient between the load change rate and the voltage recovery time within this sub-interval (the linear correlation between the two is calculated using the Pearson product-moment correlation coefficient formula); multiply the spectral entropy value by the correlation coefficient, and... Normalization is performed to obtain the voltage stress concentration index; the calculation process of the thermal accumulation steep change index is as follows: extract the temperature rise curve of the key component in dynamic testing, calculate the coefficient of variation of the temperature rise rate, that is, the ratio of the standard deviation of the temperature rise rate to the mean; determine the delay characteristics of the temperature response relative to the voltage change through cross-correlation analysis, and calculate the phase lag angle of the temperature rise relative to the voltage change rate; weight the sum of the coefficient of variation and the sine value of the phase lag angle to obtain the thermal accumulation steep change index; the predefined credibility assessment model is: boundary credibility coefficient = 1 / (1+exp(-k1×voltage stress concentration index+k2×thermal accumulation steep change index)), where k1 and k2 are model calibration parameters, which are obtained through machine learning training of historical test data.

[0053] The method for determining the main frequencies is explained below: A Fast Fourier Transform (FFT) is performed on the instantaneous voltage fluctuation sequence acquired during dynamic testing to obtain its amplitude spectrum; all frequency components are sorted from largest to smallest according to their corresponding amplitudes, and the amplitudes are accumulated sequentially until the proportion of the accumulated amplitude to the sum of the total amplitudes reaches a preset threshold (e.g., 85%); at this point, all frequency components involved in the accumulation are defined as the main frequencies of the voltage fluctuation sequence. The distribution dispersion here is a description of the process of calculating the spectral entropy value. The two are related as the process and result of the same calculation step. After determining the main frequency set and its normalized amplitude, the distribution dispersion is calculated using the information entropy formula. The result of this calculation is the spectral entropy value.

[0054] The calculation method for obtaining the phase lag angle is as follows: Data preparation: Obtain the voltage change rate sequence and the temperature rise sequence of key components simultaneously sampled within the same time period; Calculate the cross-correlation function: Perform a cross-correlation operation on two sequences; the cross-correlation function describes the degree of similarity between two signals when there is a relative offset τ on the time axis. Determine the delay time: Find the time shift τmax corresponding to the global maximum value of the cross-correlation function; τmax is interpreted as the average time delay of the temperature rise response lagging behind the voltage change excitation, which is the delay characteristic determined by cross-correlation analysis.

[0055] Calculate the phase lag angle: Based on the obtained delay characteristic τmax, it needs to be converted into phase angle form for exponential synthesis; the calculation formula is: phase lag angle φ=360°*(τmax / T); where T is the characteristic period selected during analysis that can represent the dominant dynamic process; for example, the period of the periodic disturbance signal used in dynamic testing can be used, or the main period obtained by autocorrelation analysis of the entire analysis sequence can be used; the phase lag angle intuitively reflects the inertial delay of the heat transfer process.

[0056] In one possible embodiment, the second preset threshold is dynamically determined based on the average confidence level of the prediction model on the training dataset. For example, the second threshold is 0.7 of the average boundary confidence coefficient. When the boundary confidence coefficient is lower than this threshold, the hyperparameters of the support vector regression algorithm are readjusted according to the following rules: when the voltage stress concentration index is higher than the thermal accumulation steep change index, the kernel width parameter γ is preferentially reduced to 0.01-0.05, and the penalty coefficient C is increased to 50-100; when the thermal accumulation steep change index is higher than the voltage stress concentration index, the tolerance ε of the insensitive loss function is preferentially increased to 0.8%-1.2%, and the penalty coefficient C is decreased to 1. -10 .

[0057] Example 2, see Figure 2 A structural block diagram of a DC power supply performance evaluation system is provided in this embodiment of the invention. The system includes: Range configuration module: Based on the rated output range of the target DC power supply, the rated output range is divided into multiple voltage sub-ranges, and corresponding power output indicators are set for each sub-range to ensure coverage of key operating states under different voltage ranges; Dynamic testing module: Within each voltage sub-range, dynamic voltage regulation is implemented to gradually adjust the power supply output voltage to the set range, and the dynamic response data of the power supply output is recorded in real time, including the current, voltage, temperature, instantaneous voltage fluctuations during the operation of the power supply, as well as the impact of load changes on output stability. Intelligent modeling module: Based on the collected dynamic response data, it uses nonlinear regression analysis to model the power supply's operating efficiency, power loss, and temperature changes of key components in different voltage ranges; by establishing mathematical models, it calculates the critical performance boundaries near each voltage point and predicts potential failure risks based on historical data and trends. Risk Analysis Module: By analyzing the obtained mathematical model, it identifies the performance degradation points of the power supply near the critical voltage range and establishes anomaly judgment thresholds; combined with the maximum electrical stress and thermal stress of key components, it sets the tolerance range for performance fluctuations and marks the fault risk points. Report generation module: Combines test results from various voltage ranges to generate a performance evaluation report. The performance evaluation report includes the power supply's performance, dynamic response characteristics, power loss, temperature fluctuations, and fault risk information in each voltage range, serving as a global basis for power supply performance verification.

[0058] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for evaluating the performance of a DC power supply, characterized in that, include: Step S1: Based on the rated output range of the target DC power supply, divide the rated output range into multiple voltage sub-ranges and set corresponding power output indicators for each sub-range to ensure coverage of key operating states under different voltage sub-ranges. Step S2: Within each voltage sub-range, implement dynamic voltage regulation, gradually adjust the power supply output voltage to the set range, and record the power supply output dynamic response data in real time, including the current, voltage, temperature, instantaneous voltage fluctuations during the power supply operation process, as well as the impact of load changes on output stability. Step S3: Based on the collected dynamic response data, nonlinear regression analysis is used to model the power supply's operating efficiency, power loss, and temperature changes of key components in different voltage sub-ranges. By establishing a predictive model, the critical performance boundary near each voltage point is calculated, and potential failure risks are predicted based on historical data and trends. Step S4: By analyzing the obtained prediction model, identify the performance degradation points of the power supply near the critical voltage sub-range and establish anomaly judgment thresholds; combine the maximum electrical stress and thermal stress of key components to set the tolerance range of performance fluctuations and mark the fault risk point information. Step S5: Generate a performance evaluation report by combining the test results of each voltage sub-range. The performance evaluation report includes the power supply’s performance, dynamic response characteristics, power loss, temperature fluctuation and fault risk information in each voltage sub-range, serving as the global basis for power supply performance verification.

2. The performance evaluation method for a DC power supply according to claim 1, characterized in that, In step S1, the voltage sub-intervals are divided in a non-equal manner, specifically: based on the power supply efficiency-voltage characteristic curve, the boundary of the voltage segment in the curve where the rate of change of efficiency with voltage exceeds a preset first threshold is taken as the division point of the sub-intervals.

3. The performance evaluation method for a DC power supply according to claim 2, characterized in that, The efficiency-voltage characteristic curve of the power supply is obtained by performing a pre-scan test with a constant current load and a fixed sampling interval within the rated output range and calculating the efficiency value in real time.

4. The performance evaluation method for a DC power supply according to claim 3, characterized in that, In step S2, the amplitude of the voltage step change is adaptively adjusted; the specific adjustment method is as follows: based on the prediction results of the prediction model for the instantaneous efficiency and temperature rise of the current voltage point and the target step voltage point, the predicted efficiency decrease value after the step is at the optimization target, and the final step amplitude is dynamically determined through iterative calculation.

5. The performance evaluation method for a DC power supply according to claim 1, characterized in that, In step S2, the dynamic voltage regulation adopts a predefined voltage-time waveform sequence, which includes at least one voltage step change and a voltage ramp change in each voltage sub-interval; the step amplitude is set to 10% to 20% of the voltage span of the sub-interval, and the ramp change rate is set to 0.5 to 2V / ms.

6. The performance evaluation method for a DC power supply according to claim 5, characterized in that, In step S3, the nonlinear regression analysis uses the support vector regression algorithm, with voltage value, output current value and load change rate as input features, and instantaneous power supply efficiency and temperature rise of key components as output targets, to establish a predictive model of power supply performance, and to use the predictive model to perform boundary search to obtain critical performance boundary points.

7. The performance evaluation method for a DC power supply according to claim 6, characterized in that, The support vector regression algorithm uses a radial basis function as the kernel function, with the kernel width parameter γ ranging from 0.01 to 0.

1. The tolerance ε of the insensitive loss function is set to 0.5%, and the penalty coefficient C ranges from 1 to 100. The hyperparameters required for training the prediction model are optimized and determined through a combination of grid search and cross-validation. When using the prediction model to search for critical performance boundary points, the voltage is kept constant within each voltage sub-interval, the output current value is increased incrementally, and the model predicts the corresponding instantaneous efficiency and temperature rise of key components. The coordinate point corresponding to the first instance of instantaneous efficiency decrease exceeding 5%, or the first instance of key component temperature rise exceeding 15% of the component's rated maximum allowable temperature rise, is marked as the critical performance boundary point.

8. The performance evaluation method for a DC power supply according to claim 1, characterized in that, In step S4, the setting of the anomaly judgment threshold is dynamic. The specific method is as follows: based on the predicted temperature rise ΔT(V) of the component corresponding to the current output voltage V output by the prediction model, combined with the real-time junction-to-case thermal impedance Zth(jc) and temperature coefficient k obtained from the component datasheet, the dynamic current limit value I_lim(V) = P_max / (V×(1+k×ΔT(V))) that changes with the output voltage is calculated, where P_max is the maximum allowable power consumption of the component; the value of the real-time junction-to-case thermal impedance Zth(jc) corresponds to the instantaneous thermal impedance value jointly determined by the power loss P_loss(V) and temperature rise ΔT(V) of the component under the current operating state predicted by the prediction model.

9. The performance evaluation method for a DC power supply according to claim 6, characterized in that, In step S3, after obtaining the critical performance boundary points by using the prediction model to perform boundary search, the dynamic credibility verification of the boundary points is also performed: the voltage stress concentration index and the thermal accumulation steep change index are calculated; these two indices are input into the predefined credibility evaluation model to calculate the boundary credibility coefficient; when the boundary credibility coefficient is lower than the preset second threshold, the hyperparameters of the support vector regression algorithm are readjusted and the modeling process is iterated. The calculation process of the voltage stress concentration index is as follows: within the target voltage sub-interval, calculate the spectral entropy value of the voltage fluctuation rate, that is, perform a fast Fourier transform on the instantaneous voltage fluctuation sequence collected in the dynamic test to extract the distribution dispersion of the main frequency components; calculate the correlation coefficient between the load change rate and the voltage recovery time within the sub-interval; multiply the spectral entropy value and the correlation coefficient, and perform normalization to obtain the voltage stress concentration index; The calculation process of the thermal accumulation steep change index is as follows: extract the temperature rise curve of the key component in dynamic testing, calculate the coefficient of variation of the temperature rise rate, that is, the ratio of the standard deviation of the temperature rise rate to the mean; determine the delay characteristics of the temperature response relative to the voltage change through cross-correlation analysis, and calculate the phase lag angle of the temperature rise relative to the voltage change rate; and obtain the thermal accumulation steep change index by weighted summation of the coefficient of variation and the sine value of the phase lag angle. The predefined credibility assessment model is: Boundary credibility coefficient = 1 / (1+exp(-k1×voltage stress concentration index+k2×thermal accumulation steep change index)), where k1 and k2 are model calibration parameters, which are obtained through machine learning training of historical test data.

10. A performance evaluation system for a DC power supply, used to implement the method described in any one of claims 1-9, characterized in that, include: Range configuration module: Based on the rated output range of the target DC power supply, the rated output range is divided into multiple voltage sub-ranges, and corresponding power output indicators are set for each sub-range to ensure coverage of key operating states under different voltage sub-ranges; Dynamic testing module: Within each voltage sub-range, dynamic voltage regulation is implemented to gradually adjust the power supply output voltage to the set range, and the dynamic response data of the power supply output is recorded in real time, including the current, voltage, temperature, instantaneous voltage fluctuations during the operation of the power supply, as well as the impact of load changes on output stability. Intelligent modeling module: Based on the collected dynamic response data, nonlinear regression analysis is used to model the power supply's operating efficiency, power loss, and temperature changes of key components in different voltage sub-ranges; By establishing a predictive model, the critical performance boundary near each voltage point is calculated, and potential failure risks are predicted based on historical data and trends. Risk Analysis Module: By analyzing the obtained prediction model, it identifies the performance degradation points of the power supply near the critical voltage sub-range and establishes anomaly judgment thresholds; combined with the maximum electrical stress and thermal stress of key components, it sets the tolerance range for performance fluctuations and marks the fault risk point information; Report generation module: Combines the test results of each voltage sub-range to generate a performance evaluation report; The performance evaluation report includes the power supply's performance, dynamic response characteristics, power loss, temperature fluctuation and fault risk information in each voltage sub-range, serving as a global basis for power supply performance verification.

Images