New energy automobile motor feed stability analysis system and test method

Abstract

The invention discloses a new energy automobile motor feed stability analysis system and a test method, particularly relates to the field of new energy automobile motor test, and is used for solving the problem of deviation between an evaluation result and real automobile performance caused by neglecting transient fluctuation of a bus in a dynamic process in an existing rack test. The method comprises the following steps: forming an original data set, executing transient positioning and steady-state segmentation on the original data set, constructing a bus fluctuation fragment set and a steady-state fragment index, evaluating a table relation among a ripple amplitude phase trend, duration and torque response on a bus fluctuation fragment, marking risks, generating a ripple-torque association table, and constructing a key transient script and direct-current source dynamic configuration. The method comprises the following steps of: reproducing vehicle feedback and load abrupt change on a rack, collecting a reproduction data set and evaluating a difference degree, executing single revision when the difference degree is in a risk interval, selecting a sampling bandwidth, an event marking time sequence or a direct current source dynamic configuration path, and iteratively verifying and outputting a stability judgment conclusion and a test configuration list.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle motor testing, and more specifically, to a new energy vehicle motor power supply stability analysis system and testing method. Background Technology

[0002] In the operation of new energy vehicles, the motor, as the core power unit, requires energy interaction between the motor controller and the DC bus to maintain the continuity and stability of drive. To verify the performance of the motor and controller during R&D and production, the industry typically uses bench testing to simulate the vehicle's operation on the road. Bench testing can reproduce actual driving conditions to a certain extent, helping researchers observe the performance of the motor and controller under different conditions through processes such as acceleration, deceleration, and energy feedback. This method provides a foundation for performance evaluation, but its focus is mostly on reproducing power and torque curves, often lacking sufficient monitoring and analysis of the subtle responses of the bus during dynamic processes. Especially during regenerative braking or operating condition switching phases, the transient characteristics of bus energy transfer are easily overlooked, leading to discrepancies between test results and actual applications.

[0003] Against this backdrop, existing testing methods struggle to accurately reveal the true state of motor power supply stability. Fluctuations in current and voltage occur in the bus during energy feedback or sudden load changes. While these fluctuations are brief, they directly impact the stable operation of the motor. Failure to capture and analyze these details during testing leads to overly idealized results that fail to fully reflect the vehicle's actual performance under complex road conditions. When vehicles enter congested environments, operate on inclines, or frequently switch between operating conditions over extended periods, these fluctuations can cause motor instability, abnormal controller adjustments, and even torque response delays, ultimately affecting driving safety and comfort. Therefore, the lack of systematic measurement and analysis of the relationship between bus fluctuations and power supply stability is a prominent problem in current testing methods for new energy vehicle motors. This is precisely the key technical challenge that new energy vehicle motor power supply stability analysis systems and testing methods aim to address.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies in the prior art, embodiments of the present invention provide a new energy vehicle motor power supply stability analysis system and testing method. This system synchronously collects DC bus voltage, DC bus current, controller operating status, and torque response on a unified time axis, and adds event markers for regeneration, load switching, and start-stop transitions to form an original dataset. The original dataset is then subjected to transient localization and steady-state segmentation to create a set of bus fluctuation segments and a steady-state segment index. The correlation between ripple current amplitude and phase direction, duration, and torque response is evaluated on the bus fluctuation segments to label risks and generate a ripple current-torque correlation table. Based on the ripple current-torque correlation table, key transient scripts and dynamic DC source configurations are constructed. The system simulates real vehicle feedback and load mutation processes on a test bench to collect and reproduce datasets and evaluate the difference between the reproduced datasets and the original datasets. When the difference is within the risk range, a single revision is performed on the sampling bandwidth, event marker timing, or DC source dynamic configuration. This process is repeated until the difference leaves the risk range, ultimately outputting a stability judgment conclusion and a matching test configuration list to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for testing the power supply stability of a new energy vehicle motor, comprising the following steps:

[0008] S1: Collect DC bus voltage and current, controller operating status and torque response on a unified time axis, write event tags for regeneration and load switching and start-stop transitions to obtain the raw dataset;

[0009] S2: Perform transient localization and steady-state segmentation on the original dataset to form a set of bus wave segments and a steady-state segment index, and clarify the segment boundaries and segment types;

[0010] S3: On the busbar fluctuation segment, complete the comparison evaluation of the ripple current amplitude and phase direction and duration with the synchronous torque response, which is used for segment risk labeling and written into the ripple current-torque correlation table;

[0011] S4: Based on the ripple-torque correlation table and segment risk labels, construct key transient scripts and dynamic configuration of DC sources. Reproduce the vehicle feedback and load change process on the test bench, collect the reproduction dataset and give the evaluation results of the difference between the reproduction dataset and the original dataset.

[0012] S5: When the difference is in the risk range, perform a single revision, select the revision path of sampling bandwidth, event marking timing or DC source dynamic configuration, complete the segmentation, feature extraction and reproduction evaluation again, and output the stability judgment conclusion and the matching test configuration list.

[0013] In a preferred embodiment, a single time axis is generated using a time base, and DC bus voltage, DC bus current, controller operating status and torque response are synchronously acquired. Cross-channel delay errors are eliminated by combining trigger alignment and timestamp calibration.

[0014] In a preferred embodiment, within a single time axis, event markers for regeneration, load switching, and stop-start transitions are located using a joint criterion of controller operating status and power commutation indications. Regeneration is confirmed by power direction reversal and drive command intervals, load switching is confirmed by the simultaneous fulfillment of load-side operating condition semantics and current transitions, and stop-start transitions are confirmed by the joint confirmation of controller stack entry / exit status and bus charging / discharging behavior. This ultimately forms an original dataset with event marker fields, and records the effective bandwidth of the channel and the sampling window.

[0015] In a preferred embodiment, transient localization and steady-state segmentation are performed on the original dataset. A dual-domain search is constructed around the event markers. In the time domain, transient localization is performed with the event neighborhood as the center, and in the semantic domain, steady-state confirmation is performed with the controller's running state changes. The fragment boundaries are given after the results of the two domains are cross-verified.

[0016] In a preferred embodiment, the segment types within the segment boundary are divided according to the bus energy flow direction and torque response pattern, including feedback inrush segment, feedback continuous segment, feedback exit segment, load increase segment, load decrease segment, and stop-start transition segment. For each segment, the source event, start and end timestamps, energy direction label, and torque response label are recorded to form a set of bus fluctuation segments. A steady-state segment index is established based on the segment start and end time, segment type, and source event. The index is mapped to the original dataset using key-value pairs.

[0017] In a preferred embodiment, within the set of bus fluctuation segments, the amplitude and phase trajectory of the ripple current is constructed for the DC bus voltage and DC bus current. The trajectory is jointly described by the zero-crossing sequence and the sign dwell sequence. The duration is extracted within the same segment and a correspondence is established with the synchronous torque response, so that the amplitude and phase trend of the electrical side ripple current corresponds to the rise and convergence of the mechanical side torque.

[0018] In a preferred embodiment, the main transition point of ripple current is identified within the interval from feedback triggering to torque zero crossing, and key inflection points are identified during torque rise or fall. A one-to-one mapping is established according to time sequence, and the robust representative of the mapping time difference is taken as the feedback-following offset. Within the feedback window, a closed write-back trajectory is generated in time sequence with DC bus voltage as the horizontal axis and DC bus current as the vertical axis. The write-back energy is obtained by accumulating the product of power and time along the trajectory direction. Then, the energy loop coupling area ratio is obtained by normalizing the total energy in the same window. The feedback-following offset and the energy loop coupling area ratio are fed into the sequence consistency scoring method. Using the time sequence template of regeneration triggering-write-back peaking-torque convergence as a reference, the sequential matching degree and mutual corroboration degree of parameters on the template are calculated to obtain the single-value stability coefficient. Based on this, the segment risk labeling is completed, and finally, a ripple current-torque correlation table with segment keys, parameter pairs, stability coefficients and risk labels is generated.

[0019] In a preferred embodiment, high-risk and typical segments are screened based on the ripple current-torque correlation table and compiled into a key transient script. The script structure includes trigger sequence, energy direction and target phase direction semantics. Combined with the script, a dynamic configuration of the DC source is loaded at the DC port. The configuration includes output impedance semantics, voltage drop recovery semantics and current limiting switching logic semantics. After the script is executed, a reproducible dataset is collected and aligned one-to-one with event keys and segment keys. The difference is evaluated around the feedback phase offset, energy loop coupling area ratio and stability coefficient. The difference is synthesized by three semantics: timing alignment deviation, energy coupling deviation and coefficient deviation. The overall difference conclusion is given and the sub-item difference mapping is retained.

[0020] In a preferred embodiment, when the difference degree enters the risk range, a single revision path is triggered based on the sub-item difference mapping. If the main deviation is manifested as timing alignment mismatch, event marking timing revision is performed. The revision strategy is limited to triggering edge positioning and cross-channel delay compensation. If the main deviation is manifested as high-frequency fluctuation distortion, sampling bandwidth revision is performed. The revision strategy is limited to channel anti-aliasing settings and effective bandwidth enhancement. If the main deviation is manifested as port write-back topology difference, DC source dynamic configuration revision is performed. The revision strategy is limited to single-domain fine-tuning of output impedance semantics, recovery semantics, and current limiting semantics. After the revision is completed, the process of segmentation, feature calculation, script execution, and difference evaluation is performed again along the path of steps S2-S4 until the difference degree exits the risk range. The stability judgment conclusion and the matching test configuration list are output, and the final configuration is associated with the fragment key and sealed.

[0021] A new energy vehicle motor power supply stability analysis system includes:

[0022] Data acquisition module: Collects DC bus voltage and current, controller operating status and torque response on a unified time axis, writes event tags for regeneration and load switching and start-stop transitions, and obtains raw dataset;

[0023] Transient segmentation module: Performs transient localization and steady-state segmentation on the original dataset to form a set of bus wave segments and a steady-state segment index, and clarifies segment boundaries and segment types;

[0024] Risk assessment module: Completes the comparison assessment of the ripple current amplitude and phase direction and duration with the torque response of the same period on the bus fluctuation segment set, which is used for segment risk labeling and writing into the ripple current-torque correlation table;

[0025] Script Reproduction Module: Based on the ripple-torque correlation table and segment risk labels, key transient scripts and dynamic configuration of DC sources are constructed. The actual vehicle feedback and load change process are reproduced on the test bench. The reproduction dataset is collected and the difference evaluation results between the reproduction dataset and the original dataset are given.

[0026] Revision and optimization module: When the difference is in the risk range, a single revision is performed. The revision path is selected by sampling bandwidth, event marking timing or DC source dynamic configuration. The segmentation, feature extraction and reproduction evaluation are completed again, and the stability judgment conclusion and the matching test configuration list are output.

[0027] The technical effects and advantages of the present invention, a system and testing method for analyzing the power supply stability of a new energy vehicle motor, are as follows:

[0028] The processing chain of this invention incorporates bus voltage and current, controller operating status, and torque response into the same time axis and marks them with event tags, forming a traceable data source. After transient positioning and steady-state segmentation, the bus fluctuations generated by regeneration and load switching are extracted into clear segments to avoid judgment bias caused by mixed operating conditions. Within each segment, the ripple current amplitude and duration are extracted synchronously and compared with the torque response at the same time to form a correlation table that directly reflects the stability of the power supply, realizing the transformation from "looking at the curve shape" to "looking at the correlation basis". Based on this, a key transient script and DC source dynamic configuration are constructed to reproduce the feedback and load change behavior of the actual vehicle on the test bench, and collect the ripple current and torque response relationship consistent with the whole vehicle, reducing the gap between test results and actual vehicle performance. Finally, the difference evaluation triggers a single revision path to revise and retest the sampling bandwidth, event timing, or DC source dynamic configuration, and outputs the stability judgment conclusion and the corresponding test configuration list. The resulting effects are as follows: the observability of regenerative transients is realized, the influence path of bus ripple flow on torque fluctuations is quantified, the consistency between the test bench and the whole vehicle is significantly enhanced, the stability threshold judgment has reproducible and comparable attributes, the debugging convergence process reduces invalid parameter tuning, and the output results can directly serve mass production release and platform benchmarking. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a method for testing the power supply stability of a new energy vehicle motor according to the present invention.

[0030] Figure 2 This is a schematic diagram of the structure of a new energy vehicle motor power supply stability analysis system according to the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Example 1: Figure 1 This invention provides a method for testing the power supply stability of a new energy vehicle motor, comprising:

[0033] S1: Collect DC bus voltage and current, controller operating status and torque response on a unified time axis, write event tags for regeneration and load switching and start-stop transitions to obtain the raw dataset.

[0034] S2: Perform transient localization and steady-state segmentation on the original dataset to form a set of bus fluctuation segments and a steady-state segment index, and clarify the segment boundaries and segment types.

[0035] S3: On the bus ripple fluctuation segment, complete the comparison evaluation of the ripple current amplitude and phase direction and duration with the synchronous torque response, which is used for segment risk labeling and written into the ripple current-torque correlation table.

[0036] S4: Based on the ripple-torque correlation table and segment risk labels, construct key transient scripts and dynamic configuration of DC sources. Reproduce the vehicle feedback and load change process on the test bench, collect the reproduction dataset, and give the difference evaluation results between the reproduction dataset and the original dataset.

[0037] S5: When the difference is in the risk range, perform a single revision, select the revision path of sampling bandwidth, event marking timing or DC source dynamic configuration, complete the segmentation, feature extraction and reproduction evaluation again, and output the stability judgment conclusion and the matching test configuration list.

[0038] The stability testing of electric motor power supply in new energy vehicles is crucial in research and development and production. As the core power unit, the energy interaction between the motor and the DC bus directly determines the continuity and reliability of the drive system. While traditional bench tests can simulate road conditions, they often neglect transient fluctuations in the bus during dynamic processes, leading to discrepancies between test results and actual vehicle performance. This discrepancy is particularly evident during regenerative braking, load switching, or start-stop transitions, potentially causing motor instability or controller malfunctions, affecting driving safety and comfort. To address this issue, this invention proposes a systematic approach, starting with data acquisition to ensure the synchronous recording of bus voltage, current, controller operating status, and torque response, and embedding event markers to construct a traceable foundational dataset, supporting the accuracy and consistency of subsequent analysis.

[0039] In the testing of new energy vehicle motors, accurately capturing the dynamic response of the bus is a key foundation for revealing the stability of the power supply. However, existing methods often lead to analytical biases due to asynchronous data acquisition and missing event markers. To address this, step S1 focuses on establishing multi-channel synchronous acquisition under a unified time axis and accurately locating key event markers, thereby generating a raw dataset with event marker domains, which serves as a reliable input for subsequent transient localization and steady-state segmentation.

[0040] The specific processing logic of step S1:

[0041] A single timeline is generated: using a high-precision time base as a reference, a single timeline is produced using an atomic clock or a GPS-synchronized clock source to ensure microsecond-level time resolution. A time base frequency, the reciprocal of the clock source's oscillation period, is selected to define the timeline's scale intervals. A uniform time series is generated through a frequency divider circuit, where each point in the time series is obtained by dividing an integer count by the time base frequency. This single timeline serves as a common reference for all channels, avoiding asynchronous issues caused by independent clock drift. The single timeline provides a unified timing framework, supporting subsequent data alignment.

[0042] Synchronous data acquisition is implemented for the following: DC bus voltage (the voltage value on the DC line connecting the motor controller and the power supply), DC bus current (the current value on that line), and controller operating status (the controller's internal mode indication, such as drive or regenerative mode) and torque response (the change in the mechanical torque output of the motor). A multi-channel analog-to-digital converter is used to simultaneously trigger acquisition at each sampling point on a single time axis, ensuring a consistent sampling rate (the number of samples acquired per second) across all channels, satisfying the sampling theorem to capture transient signals. Sensors are connected via hardware synchronization lines to achieve parallel acquisition, forming raw data sequences, including DC bus voltage, DC bus current, controller operating status, and torque response sequences. This ensures precise temporal correspondence of the data, laying the foundation for event judgment.

[0043] Eliminating cross-channel delay errors: A combination of trigger alignment and timestamp calibration is used to eliminate cross-channel delay errors. First, the start acquisition times of each channel are aligned using an external trigger signal, and the delay difference is calculated by subtracting the trigger time of a specific channel from the trigger time of the reference channel. Second, a timestamp is appended to each data point, and the timestamp offset is calibrated using the least squares method. This calibration method obtains the calibrated time by adding an offset coefficient to the observation time and multiplying it by the difference between the observation time and the initial reference point. The offset coefficient is obtained by fitting a known synchronization point. This calibration ensures linear correction of the timeline. Through this process, cross-channel delay errors are controlled at the nanosecond level, enhancing data consistency.

[0044] Location event markers: Within a single time axis, the event markers for regeneration, load switching, and stop-start transitions are located by combining the controller's operating status with power commutation indicators. Specific implementation: For regenerative events, the power direction reversal (the power value sign changes from positive to negative, where power is obtained by multiplying the DC bus voltage by the DC bus current) and the drive command interval (the time period during which the controller's operating state switches from drive mode to regenerative mode) are used for confirmation. The joint criterion is that the power value is less than zero and the controller's operating state enters the regenerative interval. For load switching events, the load-side operating condition semantics (based on semantic tags of torque response changes, such as acceleration or deceleration) and current transition (where the DC bus current amplitude exceeds a threshold, and confirmation is met) are used. The threshold can be determined through historical data statistics, such as calculating the percentile of the current sequence, such as using the 90th percentile as the transition threshold. For stop-start transition events, the controller's stack state (the stack push / pull indication of the controller's operating state) and bus charging / discharging behavior (the charging / discharging mode of the DC bus voltage and DC bus current) are used for confirmation. The event is marked when the controller's operating state shows a push onto the stack and the power value is zero. These event markers are written to the data sequence to form an event marker field. The located event markers provide clear time anchors, facilitating subsequent segmentation.

[0045] Step S1 synchronously collects DC bus voltage, DC bus current, controller operating status and torque response under a unified time axis to eliminate cross-channel delay errors. It locates event markers for regeneration, load switching and stop-start transitions through joint criteria, forms an original dataset with event marker fields, and records the effective bandwidth and sampling window of the channel.

[0046] Step S1 constructs an original dataset with event-marked fields by generating a single time axis, synchronously acquiring data, eliminating delay errors, and locating event markers. It also records the effective bandwidth of the channels and the sampling window to ensure high data fidelity and traceability, providing a solid data foundation for revealing the power supply stability of new energy vehicle motors under regeneration and load switching.

[0047] Step S1 has generated the raw dataset with event-marked fields, providing a synchronous and traceable multi-channel data foundation for subsequent processing. However, to reveal the specific impact of bus fluctuations on regeneration and load switching, these data must be further finely segmented. To this end, step S2 performs a dual-domain search around the event markers to achieve transient localization and steady-state segmentation, forming a set of bus fluctuation segments and a steady-state segment index, which serve as direct inputs for feature extraction and risk assessment.

[0048] The specific processing logic of step S2:

[0049] Step S2 is divided into five sub-processing steps: constructing a dual-domain search, defining fragment boundaries, classifying fragment types, forming a set of bus fluctuation fragments, and establishing a steady-state fragment index. The technical logic of each step is as follows.

[0050] A dual-domain search is constructed around event markers, including transient localization in the time domain and steady-state confirmation in the semantic domain. First, in the time domain, an event neighborhood is defined centered on the event marker. For example, the neighborhood width is set to a five-second interval before and after the event marker. Within this interval, the amplitude changes of DC bus voltage and DC bus current are scanned, and fluctuation peaks are identified as transient points. Second, in the semantic domain, changes in the controller's operating state are monitored, such as the switch from drive mode to regenerative mode, to confirm that the steady-state interval is a continuous segment where the state remains unchanged. This dual-domain search uses event markers as anchor points to ensure that the search scope focuses on key dynamic processes, avoiding the inefficiency of global scanning.

[0051] Segment boundaries are defined by cross-validation of results from both domains. Transient points in the time domain are matched with steady-state confirmation points in the semantic domain. For example, when a transient point falls near a steady-state change boundary, the timestamp of the transient point is used as the priority boundary. Cross-validation is achieved by calculating temporal overlap. If the overlap exceeds a threshold (the threshold can be determined based on the fluctuation frequency distribution of historical test data, for example, taking 80% of the distribution as the threshold to ensure coverage of typical fluctuations), the boundary is considered valid; otherwise, the neighborhood width is adjusted and the search is repeated. After the boundary is determined, it is marked with start and end timestamps to ensure that the segment boundary accurately corresponds to the data sequence in the original dataset, supporting the accuracy of subsequent type classification.

[0052] Segment Type Classification: Within the boundary, segment types are classified according to the direction of bus energy flow and torque response pattern, including feedback initiation segment, feedback continuous segment, feedback exit segment, load increment segment, load decrement segment, and start-stop transition segment. The direction of bus energy flow is determined by the sign of the power (DC bus voltage multiplied by DC bus current): positive for discharge and negative for feedback. The torque response pattern is analyzed by the slope change of the torque response; for example, a positive slope indicates an upward trend, and a negative slope indicates a downward trend. Combining both, for example, when the energy flow is negative and the torque response slope increases from zero, it is classified as a feedback initiation segment. This classification logic is based on physical meaning, ensuring that the type reflects actual operating conditions, such as the energy feedback process during regenerative braking.

[0053] A set of busbar fluctuation segments is created: For each segment, the source event, start and end timestamps, energy direction label, and torque response label are recorded, forming a set of busbar fluctuation segments. The source event is directly inherited from the event label; for example, a regeneration event corresponds to a feedback-related segment. The energy direction label is marked as positive or negative, and the torque response label is described as rising, converging, or stabilizing. The set is stored as a structured list, with each element containing the above records, ensuring direct access for risk analysis. This set captures the performance of busbar fluctuations under specific events, providing a complete subset for quantitative assessment.

[0054] Establish a steady-state fragment index: A steady-state fragment index is built using fragment start and end times, fragment type, and source event. The index maps to the original dataset using keys. The key is generated by combining the fragment start and end times, fragment type, and source event, such as a timestamp range plus a type string; the value points to the position of the corresponding sequence in the original dataset, enabling fast retrieval; this index ensures consistency and traceability, for example, by using the key to backtrack and verify the accuracy of fragment data, and supports the generation and invocation of reproduction scripts.

[0055] A dual-domain search is constructed around event markers, fragment boundaries are given and fragment types are divided to form a set of bus fluctuation fragments, and a steady-state fragment index is established to map key values ​​to the original dataset.

[0056] Step S2 transforms the original dataset into a structured set of bus fluctuation segments and a steady-state segment index through dual-domain search, boundary determination, type partitioning, set formation, and index establishment. This ensures that the transient and steady-state segmentation accurately corresponds to the regeneration and load switching scenarios in the power supply test of the new energy vehicle motor, providing a traceable segment basis for subsequent ripple current-torque correlation analysis and improving the overall reliability and efficiency of the test.

[0057] Step S2 has generated a set of bus fluctuation segments and a steady-state segment index, providing a structured data subset for subsequent feature extraction. However, to quantify the power supply stability, it is necessary to deeply analyze the fluctuation characteristics and response relationships within the segments. To this end, step S3 constructs trajectories and calculates specific parameters on the set of bus fluctuation segments, enabling table evaluation and risk labeling, and forming a ripple current-torque correlation table, which serves as a blueprint for script construction and configuration.

[0058] The specific processing logic of step S3:

[0059] Step S3 is divided into six sub-processing steps: constructing the ripple amplitude and phase trajectory, extracting the duration and establishing the table relationship, calculating the feedback phase offset, calculating the energy loop coupling area ratio, applying the sequential consistency scoring method and segment risk labeling, and generating the ripple flow-torque correlation table. The technical logic of each step is as follows.

[0060] Constructing the amplitude and phase trajectory of ripple current: Within the set of bus fluctuation segments, a ripple current amplitude and phase trajectory is constructed for the DC bus voltage and DC bus current. The trajectory is jointly described by the zero-crossing sequence and the symbol dwell order to avoid the disruption of phase continuity by carrier frequency disturbances. First, the DC bus voltage and DC bus current sequences within the segment are extracted, and the zero-crossing points of each sequence are identified, i.e., the crossover moments when the values ​​change from positive to negative or from negative to positive. Then, the symbol dwell order is recorded, i.e., the continuous dwell time periods of the sequence in the positive or negative value range. The trajectory is described by combining the two, for example, using the zero-crossing sequence as nodes and the symbol dwell order as edges to construct an ordered path. This trajectory captures the amplitude and phase dynamics of voltage and current, filters high-frequency noise, and ensures that it reflects the true energy flow.

[0061] Duration extraction and mapping: The duration of each segment is extracted and mapped to the synchronous torque response, ensuring a correspondence between the amplitude and phase trends of the electrical ripple current and the rise and convergence of the mechanical torque. The duration is obtained by calculating the difference between the start and end timestamps of the segment, directly extracted from the bus fluctuation segment set. The mapping matches the time points of the ripple current amplitude and phase trajectory with the corresponding points of the torque response. For example, when the trajectory shows a peak amplitude, it checks whether the torque response synchronously shows a rise or convergence trend, establishing a one-to-one correspondence list. This relationship emphasizes electromechanical coupling, ensuring that subsequent parameter calculations are based on a synchronous perspective.

[0062] Calculating the feedback-following offset: Identify the main transition point of the ripple current within the interval from feedback triggering to torque zero crossing, and then identify key inflection points during torque rise or fall. Establish a one-to-one mapping according to time sequence, and take the robust representative of the mapping time difference as the feedback-following offset to quantify the lag of electrical side fluctuations on the mechanical side response. Feedback triggering refers to the start of the regeneration event marker, and torque zero crossing is the moment when the torque response value is zero. The main transition point of the ripple current is determined by detecting the point of maximum amplitude change in the ripple current amplitude trajectory, and the key inflection point is found at the slope reversal point on the torque response curve. After mapping, calculate the time difference for each pair, and select the median as the robust representative to avoid the influence of extreme values. The calculation logic is that the feedback-following offset is equal to the median of all mapping time differences, where each time difference is the time of the key inflection point minus the time of the corresponding main transition point of the ripple current. All calculations are based on time units to ensure dimensional consistency. This parameter quantifies lag; the smaller the value, the more timely the response.

[0063] The energy loop coupling area ratio is calculated as follows: Within the feedback window, a closed write-back trajectory is generated sequentially along the time axis, with the DC bus voltage as the horizontal axis and the DC bus current as the vertical axis. The write-back energy is obtained by accumulating the power and time along the trajectory direction. This energy is then normalized to the total energy within the same window to obtain the energy loop coupling area ratio, which quantifies the convergence and divergence strength of the energy write-back path. The feedback window is defined as the start and end times of the relevant feedback segments; the closed write-back trajectory is formed by connecting sequence points and closing the loop; the write-back energy is calculated by integrating the power along the trajectory and multiplying it by time, with the total energy being the absolute value of the power integral within the window; normalization yields the ratio. The calculation logic is that the energy loop coupling area ratio equals the write-back energy divided by the total energy, where the write-back energy is the product of accumulated power and time along the trajectory direction, and the total energy is the integral of the absolute power value within the same window. All calculations are based on energy units to ensure dimensional consistency. This ratio reflects the path compactness; a value close to one indicates strong coupling.

[0064] The application of sequential consistency scoring and segment risk labeling: The feedback-following offset and the energy loop coupling area ratio are fed into the sequential consistency scoring method. Using a time-series template of "regeneration trigger—writeback peak—torque convergence" as a reference, the sequential matching degree and mutual corroboration degree of the parameters on the template are calculated to obtain a single-value stability coefficient, which is then used to complete segment risk labeling. Specific implementation: The time-series template is defined as the sequence of regeneration event marking, writeback peak (maximum power point of the trajectory), and torque convergence (torque response tending to stabilize). Sequential matching degree checks whether the parameter time points conform to the template sequence; mutual corroboration degree assesses the complementarity of the offset and area ratio, for example, a small offset and a high area ratio indicate a high corroboration degree. The scoring method synthesizes single values, for example, outputting the stability coefficient in a product form. If the coefficient is lower than a threshold (the threshold can be determined by simulating the distribution of stable operating conditions, for example, taking the lower 20% of the distribution as the low stability threshold to ensure risk differentiation), it is labeled as high risk. This method ensures that the coefficient reflects temporal consistency, and the labeling is used for screening.

[0065] Generating a ripple flow-torque correlation table: The final generated table contains segment keys, parameter pairs, stability coefficients, and risk labels. This table uses index keys to trace back to the original dataset and the bus fluctuation segment set, forming a unified interpretation basis. Segment keys are inherited from the steady-state segment index; parameter pairs include feedback phase offset and energy loop coupling area ratio; the table is stored in a database format, with index keys mapped to data locations to ensure traceability. This table integrates evaluation results and supports downstream script construction.

[0066] Within the set of bus fluctuation segments, construct the trajectory and duration of the ripple current amplitude and phase, establish a table relationship, calculate the ratio of feedback phase offset to energy loop coupling area, apply the sequential consistency scoring method to output the stability coefficient and mark the risk, and generate a ripple current-torque correlation table to trace back the data using the index key.

[0067] Step S3, through trajectory construction, relationship establishment, parameter calculation, scoring and labeling, and table generation, realizes the quantitative correspondence and risk identification of ripple current and torque on the set of bus fluctuation segments, providing an operable correlation basis for the stability assessment of new energy vehicle motors in energy feedback scenarios, and improving the transformation of testing from qualitative to quantitative.

[0068] Step S3 has generated a ripple flow-torque correlation table, laying a quantitative foundation for the screening of risk segments and script construction. However, to verify the consistency between the bench test and the actual vehicle, the analysis results must be transformed into an executable reproduction process. Therefore, step S4 constructs a script and configuration based on the ripple flow-torque correlation table to realize bench simulation and evaluate the differences, serving as input to trigger revisions.

[0069] The specific processing logic of step S4:

[0070] Step S4 is divided into six sub-processing steps: screening high-risk segments and typical segments, assembling key transient scripts, loading dynamic configuration of DC source, executing scripts to collect and reproduce datasets, alignment and difference evaluation, and giving overall difference conclusions and sub-item difference mappings. The technical logic of each step is as follows.

[0071] High-risk and typical segments are selected based on the ripple flow-torque correlation table. Risk labels and stability coefficients corresponding to all segment keys are extracted from the table. High-risk segments are defined as those with stability coefficients below a preset threshold. For example, the threshold can be determined through statistical distribution analysis of known stable operating condition samples, using the lower quartile of the sample stability coefficients as the threshold to ensure comprehensive coverage of potential unstable scenarios and avoid over-screening. Typical segments are selected from a representative subset of the remaining segments, such as those with stability coefficients near the median in feedback inrush segments or load increment segments, to represent average operating condition performance. This selection process ensures that the selected segments include both problem areas for diagnosis and benchmark scenarios for comparison, thus providing a balanced input basis for subsequent script assembly and avoiding the omission of key dynamic features.

[0072] Assemble the critical transient script: The script is assembled into a critical transient script, with a structure including trigger order, energy direction, and target phase direction semantics. The trigger order is based on the temporal arrangement of the source events of the selected segments; for example, segments corresponding to regeneration events are processed first, followed by segments corresponding to load switching events, to simulate the continuity of actual operating conditions. The energy direction is directly inherited from the energy direction label, explicitly marked as positive (discharge) or negative (feedback) to guide power flow control in the script. The target phase direction semantics details the expected path of the ripple current amplitude and phase direction trajectory, including the expected frequency of the zero-crossing sequence, the duration of the symbol dwell sequence, and the trend of amplitude changes, ensuring that the script semantics cover all key features of the segments. The script is compiled in an executable serialization format, such as using JSON or a custom protocol for storage, facilitating benchtop hardware parsing and execution. This assembly stage bridges the quantitative analysis to physical simulation, emphasizing semantic integrity and operability, and avoiding the loss of any segment details during the conversion process.

[0073] Dynamic DC Source Configuration Loading: Combined with scripts, dynamic DC source configuration is loaded at the DC port. This configuration includes output impedance semantics, voltage drop recovery semantics, and current limiting switching logic semantics, which together constitute port-level waveform shaping. Output impedance semantics adjusts the DC source's resistance value to match the energy direction label; for example, setting a lower impedance during negative energy flow to simulate a feedback path. Voltage drop recovery semantics defines the voltage recovery curve from fluctuation to stability, based on the amplitude recovery time and phase offset of the ripple current amplitude-phase trajectory; for example, specifying a recovery time constant to match the sign dwell order of the ripple current amplitude-phase trajectory. Current limiting switching logic semantics controls the threshold switching mechanism of the current, corresponding to the transition characteristics of the DC bus current; for example, automatically switching on and off when the current exceeds a preset limit to prevent overload. The limit is based on current sequence statistics from the original dataset. This configuration is applied to the DC port through a dedicated hardware interface (such as a programmable power controller), ensuring the waveform accurately reproduces all transient characteristics of the bus fluctuations. The three elements work together to form complete port-level shaping, avoiding simulation deviations caused by the isolated application of any configuration item.

[0074] The script execution process acquires and reproduces the dataset: After executing the script, a reproduced dataset is generated. The key transient script is run in the bench test environment, and the dynamic configuration of the DC source is applied in real time. The DC bus voltage sequence, DC bus current sequence, controller operating status sequence, and torque response sequence are acquired synchronously, forming a reproduced dataset with a structure completely consistent with the original dataset. This includes a unified single time axis, event marker fields, and records of channel effective bandwidth and sampling windows. The acquisition process uses the same synchronization mechanism as step S1, such as multi-channel analog-to-digital converters and trigger alignment, to ensure high data accuracy and zero latency. This execution step simulates the complete behavior of a real vehicle during feedback and load mutation processes. The generated comparable dataset covers the dynamic response of all selected segments, avoiding any omission of any channel or time period during acquisition.

[0075] Alignment and Dissimilarity Evaluation: One-to-one alignment is performed using event keys and fragment keys. Dissimilarity evaluation is performed around feedback offset, energy loop coupling area ratio, and stability coefficient. Dissimilarity is a semantic synthesis of three items: temporal alignment deviation, energy coupling deviation, and coefficient deviation. Detailed Implementation: Event keys and fragment keys are inherited from the ripple-torque association table and steady-state fragment index, used to match the reproduced dataset with the original dataset one by one, for example, aligning sequences within the corresponding timestamp range by fragment key; the timing alignment bias calculates the absolute difference between the feedback offset of the reproduced dataset and the original dataset, the energy coupling bias calculates the absolute difference between the ratio of the energy loop coupling area of ​​the two datasets, and the coefficient deviation calculates the absolute difference between the stability coefficients of the two datasets; the synthesis uses a weighted summation to quantify the overall difference, where each bias term is first normalized to the range of zero to one (e.g., by dividing by its maximum possible value), and then linearly added with weights. The weights are set according to engineering priorities, for example, the timing alignment bias weight is 0.4 (emphasizing timely response), the energy coupling bias weight is 0.3 (focusing on energy efficiency), and the coefficient deviation weight is 0.3 (balancing overall stability), ensuring that the synthesis considers the contribution of all dimensions and provides interpretability. The calculation logic is as follows: the overall difference is equal to the product of the temporal alignment deviation weight and the absolute difference between the replicated feedback offset and the original feedback offset, plus the product of the energy coupling deviation weight and the absolute difference between the replicated energy loop coupling area ratio and the original energy loop coupling area ratio, plus the product of the coefficient deviation weight and the absolute difference between the replicated stability coefficient and the original stability coefficient. All deviation terms are normalized to be dimensionless. This evaluation process reveals the comprehensive deviation between the simulation and the original, supports subsequent targeted optimization, and avoids any omissions in the calculation of deviation terms.

[0076] Provide overall difference conclusions and itemized difference mappings: Provide overall difference conclusions and retain itemized difference mappings for use in the next step to determine the direction of revision. Specific implementation: The overall difference conclusion is based on a comparison of the overall difference degree with a preset threshold. For example, if the overall difference degree exceeds the threshold, the conclusion is that the simulation is inconsistent and needs revision (the threshold can be determined by the average deviation level of multiple bench and real-vehicle comparisons, for example, taking the upper limit of the validation set deviation as the threshold to ensure practicality and conservatism). Itemized difference mappings record in detail the value of each deviation, the source fragment key, and the corresponding event. For example, timing alignment deviations are mapped to fragments marked with specific regeneration events, and the reasons for the deviations, such as timestamp offsets, are noted. This output is presented in a structured report format, providing clear guidance for the revision path, ensuring that all itemized information is complete, and supporting single-item triggering in step S5.

[0077] Based on the ripple-torque correlation table, the fragment is selected, the key transient script is assembled, and the DC source is dynamically configured. The dataset is then collected and reproduced, the difference is evaluated, and the overall difference conclusion and the sub-item difference mapping are given.

[0078] Step S4 transforms the ripple flow-torque correlation table into bench reproduction results and quantitative deviations by filtering fragments, assembling scripts, loading configurations, executing acquisition, evaluating differences, and mapping conclusions. This achieves a closed-loop bridge from analysis to verification in the power supply test of new energy vehicle motors, improves the accuracy and operability of stability assessment, and provides a detailed deviation basis for iterative revision.

[0079] Step S4 has generated the reproducible dataset and the difference evaluation results, providing a deviation mapping for the optimization path. However, when the deviation exceeds the acceptable range, targeted adjustments are needed to converge the test accuracy. To this end, step S5 performs individual revisions based on the itemized difference mapping, iteratively verifying until the stability requirements are met, forming the final judgment and configuration output.

[0080] The specific processing logic of step S5:

[0081] Step S5 is divided into six sub-processing steps: determining whether the difference has entered the risk range, triggering the single revision path, executing event marking timing revision, executing sampling bandwidth revision, executing DC source dynamic configuration revision, iterative verification, and outputting stability judgment conclusion. The technical logic of each step is as follows.

[0082] Determining if the variance falls into the risk zone: Based on the overall variance conclusion in step S4, determine whether the variance falls into the risk zone. Extract the overall variance value and compare it with a preset risk threshold. If the overall variance is higher than the threshold, it is confirmed to have entered the risk zone (the threshold can be determined by simulating the variance distribution of multiple stable and unstable operating conditions, for example, taking the upper quartile of the distribution as the threshold to ensure the distinction between normal fluctuations and significant deviations). This determination focuses on the stability under the energy feedback scenario, ensuring that revisions are initiated only when necessary, avoiding unnecessary iterative cycles.

[0083] Triggering a Single Revision Path: A single revision path is triggered based on the component difference mapping, following the principle of revising one item and verifying it one step at a time. Specific implementation: The relative magnitudes of timing alignment deviation, energy coupling deviation, and coefficient deviation in the component difference mapping are analyzed. For example, if the timing alignment deviation accounts for the largest proportion, an event-marked timing revision path is triggered. The principle requires revising only one item at a time and verifying its effect subsequently to isolate the influence of variables. This trigger ensures that the revision targets specific pain points in new energy vehicle motor testing, such as torque delay caused by timing mismatch.

[0084] Event marker timing revision: If the main deviation is manifested as timing alignment mismatch, event marker timing revision is performed. The revision strategy is limited to trigger edge positioning and cross-channel delay compensation. Specific implementation: Trigger edge positioning involves rescanning the sign change points of power (DC bus voltage multiplied by DC bus current) on the original dataset. For example, if the original regenerative event marker is located at a coarse position where power changes from positive to negative, it is fine-tuned to the precise sign reversal moment, typically shifted forward or backward by a few tenths of a second to match the start of the drive command interval. Cross-channel delay compensation reapplies timestamp calibration, for example, calculating the delay difference between DC bus current and torque response, and correcting the timestamps of all channels through linear interpolation to ensure that the event marker fields are aligned on a single time axis. After this revision, the event markers more accurately reflect the transient start during load switching, avoiding the amplification of torque response delay by timing deviations.

[0085] Sampling bandwidth revision: If the main deviation manifests as high-frequency fluctuation distortion, sampling bandwidth revision is performed. The revision strategy is limited to channel anti-aliasing settings and effective bandwidth enhancement. Specific implementation: Channel anti-aliasing settings suppress carrier band disturbances by configuring the cutoff frequency of the low-pass filter. For example, if the original settings cause pseudo-phase jumps in the ripple current amplitude-phase trajectory, the cutoff frequency is increased from twice its original value to match the upper limit of the effective bandwidth of the channel, while simultaneously monitoring the continuity of the zero-crossing sequence. Effective bandwidth enhancement increases the sampling rate, for example, from 1,000 samples per second in the original sampling window to 2,000 samples per second, to better capture rapid transitions in the DC bus current and ensure that the ripple current amplitude-phase trajectory is distortion-free within the feedback window. This revision optimizes the data's ability to capture high-frequency transients and is suitable for analyzing the subtle response of bus voltage in regenerative braking.

[0086] Perform DC source dynamic configuration revision: If the main deviation is reflected in the difference in port write-back morphology, perform DC source dynamic configuration revision. The revision strategy is limited to single-domain fine-tuning of output impedance semantics, recovery semantics, and current limiting semantics. Specific implementation: Output impedance semantics fine-tunes the DC source resistance value to optimize the write-back path of the energy direction label. For example, if the reproduction dataset shows scattered feedback energy flow, the impedance is reduced by 10% from its original value to enhance the convergence and dispersion of negative power. Recovery semantics adjusts the parameters of the voltage drop recovery curve, for example, modifying the recovery time constant from one second to 1.5 seconds to match the sign dwell order of the ripple current amplitude phase trajectory. Current limiting semantics refines the current switching threshold. For example, if the original threshold causes premature current transition, the threshold is increased by 5% based on the current sequence statistics of the original dataset to ensure that the current limiting switching logic is consistent with the bus charging and discharging behavior. Each single-domain fine-tuning is performed independently to avoid interfering with other configurations, ensuring that the DC source dynamic configuration more accurately reproduces the vehicle feedback morphology on the test bench.

[0087] Iterative verification and output of stability judgment conclusions: After revision, the process involves further segmentation, feature calculation, script execution, and difference evaluation along the S2-S4 path until the difference degree falls outside the risk range. A stability judgment conclusion and a matching test configuration list are then output, and the final configuration is associated with the fragment key and archived to maintain cross-platform comparability and cross-batch reproducibility. Specific implementation: The process of forming the bus fluctuation fragment set, generating the ripple current-torque association table, assembling key transient scripts, and collecting the reproducibility dataset is repeated until the overall difference degree is below the risk threshold. The stability judgment conclusion is based on the final stability coefficient; for example, if the coefficient is higher than the threshold, stability is determined. The test configuration list includes the revised sampling bandwidth, event marker timing, and DC source dynamic configuration parameters. The archived data is stored using fragment key association. This iteration ensures convergence to reliable results and supports cross-batch verification of new energy vehicle motors in the production process.

[0088] When the difference enters the risk zone, a single revision path is triggered based on the sub-item difference mapping. The revision of the event marking timing, sampling bandwidth, or DC source dynamic configuration is executed. The verification is iterated along the S2-S4 path until the risk zone is exited. The stability judgment conclusion and test configuration list are output and the configuration is sealed.

[0089] Step S5, through risk assessment, path triggering, targeted revision and iterative verification, optimizes test parameters based on sub-item difference mapping, and achieves deviation convergence and output determination in the new energy vehicle motor power supply scenario. This improves the robustness and practicality of the overall method and provides a traceable configuration basis for mass production release.

[0090] Example 2: Figure 2 This invention provides a system for analyzing the power supply stability of a new energy vehicle motor, comprising:

[0091] Data acquisition module: Collects DC bus voltage and current, controller operating status and torque response on a unified time axis, writes event flags for regeneration and load switching and start-stop transitions, and obtains raw dataset.

[0092] Transient segmentation module: Performs transient localization and steady-state segmentation on the original dataset to form a set of bus wave segments and a steady-state segment index, and clarifies the segment boundaries and segment types.

[0093] Risk assessment module: On the set of bus fluctuation segments, the direction and duration of ripple current amplitude and phase and the torque response at the same time are evaluated and used for segment risk labeling and writing into the ripple current-torque correlation table.

[0094] Script Reproduction Module: Based on the ripple-torque correlation table and segment risk labels, key transient scripts and dynamic configurations of DC sources are constructed. The actual vehicle feedback and load change process are reproduced on the test bench. The reproduction dataset is collected and the difference evaluation results between the reproduction dataset and the original dataset are given.

[0095] Revision and optimization module: When the difference is in the risk range, a single revision is performed. The revision path is selected by sampling bandwidth, event marking timing or DC source dynamic configuration. The segmentation, feature extraction and reproduction evaluation are completed again, and the stability judgment conclusion and the matching test configuration list are output.

[0096] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0097] It should be noted that the system of the present invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting a variety of hardware environments and usage requirements.

[0098] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0099] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely to distinguish one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0100] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for testing the power supply stability of a new energy vehicle motor, characterized in that, Including the following steps: S1: Collect DC bus voltage and current, controller operating status and torque response on a unified time axis, write event tags for regeneration and load switching and start-stop transitions to obtain the raw dataset; S2: Perform transient localization and steady-state segmentation on the original dataset to form a set of bus wave segments and a steady-state segment index, and clarify the segment boundaries and segment types; S3: On the busbar fluctuation segment, complete the comparison evaluation of the ripple current amplitude and phase direction and duration with the synchronous torque response, which is used for segment risk labeling and written into the ripple current-torque correlation table; S4: Based on the ripple-torque correlation table and segment risk labels, construct key transient scripts and dynamic configuration of DC sources. Reproduce the vehicle feedback and load change process on the test bench, collect the reproduction dataset and give the evaluation results of the difference between the reproduction dataset and the original dataset. S5: When the difference is in the risk range, perform a single revision, select the revision path of sampling bandwidth, event marking timing or DC source dynamic configuration, complete the segmentation, feature extraction and reproduction evaluation again, and output the stability judgment conclusion and the matching test configuration list.

2. The method for testing the power supply stability of a new energy vehicle motor according to claim 1, characterized in that: A single time axis is generated using a time base, and DC bus voltage, DC bus current, controller operating status and torque response are synchronously acquired. Cross-channel delay errors are eliminated by combining trigger alignment and timestamp calibration.

3. The method for testing the power supply stability of a new energy vehicle motor according to claim 2, characterized in that: Within a single time axis, the event markers for regeneration, load switching, and stop-start transitions are located by combining the controller's operating status and power commutation indicators. Regeneration is confirmed by power direction reversal and drive command intervals, load switching is confirmed by the simultaneous fulfillment of load-side operating condition semantics and current transitions, and stop-start transitions are confirmed by the controller's stack entry / exit status and bus charging / discharging behavior. This results in a raw dataset with event marker fields, and records the channel's effective bandwidth and sampling window.

4. The method for testing the power supply stability of a new energy vehicle motor according to claim 1, characterized in that: Transient localization and steady-state segmentation are performed on the original dataset. A dual-domain search is constructed around the event tags. Transient localization is performed in the time domain with the event neighborhood as the center, and steady-state confirmation is performed in the semantic domain with the controller's running state changes. The segment boundaries are given after cross-verification of the results from the two domains.

5. The method for testing the power supply stability of a new energy vehicle motor according to claim 4, characterized in that: Within the segment boundaries, segments are categorized according to the energy flow direction and torque response pattern of the bus, including feedback inrush segments, feedback continuous segments, feedback exit segments, load increment segments, load decrement segments, and stop-start transition segments. For each segment, the source event, start and end timestamps, energy direction label, and torque response label are recorded to form a set of bus fluctuation segments. A steady-state segment index is established based on the segment start and end time, segment type, and source event. The index is mapped to the original dataset using key-value pairs.

6. The method for testing the power supply stability of a new energy vehicle motor according to claim 1, characterized in that: Within the set of bus fluctuation segments, the amplitude and phase trajectory of the DC bus voltage and DC bus current are constructed. The trajectory is jointly described by the zero-crossing sequence and the sign dwell sequence. The duration is extracted within the same segment and a corresponding relationship is established with the synchronous torque response, so that the amplitude and phase trend of the electrical side ripple current corresponds to the rise and convergence of the mechanical side torque.

7. The method for testing the power supply stability of a new energy vehicle motor according to claim 6, characterized in that: The main transition point of ripple current is identified within the interval from feedback triggering to torque zero crossing. Key inflection points are identified during torque rise or fall. A one-to-one mapping is established according to time sequence. The robust representative of the mapping time difference is taken as the feedback-following offset. Within the feedback window, the DC bus voltage is used as the horizontal axis and the DC bus current is used as the vertical axis to generate a closed write-back trajectory in time sequence. The write-back energy is obtained by accumulating the product of power and time along the trajectory direction. Then, the energy loop coupling area ratio is obtained by normalizing the total energy in the same window. The feedback-following offset and the energy loop coupling area ratio are fed into the sequence consistency scoring method. Using the time sequence template of regeneration triggering-write-back peaking-torque convergence as a reference, the sequence matching degree and mutual corroboration degree of parameters on the template are calculated to obtain the single-value stability coefficient. Based on this, the segment risk labeling is completed. Finally, a ripple current-torque correlation table with segment keys, parameter pairs, stability coefficients and risk labels is generated.

8. The method for testing the power supply stability of a new energy vehicle motor according to claim 1, characterized in that: High-risk and typical segments were selected based on the ripple flow-torque correlation table and compiled into a key transient script. The script structure includes trigger sequence, energy direction and target phase direction semantics. Combined with the script, a dynamic configuration of the DC source was loaded at the DC port. The configuration includes output impedance semantics, voltage drop recovery semantics and current limiting switching logic semantics. After the script was executed, a reproducible dataset was collected and aligned one by one with event keys and segment keys. The difference was evaluated around the feedback phase offset, energy loop coupling area ratio and stability coefficient. The difference was synthesized by three semantics: timing alignment deviation, energy coupling deviation and coefficient deviation. The overall difference conclusion was given and the sub-item difference mapping was retained.

9. The method for testing the power supply stability of a new energy vehicle motor according to claim 8, characterized in that: When the difference enters the risk range, a single revision path is triggered based on the sub-item difference mapping. If the main deviation is manifested as timing alignment mismatch, event marking timing revision is performed. The revision strategy is limited to trigger edge positioning and cross-channel delay compensation. If the main deviation is manifested as high-frequency fluctuation distortion, sampling bandwidth revision is performed. The revision strategy is limited to channel anti-aliasing settings and effective bandwidth enhancement. If the main deviation is manifested as port write-back topology difference, DC source dynamic configuration revision is performed. The revision strategy is limited to single-domain fine-tuning of output impedance semantics, recovery semantics, and current limiting semantics. After the revision is completed, the process of segmentation, feature calculation, script execution, and difference evaluation is carried out again along the path of steps S2-S4 until the difference exits the risk range. The stability judgment conclusion and the matching test configuration list are output, and the final configuration is associated with the fragment key and sealed.

10. A new energy vehicle motor power supply stability analysis system, used to implement the new energy vehicle motor power supply stability test method according to any one of claims 1-9, characterized in that, include: Data acquisition module: Collects DC bus voltage and current, controller operating status and torque response on a unified time axis, writes event tags for regeneration and load switching and start-stop transitions, and obtains raw dataset; Transient segmentation module: Performs transient localization and steady-state segmentation on the original dataset to form a set of bus wave segments and a steady-state segment index, and clarifies segment boundaries and segment types; Risk assessment module: Completes the comparison assessment of the ripple current amplitude and phase direction and duration with the torque response of the same period on the bus fluctuation segment set, which is used for segment risk labeling and writing into the ripple current-torque correlation table; Script Reproduction Module: Based on the ripple-torque correlation table and segment risk labels, key transient scripts and dynamic configuration of DC sources are constructed. The actual vehicle feedback and load change process are reproduced on the test bench. The reproduction dataset is collected and the difference evaluation results between the reproduction dataset and the original dataset are given. Revision and optimization module: When the difference is in the risk range, a single revision is performed. The revision path is selected by sampling bandwidth, event marking timing or DC source dynamic configuration. The segmentation, feature extraction and reproduction evaluation are completed again, and the stability judgment conclusion and the matching test configuration list are output.

Images