Energy-saving DC power supply load

Abstract

The invention discloses an energy-saving direct-current power supply load, particularly relates to the technical field of semiconductor testing, and is used for solving the problems of test waveform distortion and lag caused by mutual interference of an energy recovery system and load current control in a dynamic test of an existing direct-current electronic load. A circuit monitoring module monitors direct current bus voltage ripples and load current tracking errors, a track construction module constructs a dynamic working point track in a phase plane and calculates the divergence of the dynamic working point track relative to a stable operation area, and a feature recognition module recognizes local geometric features of the track when the divergence exceeds a stability threshold value. The weight evaluation module maps the geometrical characteristics into state categories and evaluates collaborative weights, the interaction analysis module analyzes an interaction mode of an energy recovery power regulation rate and a load current change rate according to the collaborative weights, and the collaborative execution module executes collaborative operation of load current control and energy recovery power scheduling. The dynamic test accuracy is improved, and the energy recovery efficiency is optimized.

Description

Technical Field

[0001] This invention relates to the field of semiconductor testing technology, and in particular to an energy-saving DC power supply load. Background Technology

[0002] In the production verification of core components of new energy vehicle electronic control systems, such as motor controllers and on-board chargers, long-term aging tests of high-power semiconductor devices are essential to ensure their vehicle-grade reliability, and DC electronic loads are commonly used as energy-consuming devices. To cope with the complex power variation conditions during vehicle operation, aging tests need to simulate dynamic load curves, including acceleration and braking energy recovery. Simultaneously, to reduce testing energy consumption, the industry has introduced energy recovery technology into the load to convert and reuse the test electrical energy. In existing technical solutions, electronic loads integrating energy recovery functions can balance energy efficiency and basic functionality during steady-state testing.

[0003] However, when the load device executes highly dynamic test waveforms to simulate real vehicle operating conditions, the power scheduling process of its energy recovery system will couple with the rapid tracking control of the load current. This system-level interaction causes distortion and lag in the actual response waveform of the load, compromising the accuracy of dynamic testing. This not only makes the performance evaluation results of automotive-grade semiconductor components under simulated operating conditions inaccurate, but the abnormal operating states it causes also pose a potential risk to the durability of the test system and the device under test, ultimately restricting the reliable application of high-efficiency testing equipment in the aging test of key components of new energy vehicles. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing an energy-saving DC power supply load.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: An energy-saving DC power supply load includes: The circuit monitoring module is used to monitor the DC bus voltage ripple and load current tracking error of the energy recovery circuit. The trajectory construction module is used to construct a dynamic operating point trajectory in the phase plane composed of the energy recovery power and the load current change rate based on the DC bus voltage ripple and load current tracking error, and to calculate the divergence of the dynamic operating point trajectory relative to the stable operating region. The feature recognition module is used to identify the local geometric features of the dynamic working point trajectory when the divergence exceeds the stability threshold. The weight evaluation module is used to map local geometric features to state categories that characterize the dynamic quality of the system, and evaluate the synergistic weight of energy recovery efficiency and load tracking accuracy under the current operating conditions based on the state categories. The interactive analysis module is used to analyze the interaction mode between the energy recovery power regulation rate and the load current change rate based on the cooperative weight. The collaborative execution module is used to perform collaborative operations of load current control and energy recovery power scheduling based on the interaction mode.

[0006] Furthermore, the DC bus voltage ripple and load current tracking error of the energy recovery circuit are monitored, including: The DC bus voltage is continuously sampled by a voltage sensor, and digital filtering is used to separate the DC bus voltage ripple. Meanwhile, the actual current value of the load circuit is obtained through a current sensor, and the actual current value is compared with the reference current value in the dynamic test waveform in real time to calculate the load current tracking error.

[0007] Furthermore, based on the DC bus voltage ripple and load current tracking error, a dynamic operating point trajectory is constructed in the phase plane formed by the energy recovery power and the load current change rate, and the divergence of the dynamic operating point trajectory relative to the stable operating region is calculated, including: Real-time energy recovery power is calculated based on the DC bus voltage ripple amplitude and the operating status of the energy recovery circuit. The load current change rate is obtained by performing time-domain differentiation on the load current tracking error. The real-time energy recovery power and the load current change rate are used to form two-dimensional coordinate points. The continuously collected coordinate points are connected in the phase plane according to the time sequence to form a dynamic working point trajectory. An elliptical stable operating region is established in the phase plane, formed by the aggregation of operating points under historical normal operating conditions; The divergence is quantified by calculating the standard deviation of the minimum distance sequence of each sampling point of the dynamic operating point trajectory relative to the boundary of the elliptical stable operating region. An increase in the standard deviation of the minimum distance sequence indicates that the dynamic operating point trajectory is evolving towards an unstable state.

[0008] Furthermore, the load current change rate is obtained by time-domain differentiation of the load current tracking error, which includes: using the backward difference method to numerically differentiate the continuously sampled load current tracking error sequence, subtracting the load current tracking error values ​​of two adjacent sampling periods and dividing by the sampling time interval to obtain a discrete sequence of the load current change rate, and smoothing the discrete sequence of the load current change rate through a first-order low-pass digital filter to suppress measurement noise interference.

[0009] Furthermore, establishing an elliptical stable operating region in the phase plane formed by the aggregation of operating points under historical normal operating conditions includes: collecting a set of coordinate points of energy recovery power and load current change rate in the phase plane during the historical normal operating phase, calculating the mean and covariance matrix of the coordinate point set in the two coordinate axis directions, and establishing an elliptical region boundary containing a preset percentage of normal operating points with the mean as the center and the eigenvector of the covariance matrix as the axis.

[0010] Furthermore, when the divergence exceeds the stability threshold, the local geometric features of the dynamic operating point trajectory are identified, including: When the divergence exceeds the stability threshold, the trajectory segment consisting of the latest continuous sampling points of the dynamic operating point trajectory is extracted in the phase plane. Calculate the sequence of changes in the turning angle of the line connecting adjacent sampling points in the trajectory segment; Based on the statistical characteristics of the steering angle change sequence, trajectory segments are classified into one of the following geometric features: spiral convergence, straight divergence, or smooth gliding.

[0011] Furthermore, based on the statistical characteristics of the turning angle change sequence, the trajectory segment is classified into one of the following geometric features: spiral convergence, linear divergence, or smooth gliding. These features include: calculating the standard deviation and mean of the turning angle change sequence of the trajectory segment; when the standard deviation is greater than the angle threshold and the mean is close to zero, it is determined to be a spiral convergence pattern; when the standard deviation is less than the angle threshold and the mean is positive, it is determined to be a linear divergence pattern; when the standard deviation is less than the angle threshold and the mean is close to zero, it is determined to be a smooth gliding pattern. The condition for the mean to be close to zero is that the absolute value of the mean is less than 5 degrees, and the condition for the mean to be positive is that the mean is greater than 5 degrees.

[0012] Furthermore, local geometric features are mapped to state categories characterizing the dynamic quality of the system, and the synergistic weights of energy recovery efficiency and load tracking accuracy under the current operating conditions are evaluated based on these state categories, including: When the local geometric features are spiral convergent, the state category is determined to be a stable oscillation state, and a higher weight is assigned to the energy recovery efficiency. When the local geometric features are linearly divergent, the state category is determined to be a rapid instability state, and a higher weight is assigned to the load tracking accuracy. When the local geometry is a smooth gliding pattern, the state category is determined as an efficient cruise state, and the highest weight is assigned to energy recovery efficiency.

[0013] Furthermore, based on the synergistic weighting analysis, the interaction mode between the energy recovery power regulation rate and the load current change rate is analyzed, including: When the collaborative weighting is tilted toward energy recovery efficiency, the rate of change of load current is limited and the rate of energy recovery power regulation is allowed to respond preferentially. When the collaborative weighting is tilted toward load tracking accuracy, the rate of energy recovery power regulation is limited and the rate of load current change is allowed to respond preferentially. When the collaborative weight is in the equilibrium region, a proportional following relationship is established between the energy recovery power regulation rate and the load current change rate.

[0014] Furthermore, the coordinated operation of load current control and energy recovery power scheduling based on the interaction mode includes: When the interaction mode is to limit the rate of change of load current, a smoothing command is output to the load current controller and the response limit on the energy recovery power regulator is removed. When the interaction mode is to limit the energy recovery power regulation rate, a slow change command is output to the energy recovery power regulator and the response limit to the load current controller is removed. When the interaction mode is proportional follower, load current change command and energy recovery power adjustment command are sent synchronously and the change rate ratio of the two is kept consistent.

[0015] The beneficial effects of this invention are: 1. By real-time monitoring of the DC bus voltage ripple and load current tracking error of the energy recovery circuit, and constructing a dynamic operating point trajectory in the phase plane formed by the energy recovery power and the load current change rate, the system can accurately capture the changes in operating status during dynamic aging tests. By calculating the divergence of the trajectory relative to the stable operating region, the system's unstable trends can be identified in a timely manner, thus providing early warning of potential anomalies in the aging tests of automotive-grade semiconductor components. The geometric feature-based identification method effectively avoids the waveform distortion and hysteresis problems caused by the coupling interference between the energy recovery system and the load current control in traditional tests, significantly improving the accuracy and reliability of dynamic tests and ensuring a more realistic and credible performance evaluation of semiconductor devices under simulated real-world conditions.

[0016] 2. By mapping local geometric features to state categories that characterize the dynamic quality of the system, and evaluating the synergistic weights of energy recovery efficiency and load tracking accuracy based on the state categories, intelligent coordination of energy recovery power regulation and load current tracking is achieved. The interaction mode is analyzed according to the synergistic weights, and corresponding synergistic operations are executed. This enables the energy recovery system and load control to adaptively adjust under rapidly changing test waveforms, reducing mutual interference. This not only optimizes test energy consumption and improves energy recovery efficiency, but also ensures rapid and accurate tracking of load current. As a result, the system remains stable during long-term aging tests, extending the service life of test equipment and devices under test, and providing effective support for the reliability verification of key components of new energy vehicles. Attached Figure Description

[0017] Figure 1This is a schematic diagram of the structure of an energy-saving DC power supply load according to the present invention; Figure 2 This is a flowchart illustrating the identification of local geometric features of dynamic working point trajectories in this invention. Detailed Implementation

[0018] 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.

[0019] Example: Figure 1 A schematic diagram of an energy-saving DC power supply load according to the present invention is provided. The energy-saving DC power supply load includes: The circuit monitoring module is used to monitor the DC bus voltage ripple and load current tracking error of the energy recovery circuit. The trajectory construction module is used to construct a dynamic operating point trajectory in the phase plane composed of the energy recovery power and the load current change rate based on the DC bus voltage ripple and load current tracking error, and to calculate the divergence of the dynamic operating point trajectory relative to the stable operating region. The feature recognition module is used to identify the local geometric features of the dynamic working point trajectory when the divergence exceeds the stability threshold. The weight evaluation module is used to map local geometric features to state categories that characterize the dynamic quality of the system, and evaluate the synergistic weight of energy recovery efficiency and load tracking accuracy under the current operating conditions based on the state categories. The interactive analysis module is used to analyze the interaction mode between the energy recovery power regulation rate and the load current change rate based on the cooperative weight. The collaborative execution module is used to perform collaborative operations of load current control and energy recovery power scheduling based on the interaction mode.

[0020] The DC bus voltage ripple and load current tracking error of the energy recovery circuit are monitored, specifically as follows: When monitoring the DC bus voltage ripple and load current tracking error of the energy recovery circuit, the DC bus voltage is continuously sampled using a voltage sensor. The voltage sensor is, for example, an isolated Hall effect voltage sensor, with a measurement range covering 0 volts to 120% of the rated DC bus voltage. The sampling frequency is set to, for example, at least 10 times the switching frequency of the energy recovery circuit. The raw DC bus voltage data obtained from continuous sampling is sent to a digital signal processor for digital filtering. The digital filtering uses, for example, a Butterworth low-pass filter with a cutoff frequency of half the switching frequency. By convolving the raw sampled data with the filter coefficients, the AC ripple component in the DC bus voltage is separated. The amplitude of the DC bus voltage ripple is obtained by calculating the peak-to-peak value of the filtered AC ripple component over one switching cycle.

[0021] Simultaneously, the actual current value of the load circuit is acquired through a current sensor, such as a closed-loop Hall effect current sensor, with a bandwidth no less than five times the maximum frequency of load current change. The sampling period of the actual current value is synchronized with the DC bus voltage sampling, with a sampling interval set to, for example, 100 microseconds. The acquired actual current value is compared in real time with a reference current value in a dynamic test waveform stored in non-volatile memory, containing current change curves simulating the acceleration, braking, and energy recovery conditions of a new energy vehicle. The real-time comparison process is completed within each sampling period. The load current tracking error is obtained by subtracting the reference current value from the actual current value. This error value is a signed value; a positive error indicates that the actual current is greater than the reference current, and a negative error indicates that the actual current is less than the reference current.

[0022] During the monitoring of DC bus voltage ripple, the output voltage signal of the voltage sensor is converted into a digital quantity by an analog-to-digital converter (ADC). The ADC has a resolution of at least 12 bits, and the conversion rate is matched to the sampling frequency. Digital filtering is implemented in a digital signal processor (DSP). The filter coefficients are adaptively adjusted according to the actual operating frequency of the energy recovery circuit. When a change in the switching frequency is detected to exceed, for example, 5%, the filter coefficients are automatically updated to maintain accurate ripple separation. The effective value of the DC bus voltage ripple is obtained by calculating the root mean square (RMS) value of the separated AC ripple component over a complete power frequency cycle.

[0023] During the calculation of load current tracking error, the reference current value is generated in real time based on the preset curve of the dynamic test waveform and the actual test progress, which is recorded by an internal clock counter. The comparison between the actual current value and the reference current value is performed synchronously at each sampling moment. The value of the load current tracking error is range-limited; when the calculated error value exceeds 80% of the current sensor's range, the error value is limited to the sensor's maximum measurable range. The continuous sampling sequence of the load current tracking error is stored in a circular buffer, the length of which can hold at least one complete dynamic test cycle of data for subsequent trajectory construction and analysis.

[0024] Monitoring of DC bus voltage ripple also includes ripple spectrum analysis. By performing a Fast Fourier Transform on the separated AC ripple components, the main frequency components and their amplitudes of the ripple are obtained. The results of the ripple spectrum analysis are used to evaluate the switching state and harmonic characteristics of the energy recovery circuit, and a protection mechanism is activated when abnormal harmonic components are detected. Monitoring of load current tracking error also includes error statistics, calculating the mean, variance, and peak value of the error. These statistical characteristics are used to evaluate the overall performance and quality of load current tracking.

[0025] The calibration process for voltage and current sensors is performed automatically and periodically, with a calibration cycle set to, for example, 24 hours. During calibration, the sensor output is compared with a standard reference value, and the sensor gain and offset parameters are adjusted. The alarm threshold for DC bus voltage ripple is set as a percentage of the rated DC bus voltage. This percentage is determined by statistically analyzing the DC bus voltage ripple data from long-term operation under typical conditions, taking the 95th percentile of the ripple peak-to-peak distribution; for example, this typical value is 5% of the rated voltage. The warning threshold for load current tracking error is set as a percentage of the rated load current. This percentage is determined by analyzing the maximum permissible deviation of the load at rated tracking accuracy; for example, this typical value is 3% of the rated current. When the detected ripple or error exceeds the corresponding threshold, the abnormal event is recorded and the corresponding processing procedure is triggered.

[0026] Real-time monitoring data of DC bus voltage ripple is uploaded to the host computer monitoring system via a communication interface, such as an isolated CAN bus, with the communication cycle synchronized with the sampling cycle. The monitoring results of load current tracking error are also used for load equipment performance evaluation. When the tracking error continuously exceeds the set range for multiple consecutive sampling cycles, the load control parameters are automatically adjusted to improve tracking performance. The timing of the entire monitoring process is controlled by a high-precision timer to ensure strict synchronization of voltage sampling, current sampling, and error calculation, avoiding additional measurement errors introduced by timing deviations.

[0027] Based on the DC bus voltage ripple and load current tracking error, a dynamic operating point trajectory is constructed in the phase plane formed by the energy recovery power and the load current change rate, and the divergence of the dynamic operating point trajectory relative to the stable operating region is calculated. The specific implementation is as follows: When constructing a dynamic operating point trajectory based on DC bus voltage ripple and load current tracking error, the real-time energy recovery power is first calculated according to the DC bus voltage ripple amplitude and the operating status of the energy recovery circuit. The DC bus voltage ripple amplitude is determined by the peak-to-peak value of the ripple obtained during monitoring, and the operating status of the energy recovery circuit is obtained by detecting the duty cycle and conduction state of the power switching devices. The real-time energy recovery power is calculated using the product of the effective value of the ripple voltage and the loop current. The effective value of the ripple voltage is obtained by dividing the ripple amplitude by the square root of two, and the loop current is measured by the output current sensor of the energy recovery circuit. The calculated real-time energy recovery power is converted to standard power units and stored in a temporary register for subsequent processing.

[0028] When obtaining the load current change rate by time-domain differentiation of the load current tracking error, the backward difference method is used to numerically differentiate the continuously sampled load current tracking error sequence. The load current tracking error sequence comes from the error data sequence obtained during monitoring, with a fixed sampling time interval, for example, 100 microseconds. The numerical differentiation process subtracts the load current tracking error values ​​of two adjacent sampling periods and divides the result by the sampling time interval to obtain a discrete sequence of the load current change rate. This discrete sequence is smoothed by a first-order low-pass digital filter. The filter's cutoff frequency is set according to the maximum frequency of load current change, for example, set to half of the maximum frequency of change, to suppress measurement noise interference.

[0029] When constructing a two-dimensional coordinate system using real-time energy recovery power and load current change rate, real-time energy recovery power is used as the x-axis and load current change rate as the y-axis. A two-dimensional coordinate point is generated for each sampling period, and the numerical range of the coordinate points is normalized according to the system's rated parameters. The continuously acquired coordinate points are connected in a time sequence in the phase plane to form a dynamic operating point trajectory. The connection method uses straight line segments to sequentially connect adjacent coordinate points. The trajectory line width is set to, for example, 2 pixels, and the color gradually changes according to the age of the trajectory. The data points of the dynamic operating point trajectory are stored in a circular buffer, with a buffer capacity set to store at least the trajectory data of the most recent, for example, 1000 sampling periods.

[0030] When establishing an elliptical stable operating region in the phase plane, a set of coordinate points representing the energy recovery power and load current change rate during historical normal operation phases is collected in the phase plane. The historical normal operation phase refers to the period during which the system operates without faults during calibration testing, lasting at least, for example, 24 hours. The coordinate point set contains tens of thousands of normal operating points, and the mean and covariance matrix of the coordinate point set are calculated along both coordinate axes. Using the mean as the center point, the eigenvectors of the covariance matrix as the axes, and the eigenvalues ​​as the axial radii, an elliptical region boundary is established. The elliptical region boundary includes a preset percentage of normal operating points. This preset percentage is determined through statistical learning, specifically using a 95% confidence level. That is, the elliptical region is determined by the covariance matrix and mean of the normal operating point set, covering 95% of the sample points.

[0031] When quantifying divergence by calculating the standard deviation of the minimum distance sequence of each sampling point on the dynamic operating point trajectory relative to the boundary of the elliptical stable operating region, the minimum Euclidean distance from each trajectory sampling point to the boundary of the elliptical region is first calculated. The minimum Euclidean distance is calculated using the distance formula from a point to an ellipse, obtaining the nearest boundary point by solving a quadratic equation. The minimum distances of all sampling points are grouped into a sequence, and the standard deviation of this sequence is calculated as the quantification value of divergence. An increase in the standard deviation of the minimum distance sequence indicates that the dynamic operating point trajectory is evolving towards an unstable state. The divergence alarm threshold is determined through historical data analysis; for example, by collecting historical normal operating data to calculate the mean and standard deviation of the divergence, the divergence alarm threshold is set to the mean plus three times the standard deviation. When the divergence exceeds this threshold, an alarm is triggered.

[0032] The calculation of real-time energy recovery power considers the efficiency factor of the energy recovery circuit, and the calculation coefficients are adjusted according to the circuit's operating state using a lookup table method. The data for the lookup table method comes from circuit characteristic tests, and the efficiency coefficients under different operating states are pre-stored in the lookup table. The differential processing of the load current change rate adopts anti-aliasing measures. Before the differential operation, the load current tracking error sequence is subjected to anti-aliasing filtering, and the cutoff frequency of the anti-aliasing filter is set to 1 / 2 of the sampling frequency according to the sampling theorem.

[0033] The elliptical stable operating region is established using an incremental learning mechanism, automatically updating region parameters when the system's operating state changes. Incremental learning is implemented through a sliding window, with the window size set to, for example, 1000 recent operating points. When the matching degree between a new operating point and the existing region falls below a matching degree threshold, a region parameter update is triggered. The matching degree threshold is determined experimentally, for example, by statistically analyzing the distance distribution between the new operating point and the region's centroid, setting the matching degree threshold to a 90% matching ratio. An update is performed when the matching ratio falls below this threshold.

[0034] The quantification process of divergence includes trend analysis, which determines the divergence trend by calculating the slope of the minimum distance sequence. The slope calculation uses the least squares method to fit distance data across multiple consecutive sampling periods; a continuously positive slope indicates a strengthening divergence trend. The divergence alarm threshold is set according to system safety requirements. For example, reliability analysis may determine that a divergence exceeding the normal range by 200% is an alarm condition; when the divergence exceeds this threshold, safety protection is triggered.

[0035] The display update frequency of the dynamic working point trajectory is consistent with the sampling frequency, with the trajectory display updated once per sampling period. Trajectory data is stored using a segmented storage strategy, dividing the trajectory data into time segments and storing them in different storage areas for easy subsequent traceability and analysis. The coordinate range of the phase plane is dynamically adjusted according to the system's operating range, automatically expanding the coordinate range when the working point is detected approaching the boundary.

[0036] The boundary of the elliptical region is displayed visually, with the boundary line color set to, for example, green and the boundary line style set to dashed. A boundary warning is triggered when the dynamic working point trajectory crosses the boundary of the elliptical region. The trigger condition for the boundary warning is set to multiple consecutive sampling points located outside the region. The duration threshold for determining whether a point is outside the region is determined by the system response characteristics, for example, set to 10 consecutive sampling points. A warning is triggered when the trajectory points continuously exceed this threshold.

[0037] The calculation of divergence also includes distinguishing between short-term and long-term divergence. Short-term divergence is calculated using the standard deviation of the most recent 50 sampling points, while long-term divergence is calculated using the standard deviation of the most recent 500 sampling points. The ratio of the two divergences is used to determine the abrupt change characteristics of the system state. The ratio threshold is determined through analysis of historical fault data. For example, by analyzing the changing patterns of the ratio in historical fault data, the ratio threshold is set to 2.0. When the real-time ratio exceeds this threshold, a sudden change in the system state is determined. The specific value of the ratio threshold is calibrated by statistically analyzing the distribution characteristics of the ratio in historical fault events to ensure that the threshold can effectively distinguish between normal and abnormal states.

[0038] The timing control of the entire construction process is synchronized with the sampling period, with one trajectory update and divergence calculation completed in each sampling period. All intermediate variables used in the calculation are stored in floating-point format, ensuring that the calculation accuracy meets the system control requirements. The execution time of all calculation steps has been optimized to ensure that all calculations are completed within a single sampling period.

[0039] Figure 2 A flowchart for identifying local geometric features of a dynamic operating point trajectory according to the present invention is provided. When the divergence exceeds the stability threshold, the local geometric features of the dynamic operating point trajectory are identified. The specific implementation is as follows: When the divergence exceeds the stability threshold, a trajectory segment consisting of the latest continuous sampling points of the dynamic operating point trajectory is extracted from the phase plane. The stability threshold is determined through historical operating data analysis. For example, by collecting the divergence data sequence of the system under normal operating conditions, calculating the statistical distribution characteristics of the sequence, and setting the stability threshold to twice the mean of the divergence plus one standard deviation, the trajectory segment extraction process is initiated when the real-time calculated divergence exceeds this stability threshold. The extraction length of the trajectory segment is set according to the dynamic characteristics of the system, for example, it is set to include the most recent 50 continuous sampling points. These sampling points come from the latest stored data points in the dynamic operating point trajectory buffer and are arranged in chronological order to form the trajectory segment. The extracted trajectory segment contains the coordinate values ​​of the energy recovery power and the load current change rate of each sampling point. These coordinate values ​​are directly read from the dynamic operating point trajectory data in the phase plane.

[0040] When calculating the turning angle change sequence of the lines connecting adjacent sampling points in a trajectory segment, the direction vector of the line connecting adjacent sampling points in the trajectory segment is first calculated. The direction vector is obtained by subtracting the coordinates of the previous sampling point from the coordinates of the subsequent sampling point. The angle between each adjacent line is then calculated as the turning angle. The turning angle is calculated using the vector angle formula, by calculating the dot product and magnitude of two adjacent direction vectors, and then applying the inverse cosine function to obtain the angle value, which ranges from 0 degrees to 180 degrees. The turning angles between all adjacent lines are arranged in chronological order to form a turning angle change sequence. The sequence length is one less than the number of sampling points in the trajectory segment, and the sequence data is stored in a temporary array for subsequent analysis.

[0041] When classifying trajectory segments into one of the geometric features—spiral convergence, linear divergence, or smooth gliding—based on the statistical characteristics of the steering angle change sequence, the standard deviation and mean of the steering angle change sequence are first calculated. The standard deviation is calculated by taking the square root of the average of the squared deviations of each angle value from the mean in the sequence, and the mean is calculated by the arithmetic mean of all angle values ​​in the sequence. The angle threshold is determined experimentally. Specifically, under three known states—stable oscillation, linear divergence, and smooth gliding—a large number of trajectory segment samples are collected. The standard deviation of the steering angle change sequence for each sample is calculated, and the critical value of the standard deviation that most effectively distinguishes the three forms is taken as the angle threshold, for example, 15 degrees. When the standard deviation of the steering angle change sequence is greater than the angle threshold and the mean is close to zero, it is determined to be a spiral convergence pattern; the condition for a mean close to zero is set to an absolute value of less than 5 degrees. When the standard deviation of the steering angle change sequence is less than the angle threshold and the mean is positive, it is determined to be a linear divergence pattern; the condition for a positive mean is set to a mean greater than 5 degrees. When the standard deviation of the steering angle change sequence is less than the angle threshold and the mean is close to zero, it is judged as a smooth gliding pattern. The condition for the mean to be close to zero is set to the absolute value of the mean being less than 5 degrees.

[0042] In the geometric feature classification process, a classification confidence assessment is set up. The reliability of the classification results is evaluated by calculating the degree of matching between statistical features and classification conditions. The confidence assessment uses a weighted scoring method, calculating scores based on the deviations of the standard deviation and mean from thresholds. Specifically, the score calculation first calculates the relative deviation of the standard deviation from the angle threshold by subtracting the angle threshold from the standard deviation, taking the absolute value, and then dividing by the angle threshold. Next, the absolute deviation of the mean from zero is calculated by taking the absolute value of the mean. The deviation of the standard deviation is multiplied by its weight, and the deviation of the mean is multiplied by its weight. The two weighted deviations are then added to obtain the total deviation. The weights of the standard deviation and the mean are determined experimentally; for example, the standard deviation weight is set to 0.7, and the mean weight is set to 0.3. The score is calculated by mapping the total deviation to a range of 0 to 100, for example, using a linear mapping function. When the total deviation is 0, the score is 100; when the total deviation exceeds the maximum allowable deviation, the score is 0. The maximum permissible deviation is determined through historical data statistics, for example, set to 2.0. The final score is used for comparison with a confidence threshold, which is determined through receiver operating characteristic curve analysis to balance the false alarm rate and the false negative rate, for example, set to 60 points. When the score is lower than the confidence threshold, the trajectory segment is required to be re-extracted for analysis.

[0043] During trajectory segment extraction, data validity verification is considered, checking whether the coordinates of each sampling point are within a reasonable range. For example, the energy recovery power does not exceed the system's maximum rated power, and the load current change rate does not exceed the maximum allowable change rate. When abnormal sampling points are found, linear interpolation is used for data repair to ensure the continuity of the trajectory segments. The calculation of the steering angle change sequence includes data smoothing, applying a moving average filter to the original steering angle sequence with a filter window size of 3 data points to reduce the impact of random fluctuations.

[0044] The classification results are stored in association with the trajectory segment data, including the trajectory segment's timestamp, number of sampling points, geometric feature type, and confidence score, for subsequent weight evaluation. The statistical feature calculation of the turning angle change sequence also includes outlier handling; when angle values ​​exceed a reasonable range in the sequence, median filtering is used instead of mean calculation, with the reasonable range set to 0 to 180 degrees. Geometric feature classification employs decision tree logic, first determining the relationship between the standard deviation and the angle threshold, and then further subdividing the classification based on the sign and magnitude of the mean. All calculations are implemented in a digital signal processor, with the calculation cycle synchronized with the sampling cycle to ensure real-time performance.

[0045] The triggering conditions for trajectory segment extraction include stability threshold comparison and duration verification. Extraction is only performed when the divergence exceeds the stability threshold for multiple consecutive sampling periods. The duration threshold is set to 5 consecutive sampling periods. Radians are used as the internal unit for steering angle calculation, but are converted to degrees for display in the output. A timestamp and trajectory segment identifier are appended to the geometric feature classification results for easy comparison and analysis with historical data.

[0046] The entire recognition process includes an error handling mechanism, automatically retrying when trajectory segment capture fails or steering angle calculation is abnormal, with a maximum of 3 retries. The execution time of the recognition process has been optimized to ensure that all calculation steps are completed within a single sampling period. All intermediate variables and result data are stored in non-volatile memory, supporting data recovery after system restart.

[0047] Local geometric features are mapped to state categories characterizing the dynamic quality of the system, and the synergistic weight of energy recovery efficiency and load tracking accuracy under the current operating conditions is evaluated based on the state categories. Specifically, the implementation is as follows: When mapping local geometric features to state categories characterizing the dynamic quality of a system, a correspondence is established based on the classification results obtained during the geometric feature identification process. When the local geometric feature is a spiral convergent shape, the state category is determined to be a stable oscillation state. This mapping is based on the dynamic characteristic of the system, represented by the spiral convergent shape, undergoing damped oscillations near the equilibrium point. When the local geometric feature is a linear divergent shape, the state category is determined to be a rapid instability state. This mapping is based on the dynamic characteristic of the system, represented by the linear divergent shape, where the operating point rapidly deviates from the normal range along a fixed direction. When the local geometric feature is a smooth gliding shape, the state category is determined to be a high-efficiency cruise state. This mapping is based on the dynamic characteristic of the system, represented by the smooth gliding shape, where the operating point moves smoothly along a stable trajectory. During the state category mapping process, a correspondence table between geometric features and dynamic qualities is established, and the mapping results, along with timestamps, are stored in the state record area.

[0048] When evaluating the synergistic weight of energy recovery efficiency and load tracking accuracy under the current operating conditions based on state categories, a weight allocation strategy is determined according to the mapped state categories.

[0049] When the state category is stable oscillation, a higher weight is assigned to energy recovery efficiency. This weight value, Weff, is calculated using the formula Weff = 0.5 + 0.2 × k, where k is the normalized value of the DC bus voltage ripple attenuation coefficient within the current oscillation cycle (0 ≤ k ≤ 1). The load tracking accuracy weight is 1 - Weff. For example, under typical attenuation conditions, the energy recovery efficiency weight is approximately 0.7, and the load tracking accuracy weight is approximately 0.3.

[0050] It is worth noting that when the state category is a stable oscillation state, a higher weight, Weff, is assigned to energy recovery efficiency. Weff is calculated using a formula, where the baseline weight of 0.5 is derived from historical operating data statistics of the system under slight oscillation conditions, representing the initial balance point between energy recovery efficiency and load tracking accuracy. The adjustment coefficient of 0.2 is determined through system stability analysis. The analysis process includes testing the impact of different coefficient values ​​on the system recovery speed in a simulated oscillation environment, and finally selecting the maximum effective coefficient that enables the system to recover stability within a typical oscillation period. The normalized decay coefficient is calculated by the ratio of the current oscillation amplitude to the initial oscillation amplitude, and its value is strictly limited to between 0 and 1. Calculated using this formula, under typical decay conditions, the weight of energy recovery efficiency is approximately 0.7, and the weight of load tracking accuracy is approximately 0.3, ensuring that the system prioritizes energy recovery efficiency while maintaining necessary stability under oscillation conditions.

[0051] When the state category is rapid instability, a higher weight is assigned to load tracking accuracy. This weight value, Wprec, is calculated using the formula Wprec = 0.5 + 0.3 × m, where m is the normalized value of the distance between the current operating point and the boundary of the stable operating region (0 ≤ m ≤ 1, the closer the distance, the larger the value of m). The weight for energy recovery efficiency is 1 - Wprec. For example, when approaching the boundary, the weight for load tracking accuracy is approximately 0.8, and the weight for energy recovery efficiency is approximately 0.2.

[0052] It is worth noting that when the state category is rapid instability, a higher weight, Wprec, is assigned to load tracking accuracy. Wprec is calculated using a formula, where the baseline weight of 0.5 aligns with the initial balance design principle of maintaining consistency with the oscillation state; the adjustment coefficient of 0.3 is determined through instability recovery testing. The testing process includes verifying the correction effect of different coefficient values ​​under simulated instability conditions, ultimately selecting the maximum coefficient that enables the system to initiate effective corrective measures within the shortest control cycle after instability is detected; the normalized distance value is calculated as the ratio of the shortest distance from the current operating point to the boundary of the stable region to the maximum observation distance, with its value strictly limited to between 0 and 1. Calculated using this formula, the load tracking accuracy weight is approximately 0.8 and the energy recovery efficiency weight is approximately 0.2 when approaching the boundary of the stable region, ensuring that the system prioritizes load tracking accuracy in the instability state to prevent further deterioration of the state.

[0053] When the state category is high-efficiency cruise, the highest weight is assigned to energy recovery efficiency, with the weight value Weffmax directly set to 0.9, and the load tracking accuracy weight set to 0.1, in order to maximize energy recovery benefits.

[0054] It is worth noting that when the state category is high-efficiency cruise, the highest weight, Weffmax, is assigned to energy recovery efficiency. Weffmax is directly set to 0.9. This value is determined through energy recovery benefit analysis. The analysis process includes testing the energy recovery efficiency change curves under different weight values. It was determined that when the weight reaches 0.9, the energy recovery efficiency is close to the saturation point. Beyond this value, the efficiency improvement decreases significantly, while the load tracking accuracy begins to drop sharply. At the same time, the load tracking accuracy weight is set to 0.1. This value is determined based on system safety requirements to ensure that the system maintains the minimum necessary load monitoring capability while maximizing energy recovery benefits, preventing the system from completely abandoning its tracking responsibility. This fixed weight allocation scheme enables the system to maximize energy recovery benefits in high-efficiency cruise mode while ensuring basic operational safety.

[0055] The weight allocation process considers the historical state sequence of the system's operation, dynamically adjusting the weights by analyzing the development trends of multiple consecutive state categories. For example, when the system is in a stable oscillation state for multiple consecutive cycles, the weight allocation ratio for energy recovery efficiency is gradually increased. The magnitude of the weight adjustment is determined based on the state duration and oscillation amplitude; the longer the duration and the smaller the oscillation amplitude, the greater the increase in the weight of energy recovery efficiency. The weight calculation uses a sliding window averaging method, with the window size set, for example, to 10 state cycles, and the duration of each state cycle being the interception period of the trajectory segment.

[0056] The evaluation of collaborative weights also includes weight smoothing to prevent abrupt changes in weight values ​​during state category transitions. Smoothing employs a first-order inertial element, with the time constant set according to the system's dynamic response speed, for example, five times the state update cycle. Weight values ​​are stored in floating-point format, with precision retained to two decimal places. After each weight update, the update timestamp and corresponding state category are recorded. A weight rationality check is implemented during the weight evaluation process. When the calculated weight value exceeds a preset range, a limiting process is performed. The preset range is determined by system safety requirements; for example, the lower limit for energy recovery efficiency weight is 0.1, and the upper limit is 0.9.

[0057] The state category mapping is implemented using a lookup table method to establish a dictionary of correspondences between geometric feature types and state categories. The dictionary data is obtained through system identification experiments, recording the correspondence between geometric features and system dynamic qualities under different operating conditions. Mapping confidence is recorded simultaneously during the mapping process. When the confidence of geometric feature classification is low, a conservative strategy is adopted for state category mapping, prioritizing mapping to rapidly unstable states. The weight evaluation algorithm employs a rule-based reasoning mechanism. The rule base contains empirical values ​​for weight allocation corresponding to various state categories, which are determined through a combination of expert knowledge and experimental data.

[0058] The output format of the collaborative weights uses standardized data packets, including energy recovery efficiency weights, load tracking accuracy weights, state category codes, and timestamps. These data packets are transmitted to subsequent processing stages via a communication interface, with the transmission cycle consistent with the state update cycle. Weight trend prediction is implemented during the weight evaluation process. By analyzing the changing patterns of historical weight sequences, the weight change trends for the next few cycles are predicted. The prediction algorithm employs an autoregressive model, with model parameters trained using historical data. The prediction results are used to adjust the control strategy in advance.

[0059] The weight allocation strategy is optimized using an online learning mechanism, automatically adjusting weight values ​​based on system performance feedback after weight allocation. Performance feedback is obtained by evaluating the actual performance of energy recovery efficiency and load tracking accuracy; when actual performance falls short of expectations, the weight allocation ratio is adjusted. Online learning employs gradient descent, with the learning rate adaptively adjusted according to the system's operating state—a smaller learning rate in a stable state and a larger learning rate in a transitional state. The real-time performance of weight evaluation is guaranteed by optimizing algorithm execution time; all calculations are completed within the state update cycle.

[0060] The reliability of state category mapping is ensured through multiple verifications; state mapping is only performed when the geometric feature classification results of multiple consecutive trajectory segments are consistent. The weight evaluation process considers the relative position of the system's operating point and the stable operating region; when the operating point is close to the region boundary, the load tracking accuracy weight is appropriately increased. Weight values ​​are displayed using a visual interface, with different colors indicating different weight ranges for easy monitoring by operators. The entire mapping and evaluation process has data loss protection, and important parameters are stored in non-volatile memory.

[0061] Based on the collaborative weighting analysis, the interaction mode between the energy recovery power regulation rate and the load current change rate is analyzed, and the specific implementation is as follows: When analyzing the interaction mode between the energy recovery power regulation rate and the load current change rate based on the collaborative weighting, the real-time values ​​of the energy recovery efficiency weight and the load tracking accuracy weight are first obtained from the weight evaluation process. The judgment condition for the collaborative weighting to tilt towards energy recovery efficiency is determined by setting a weight threshold. The weight threshold is obtained through multi-objective system optimization testing. Specifically, by running the system under different dynamic test conditions and collecting data on energy recovery efficiency, load tracking accuracy, and system stability, a mapping relationship between weight allocation and system performance is established. Regression analysis is used to determine the weight range that optimizes energy recovery efficiency. The weight threshold is set to indicate a tilt towards energy recovery efficiency when the energy recovery efficiency weight is greater than 0.7 and the load tracking accuracy weight is less than 0.3. This threshold ensures that the system still maintains the necessary stability margin when the energy recovery efficiency reaches its peak. The specific implementation of limiting the load current change rate is to set an upper limit value for the load current change rate. This upper limit value is dynamically adjusted based on the system's rated parameters and current operating conditions. For example, it can be obtained by querying a pre-stored load current change rate limit table. The limit table data is determined based on the system's safe operating range. The calculation of the upper limit value for the load current change rate takes into account the thermal inertia and electrical characteristics of the load equipment. The upper limit value is dynamically adjusted by monitoring the load temperature and current parameters in real time. Simultaneously, the energy recovery power regulation rate is allowed to respond preferentially. This is achieved by removing the restriction on the energy recovery power regulation rate and increasing its control priority. The priority setting is completed by adjusting the task scheduling order in the control loop. The increase in control priority is linearly adjusted according to the specific value of the energy recovery efficiency weight.

[0062] When the collaborative weighting is tilted towards load tracking accuracy, the judgment condition is that the load tracking accuracy weight is greater than 0.7 and the energy recovery efficiency weight is less than 0.3. The weight threshold is set by analyzing the system response characteristics under rapidly changing load conditions. Specifically, the weight distribution when the load tracking error is minimized is recorded in step load and ramp load tests. Cluster analysis is used to identify the weight characteristics corresponding to high-precision tracking, ensuring that tracking accuracy is prioritized when the load changes drastically. Specific measures to limit the energy recovery power regulation rate include setting a maximum allowable value for the energy recovery power regulation rate. This value is calculated by real-time monitoring of DC bus voltage stability and energy recovery circuit status. For example, the limit value is dynamically adjusted based on the ratio of DC bus voltage ripple amplitude to rated voltage. The limit value is calculated using a piecewise linear function, and the maximum allowable value is reduced accordingly when the ripple amplitude increases. Simultaneously, the load current change rate is allowed to respond preferentially. This is achieved by maximizing the response level of the load current controller and reducing its control delay. The reduction in control delay is achieved by optimizing the control algorithm execution path, such as using pre-calculation and caching techniques to reduce real-time calculation time.

[0063] When the collaborative weights are in the equilibrium region, the equilibrium region is defined by statistically analyzing the distribution characteristics of weight values ​​in historical normal operation data. Specifically, the weight sequence during the stable operation phase of the system is collected, and the probability distribution of weight values ​​is calculated using the kernel density estimation method. The region with the highest distribution density is selected as the equilibrium range. When both the energy recovery efficiency weight and the load tracking accuracy weight are within the range of 0.4 to 0.6, it is determined to be in the equilibrium region. This range covers the concentrated area of ​​weight distribution under normal operating conditions. At this time, a proportional following relationship is established between the energy recovery power regulation rate and the load current change rate. The proportional coefficient is calculated based on the ratio of the two weights. For example, the proportional coefficient between the energy recovery power regulation rate and the load current change rate is set as the ratio of the energy recovery efficiency weight and the load tracking accuracy weight. The proportional following relationship is implemented using a closed-loop control method. By comparing the deviation between the actual values ​​of the two rates and the proportional relationship in real time, a proportional-integral (PI) controller is used for dynamic correction. The parameters of the PI controller are tuned according to the system response characteristics.

[0064] The interaction mode analysis process includes smooth transition handling for mode switching. When the cooperative weights switch between different tilt states, a gradual adjustment of the rate limit and proportional coefficient is used to avoid abrupt changes in control commands. The gradual adjustment employs a first-order inertial element, with the time constant set according to the system's dynamic characteristics, for example, 10 times the control cycle. The specific value of the time constant is determined through system step response testing. A hysteresis interval is set for mode switching to prevent frequent switching near weight boundaries. The size of the hysteresis interval is determined experimentally, for example, set to 0.05 of the weight value. The setting of the hysteresis interval is based on a trade-off between system stability and response speed requirements.

[0065] The rate limit is set considering the system's real-time operating status. When the system's operating point is detected to be approaching the boundary of the stable operating region, the rate limit range is automatically tightened. The dynamic adjustment of the limit is based on the distance between the operating point and the boundary; the closer the distance, the smaller the limit. The adjustment curve uses a linear or non-linear function, the specific function form of which is determined through system simulation optimization. A deviation tolerance is set to maintain the proportional following relationship. When the deviation between the actual rate ratio and the target ratio exceeds the tolerance, recalibration is triggered. The tolerance value is set according to the system accuracy requirements, for example, 5% of the target ratio. The determination of the tolerance value considers sensor accuracy and control system resolution.

[0066] The output of the interaction mode includes the current mode type, rate limit value, proportional coefficient, and control priority parameters. The output data is encapsulated in a structure and passed to subsequent control components via shared memory. Anomaly detection is implemented during mode analysis. When the cooperative weight value is abnormal or the rate calculation is incorrect, the system switches to a default safe mode, which employs a conservative rate limiting strategy. The anomaly detection threshold is determined through historical fault data analysis. For example, anomalies are triggered when the weight value exceeds a reasonable range or the rate value exceeds physical limits. The reasonable range is determined through system design parameters and operational experience.

[0067] The execution cycle of the pattern analysis algorithm is consistent with the weight update cycle, with a complete interaction pattern analysis performed once per weight update cycle. Intermediate variables used in the calculation are normalized to ensure numerical computation stability. All pattern parameters are stored in non-volatile memory, supporting online modification and saving. The pattern analysis process includes a self-diagnostic function, periodically checking the algorithm logic and parameter integrity. The self-diagnostic cycle is set to, for example, 1000 analysis cycles, and the self-diagnostic content covers data range checks and logical path verification.

[0068] The interactive mode visualization provides an interface displaying the current mode status, rate limit value, and proportional relationship curve. The display update frequency is synchronized with the mode analysis cycle, facilitating operator monitoring and intervention. The mode analysis logic is implemented using a state machine, with state transition conditions based on cooperative weight values ​​and the system operating point location. The state machine design includes all possible mode switching paths. The entire analysis process optimizes computational complexity, ensuring all calculations are completed within the specified control cycle.

[0069] The coordinated operation of load current control and energy recovery power scheduling is performed based on the interaction mode, specifically as follows: When performing coordinated operation of load current control and energy recovery power scheduling based on the interaction mode, the current mode type and corresponding control parameters are first obtained from the interaction mode analysis process. When the interaction mode is to limit the load current change rate, a smoothing command is output to the load current controller. The smoothing command is generated using a trajectory planning algorithm, which calculates a smooth transition curve based on the current load current value and the target current value. The slope limit of the transition curve is set according to the upper limit of the load current change rate determined during the interaction mode analysis process. This is achieved by converting the upper limit of the load current change rate into the maximum current change per unit time, and then generating a step-type or ramp-type smoothing command based on this change. The step size limit of the step-type smoothing command is determined through system step response testing. During the test, current step signals of different lengths are applied, and the system overshoot and settling time are measured. The step size value that minimizes the overshoot to no more than 5% and minimizes the settling time is selected and set to 5% of the rated current value. The slope limit of the ramp-type smoothing command is determined through spectrum analysis. A Fourier transform is performed on the load current change process to identify the main frequency components. Based on the system's disturbance rejection capability, the slope limit is set to 10% of the rated current change per second. This value effectively suppresses high-frequency disturbances while ensuring response speed. Simultaneously, the response restrictions on the energy recovery power regulator are removed. Specifically, the response priority of the energy recovery power regulator is elevated to the highest level, and its regulation rate restriction is removed. This allows the energy recovery power regulator to freely adjust its output power according to the system state. The magnitude of the priority increase is linearly adjusted according to the specific value of the energy recovery efficiency weight. For example, when the energy recovery efficiency weight reaches 0.9, the response priority is set to the highest level.

[0070] When the interaction mode is to limit the energy recovery power regulation rate, a gradual change command is output to the energy recovery power regulator. The generation of the gradual change command adopts a rate-limiting algorithm. This algorithm is based on the maximum allowable value of the energy recovery power regulation rate determined during the interaction mode analysis, and limits the rate of change of the power regulation command to within this allowable value range. For example, gradual control is achieved by setting the maximum change amplitude per second of the power regulation command. The maximum change amplitude per second is determined according to the thermal capacity and electrical characteristics of the energy recovery circuit, for example, set to 5% of the rated power. At the same time, the response limitation on the load current controller is removed. Specifically, the response delay of the load current controller is minimized, and the requirement for smoothing its output changes is eliminated, enabling the load current controller to respond quickly to changes in load demand. The reduction of response delay is achieved by optimizing the control algorithm execution path, for example, by using pre-calculation and caching techniques to reduce real-time calculation time.

[0071] When the interaction mode is proportional-following, load current change commands and energy recovery power adjustment commands are sent synchronously. The synchronization mechanism uses timestamp alignment to ensure that the two commands are issued within the same control cycle. Maintaining the consistency of their change rates is achieved by calculating the actual ratio of the load current change rate to the energy recovery power adjustment rate in real time and comparing this ratio with the target ratio determined during the interaction mode analysis. When a deviation is detected, the load current change command or energy recovery power adjustment command is fine-tuned to maintain the proportional relationship. For example, if the actual ratio is lower than the target ratio, the load current change rate is appropriately increased or the energy recovery power adjustment rate is decreased. The fine-tuning magnitude is determined based on the deviation, for example, by using a proportional control algorithm to calculate the adjustment amount.

[0072] The collaborative operation process includes validating the output of commands. Verification methods include checking whether the command value is within the device's permissible operating range. For example, the load current command value must not exceed the maximum permissible current of the load device, and the energy recovery power command value must not exceed the maximum processing capacity of the energy recovery circuit. When a command value is detected to be out of bounds, a limiting process is used to restrict the command value to a safe range. The limiting threshold is determined based on the device's technical parameters, such as by consulting the device specifications to obtain the maximum permissible value. Command transmission employs a redundancy check mechanism to ensure the integrity of control commands during transmission. The check algorithm uses a cyclic redundancy check (CRC) code, and the check code length is set according to communication reliability requirements, for example, a 16-bit check code.

[0073] The timing control of collaborative operations is strictly synchronized with the system control cycle, with a complete instruction generation and output process executed once per control cycle. The computational complexity of the instruction generation algorithm has been optimized to ensure that all calculations are completed within the control cycle. The instruction output interface adopts a standardized communication protocol, and the protocol data frame contains the instruction value, timestamp, and verification information. Real-time monitoring is implemented during collaborative operations, with monitoring parameters including instruction execution status, device response time, and control deviation. When an anomaly is detected, an emergency handling procedure is initiated, which includes switching to a safe control mode. The parameters of the safe control mode are determined through system security analysis.

[0074] The command smoothing process employs multiple smoothing algorithms, including moving average filtering, exponential smoothing, and polynomial fitting, selecting the most suitable algorithm based on the system's dynamic characteristics. The generation of gradually changing commands considers the system's inertial characteristics, using different gradually changing curves under different operating conditions. The parameters of these curves are determined through system identification experiments. An adaptive adjustment mechanism is implemented to maintain the proportional following relationship. When the system operating point changes significantly, the proportional coefficient is automatically recalibrated. The calibration period is set according to the system's frequency of change, for example, once every 10 control cycles.

[0075] The collaborative operation process achieves a seamless transition between modes. When the interaction mode changes, control commands are adjusted gradually to avoid impacting the system. The duration of the transition process is set according to system inertia, for example, completing a smooth transition of commands within several control cycles after the mode switch. All control parameters are stored in non-volatile memory, supporting online modification and persistent storage. The collaborative operation process includes a comprehensive status logging function, recording the command value, mode type, and system response data for each control cycle for subsequent analysis and optimization.

[0076] The command output stage includes safety interlock protection, automatically switching to safety control mode when a system anomaly or equipment failure is detected. Safety control mode employs conservative control strategies, such as limiting load current and energy recovery power within rated ranges and reducing system response speed. The trigger conditions for safety interlocks are determined through fault tree analysis, covering various possible abnormal situations. A collaborative operation visualization interface displays control commands, equipment status, and system parameters in real time, facilitating operator monitoring and intervention.

[0077] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0078] It should be noted that this 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 various hardware environments and usage requirements.

[0079] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0080] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0081] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0082] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0083] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0084] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0085] The above are merely specific embodiments 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 scope of the technology 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.

[0086] In conclusion, the above are merely preferred embodiments of the present invention and are 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. An energy-saving DC power supply load, characterized in that, include: The circuit monitoring module is used to monitor the DC bus voltage ripple and load current tracking error of the energy recovery circuit. The trajectory construction module is used to construct a dynamic operating point trajectory in the phase plane composed of the energy recovery power and the load current change rate based on the DC bus voltage ripple and load current tracking error, and to calculate the divergence of the dynamic operating point trajectory relative to the stable operating region. The feature recognition module is used to identify the local geometric features of the dynamic working point trajectory when the divergence exceeds the stability threshold. The weight evaluation module is used to map local geometric features to state categories that characterize the dynamic quality of the system, and evaluate the synergistic weight of energy recovery efficiency and load tracking accuracy under the current operating conditions based on the state categories. The interactive analysis module is used to analyze the interaction mode between the energy recovery power regulation rate and the load current change rate based on the cooperative weight. The collaborative execution module is used to perform collaborative operations of load current control and energy recovery power scheduling based on the interaction mode.

2. The energy-saving DC power supply load according to claim 1, characterized in that, Monitoring the DC bus voltage ripple and load current tracking error of the energy recovery circuit includes: The DC bus voltage is continuously sampled by a voltage sensor, and digital filtering is used to separate the DC bus voltage ripple. Meanwhile, the actual current value of the load circuit is obtained through a current sensor, and the actual current value is compared with the reference current value in the dynamic test waveform in real time to calculate the load current tracking error.

3. The energy-saving DC power supply load according to claim 1, characterized in that, Based on the DC bus voltage ripple and load current tracking error, a dynamic operating point trajectory is constructed in the phase plane formed by the energy recovery power and the load current change rate. The divergence of the dynamic operating point trajectory relative to the stable operating region is calculated, including: Real-time energy recovery power is calculated based on the DC bus voltage ripple amplitude and the operating status of the energy recovery circuit. The load current change rate is obtained by performing time-domain differentiation on the load current tracking error. The real-time energy recovery power and the load current change rate are used to form two-dimensional coordinate points. The continuously collected coordinate points are connected in the phase plane according to the time sequence to form a dynamic working point trajectory. An elliptical stable operating region is established in the phase plane, formed by the aggregation of operating points under historical normal operating conditions; The divergence is quantified by calculating the standard deviation of the minimum distance sequence of each sampling point of the dynamic operating point trajectory relative to the boundary of the elliptical stable operating region. An increase in the standard deviation of the minimum distance sequence indicates that the dynamic operating point trajectory is evolving towards an unstable state.

4. The energy-saving DC power supply load according to claim 3, characterized in that, The process of time-domain differentiation of load current tracking error to obtain load current change rate includes: using the backward difference method to numerically differentiate the continuously sampled load current tracking error sequence, subtracting the load current tracking error values ​​of two adjacent sampling periods and dividing by the sampling time interval to obtain a discrete sequence of load current change rate, and smoothing the discrete sequence of load current change rate through a first-order low-pass digital filter to suppress measurement noise interference.

5. The energy-saving DC power supply load according to claim 3, characterized in that, Establishing an elliptical stable operating region in the phase plane, formed by the aggregation of operating points under historical normal operating conditions, includes: collecting a set of coordinate points for energy recovery power and load current change rate in the phase plane during historical normal operating phases; calculating the mean and covariance matrix of the coordinate point set in the two coordinate axes; and establishing an elliptical region boundary containing a preset percentage of normal operating points with the mean as the center and the eigenvector of the covariance matrix as the axis.

6. The energy-saving DC power supply load according to claim 1, characterized in that, When the divergence exceeds the stability threshold, identify the local geometric features of the dynamic operating point trajectory, including: When the divergence exceeds the stability threshold, the trajectory segment consisting of the latest continuous sampling points of the dynamic operating point trajectory is extracted in the phase plane. Calculate the sequence of changes in the turning angle of the line connecting adjacent sampling points in the trajectory segment; Based on the statistical characteristics of the steering angle change sequence, trajectory segments are classified into one of the following geometric features: spiral convergence, straight divergence, or smooth gliding.

7. The energy-saving DC power supply load according to claim 6, characterized in that, Based on the statistical characteristics of the steering angle change sequence, trajectory segments are classified into one of the following geometric features: spiral convergence, linear divergence, or smooth gliding. These features include: calculating the standard deviation and mean of the steering angle change sequence of the trajectory segment; when the standard deviation is greater than the angle threshold and the mean is close to zero, it is determined to be a spiral convergence pattern; when the standard deviation is less than the angle threshold and the mean is positive, it is determined to be a linear divergence pattern; when the standard deviation is less than the angle threshold and the mean is close to zero, it is determined to be a smooth gliding pattern. The condition for the mean to be close to zero is that the absolute value of the mean is less than 5 degrees, and the condition for the mean to be positive is that the mean is greater than 5 degrees.

8. The energy-saving DC power supply load according to claim 1, characterized in that, Local geometric features are mapped to state categories characterizing the dynamic quality of the system, and the synergistic weights of energy recovery efficiency and load tracking accuracy under the current operating conditions are evaluated based on these state categories, including: When the local geometric features are spiral convergent, the state category is determined to be a stable oscillation state, and a higher weight is assigned to the energy recovery efficiency. When the local geometric features are linearly divergent, the state category is determined to be a rapid instability state, and a higher weight is assigned to the load tracking accuracy. When the local geometry is a smooth gliding pattern, the state category is determined as an efficient cruise state, and the highest weight is assigned to energy recovery efficiency.

9. The energy-saving DC power supply load according to claim 1, characterized in that, The interaction mode between the energy recovery power regulation rate and the load current change rate is analyzed based on the synergistic weighting, including: When the collaborative weighting is tilted toward energy recovery efficiency, the rate of change of load current is limited and the rate of energy recovery power regulation is allowed to respond preferentially. When the collaborative weighting is tilted toward load tracking accuracy, the rate of energy recovery power regulation is limited and the rate of load current change is allowed to respond preferentially. When the collaborative weight is in the equilibrium region, a proportional following relationship is established between the energy recovery power regulation rate and the load current change rate.

10. An energy-saving DC power supply load according to claim 1, characterized in that, The coordinated operation of load current control and energy recovery power scheduling is performed based on the interaction mode, including: When the interaction mode is to limit the rate of change of load current, a smoothing command is output to the load current controller and the response limit on the energy recovery power regulator is removed. When the interaction mode is to limit the energy recovery power regulation rate, a slow change command is output to the energy recovery power regulator and the response limit to the load current controller is removed. When the interaction mode is proportional follower, load current change command and energy recovery power adjustment command are sent synchronously and the change rate ratio of the two is kept consistent.

Images