A software reconfigurable digital power management method and system

Abstract

The application relates to a software reconfigurable digital power management method and system, which can accurately divide the steady state and transient switching interval by monitoring the target load in real time and combining historical data to generate a load behavior sequence, can predict the next working state of the load according to the voltage and current change time sequence and generate power demand information, and can synchronously calculate a joint parameter set after comparison of the power demand information and the current output capacity, and can complete parameter updating and replacement in the mute period of the power switch, so that waveform distortion and system oscillation caused by time sequence conflicts can be effectively avoided; meanwhile, the parameters are real-timely corrected according to the voltage difference between the actual output and the target output, so that the output voltage is stably kept in the allowable range; in conclusion, the application can convert passive compensation into active regulation, improves the monitoring lag and transient distortion problem, has the effects of improving the dynamic response speed and energy regulation precision, and enhancing the system operation reliability and power density.

Description

Technical Field

[0001] This application relates to the technical field of digital power control, and in particular to a software-reconfigurable digital power management method and system. Background Technology

[0002] With the widespread application of high-performance loads such as artificial intelligence computing chips, 5G communication equipment, and automated production equipment, power conversion systems are facing the dual challenges of response speed and energy regulation. Traditional analog power supplies or fixed-parameter digital power supplies generally adopt a static linear control architecture, whose core control strategy and drive configuration are fixed during the equipment manufacturing stage, making it difficult to adapt to the wide-range power fluctuations and high-frequency state switching characteristics of modern loads. When dealing with sudden load changes, such power supplies usually only suppress voltage transients by increasing the output filter capacitor capacity, resulting in increased system size, decreased heat dissipation performance, and limited power density improvement.

[0003] Currently, some existing digital power supplies have introduced preset parameter group switching mechanisms, but they still have significant technical limitations: First, existing load status monitoring mechanisms are lagging. Most solutions are based on output voltage deviation thresholds to trigger control adjustment, and compensation is only performed after a significant voltage drop or overshoot, which is a passive response method and cannot suppress transient disturbances caused by load changes. At the same time, existing detection methods lack in-depth analysis of the historical operating characteristics of the load, making it difficult to identify the periodic patterns and long-term trends of the load. The controller cannot predict changes in power demand in advance and cannot update parameter configurations at the optimal time. Second, the dynamic reconfiguration stability of the control logic in existing technologies is insufficient. The loop parameters, pulse width modulation timing, and power switch drive parameters of the digital controller are strongly coupled. Directly rewriting key parameters while the power supply is continuously operating can easily cause output waveform distortion, control signal interference, and system oscillations due to conflicts between parameter update timing and power device switching timing. This non-smooth switching can affect electromagnetic compatibility performance at best, and lead to power bridge arm shoot-through at worst, posing safety risks to load equipment and power supply hardware. Summary of the Invention

[0004] To address the aforementioned shortcomings, this application provides a software-reconfigurable digital power management method and system.

[0005] The above-mentioned objective of this application is achieved through the following technical solution: A software-reconfigurable digital power management method, applied to a power conversion device including a power switch and a digital controller, includes the following steps: The target load associated with the output terminal of the power conversion device is monitored to obtain the real-time voltage and real-time current of the target load, and the target load behavior sequence is generated by combining the historical load data. The steady-state operating range and transient switching range of the target load are determined based on the target load behavior sequence, and the voltage change time sequence and current change time sequence of the transient switching range are extracted. Based on the timing of voltage and current changes, the next operating state of the target load is deduced and the corresponding power demand information is generated. The first control loop parameters of the power conversion device and the first drive parameters of the power switch are obtained. Based on the first control loop parameters and the first drive parameters, the power output capacity information is obtained, and the power demand information is compared with the power output capacity information. When the comparison result does not meet the preset adaptation conditions, or when a state switching command is received from the target load, a joint parameter set is synchronously calculated based on the power demand information. The joint parameter set includes the second control loop parameters and the second drive parameters. Obtain the power switch action sequence, and input the joint parameter set into the digital controller based on the preset silent period associated with the power switch action sequence to replace the first control loop parameters and the first drive parameters. Based on the voltage difference between the actual output voltage and the target output voltage of the power conversion device using a set of combined parameters, the set of combined parameters is corrected to keep the actual output voltage within the preset allowable range of the target output voltage.

[0006] The second objective of this invention is achieved through the following technical solution: A software-reconfigurable digital power management system includes: The behavior sequence generation module is used to monitor the target load associated with the output terminal of the power conversion device, obtain the real-time voltage and real-time current of the target load, and generate the target load behavior sequence by combining the historical load data. The timing sequence extraction module is used to determine the steady-state operating range and transient switching range of the target load based on the target load behavior sequence, and to extract the voltage change timing sequence and current change timing sequence of the transient switching range. The demand information generation module is used to deduce the next working state of the target load and generate corresponding power demand information based on the voltage change time sequence and the current change time sequence. The power information comparison module is used to obtain the current first control loop parameters of the power conversion device and the current first drive parameters of the power switch, obtain the power output capacity information based on the first control loop parameters and the first drive parameters, and compare the power demand information with the power output capacity information. The joint parameter calculation module is used to synchronously calculate a set of joint parameters based on power demand information when the comparison result does not meet the preset adaptation conditions or when a state switching command is received from the target load. The set of joint parameters includes the second control loop parameters and the second drive parameters. The joint parameter loading module is used to obtain the power switch action sequence and input the joint parameter set into the digital controller based on the preset silent period associated with the power switch action sequence, so as to replace the first control loop parameters and the first drive parameters. The joint parameter correction module is used to correct the joint parameter set based on the voltage difference between the actual output voltage and the target output voltage of the power conversion device, so that the actual output voltage is maintained within the preset allowable range of the target output voltage.

[0007] In summary, the software-reconfigurable digital power management method and system provided in this application can transform passive compensation into active regulation by identifying load behavior sequences and predicting states, thereby solving the problems of load monitoring lag and transient distortion. By comparing power demand and output capacity, the control loop and drive parameters are optimized collaboratively, which can adapt to wide-range power fluctuations and high-frequency state switching. By updating parameters during the power switch's quiet period, timing conflicts and system oscillations can be avoided, improving the safety and stability of reconfiguration. Finally, by using voltage difference closed-loop correction, the output voltage is kept stable, which improves dynamic response and regulation accuracy, as well as the system power density and operational reliability. Attached Figure Description

[0008] Figure 1 This is a flowchart of an embodiment of a software-reconfigurable digital power management method according to this application; Figure 2 This is a flowchart of an implementation of step S10 in an embodiment of a software reconfigurable digital power management method of this application; Figure 3 This is a schematic diagram comparing the dynamic response of output voltage under load change conditions in an embodiment of the software reconfigurable digital power management method of this application with that of the prior art. Detailed Implementation

[0009] The following is in conjunction with the appendix Figures 1-3 This application will be described in further detail.

[0010] In one embodiment, such as Figure 1 As shown, this application discloses a software-reconfigurable digital power management method, applied to a power conversion device including a power switch and a digital controller, specifically including the following steps: S10: Monitor the target load associated with the output terminal of the power conversion device, obtain the real-time voltage and real-time current of the target load, and generate a target load behavior sequence by combining the historical load data. In this embodiment, a power conversion device refers to a device capable of converting electrical energy from one form to another, such as converting direct current to alternating current, or converting electrical energy of one voltage level to electrical energy of another voltage level. A power conversion device typically includes a power switch, a digital controller, and other auxiliary circuits. A power switch is a semiconductor device, such as a MOSFET or IGBT, used in the power conversion device to control the flow of electrical energy. The power switch receives drive signals to control its on / off state, thereby achieving the conversion and regulation of electrical energy. A digital controller is the main control unit in the power conversion device. It processes sensor feedback signals through digital algorithms and generates control commands such as pulse width modulation signals to control the operation of the power switch, thereby regulating the output of the power conversion device. The target load refers to the external equipment or system powered by the power conversion device, whose demand for electrical energy may vary over time. The target load behavior sequence refers to a data sequence generated by monitoring the real-time voltage and current of the target load and combining it with historical data. This sequence represents the electrical characteristics and behavior patterns of the target load under different operating modes.

[0011] Specifically, step S10 can be achieved by connecting a voltage sensor and a current sensor to the output of the power conversion device. Specifically, the voltage sensor and the current sensor periodically sample the real-time voltage and real-time current of the target load. The historical load data can be stored as the average voltage and current values ​​over a period of time. The target load behavior sequence can be composed of real-time sampled values ​​and average values ​​arranged in chronological order to provide an understanding of the basic behavior state of the load.

[0012] S20: Determine the steady-state operating range and transient switching range of the target load based on the target load behavior sequence, and extract the voltage change time sequence and current change time sequence of the transient switching range; In this embodiment, the steady-state operating range refers to the range in which the target load operates under relatively small changes in power demand. Within the steady-state operating range, the electrical parameters of the target load change little. The transient switching range refers to the transition range experienced by the target load when the power demand changes significantly, such as switching from light load to heavy load, or switching from one operating mode to another. The voltage change timing refers to the specific process and important time points of the change in the output voltage of the target load over time within the transient switching range. The current change timing refers to the specific process and important time points of the change in the output current of the target load over time within the transient switching range.

[0013] Specifically, step S20 can use a set threshold for judgment. For example, when the fluctuation amplitude of real-time voltage and current is continuously less than a certain small threshold for a period of time, the time period is determined as the steady-state operating range; when the fluctuation amplitude exceeds a certain large threshold, the time period is determined as the transient switching range. In the transient switching range, the voltage and current values ​​of all sampling points can be recorded as the voltage change timing sequence and the current change timing sequence.

[0014] S30: Based on the voltage and current change timing, deduce the next working state of the target load and generate the corresponding power demand information. In this embodiment, the power demand information refers to a set of data used to describe the power demand of the target load, which is derived from the next operating state of the target load. Examples of such data include target voltage, target current, and maximum allowable rate of change of current.

[0015] Specifically, step S30 can be completed by consulting a preset lookup table, which associates common voltage change timing and current change timing patterns with the next operating state of the defined target load, such as light load, medium load, heavy load, etc.; the power demand information can be obtained from the preset parameter table according to the deduced next operating state to obtain the corresponding target voltage and target current values.

[0016] S40: Obtain the current first control loop parameters of the power conversion device and the current first drive parameters of the power switch, obtain the power output capacity information based on the first control loop parameters and the first drive parameters, and compare the power demand information with the power output capacity information; In this embodiment, the first control loop parameter refers to the control loop parameter currently used by the digital controller, such as the gain parameter of the PID controller, which determines the response characteristics of the power conversion device; the first drive parameter refers to the drive parameter currently used by the power switch, such as the dead time of the pulse width modulation signal, the switching frequency, etc., which affects the switching loss and efficiency of the power switch; the power output capability information refers to the power output range and response capability that the power conversion device can provide under the current first control loop parameter and first drive parameter.

[0017] Specifically, step S40 can obtain the current first control loop parameters and first drive parameters by reading the registers inside the digital controller. At the same time, the power output capability information can be estimated based on the obtained parameters using calibration or empirical formulas to determine the maximum output power and maximum output current of the power conversion device. During comparison, the target power and target current in the power demand information can be compared to see if they are less than or equal to the maximum output power and maximum output current in the power output capability information.

[0018] S50: When the comparison result does not meet the preset adaptation conditions, or when a state switching command is received from the target load, a joint parameter set is synchronously calculated based on the power demand information. The joint parameter set includes the second control loop parameters and the second drive parameters. In this embodiment, the adaptation condition refers to a set of set standards used to determine whether the current power output capacity of the power conversion device meets the power demand of the target load; the state switching instruction refers to an electrical signal instruction from the target load or external system end used to indicate that the target load has changed its working state; the joint parameter set refers to the replacement control loop parameters and drive parameters synchronously calculated based on the power demand information, used to adjust the performance of the power conversion device to adapt to changes in the target load; the second control loop parameter refers to the control loop parameters in the joint parameter set used to replace the first control loop parameter; and the second drive parameter refers to the drive parameters in the joint parameter set used to replace the first drive parameter.

[0019] Furthermore, the preset adaptation conditions may include the following three indicators: The first item is the load state stability condition. If the rate of change of load power is consistently lower than a preset threshold within a number of consecutive sampling periods, it indicates that the load has entered the steady state range. The typical range of the number of consecutive judgment periods is 10 to 50 switching periods. The power change rate threshold can be set proportionally according to the rated power of the device, with a typical value of about 0.5% of the rated power per switching period. The second condition is the prediction error convergence condition. The root mean square value of the rolling prediction error within the statistical window is lower than the preset error upper limit, indicating that the prediction process has fully fit the current load behavior. The typical value of the error upper limit can be set to 2% of the rated power, and the length of the statistical window can be adaptively adjusted according to the load change rate. The third condition is the validity condition of the parameter solution. The joint parameter solution result must meet the closed-loop stability constraint. Specifically, the phase margin of the closed-loop system corresponding to the obtained control loop parameters is not less than 45 degrees and the gain margin is not less than 6dB. If the solution result does not meet the above constraints, the parameter update is abandoned and the controller continues to use the current valid parameter set and re-triggers the solution in the next evaluation cycle.

[0020] Specifically, when the comparison result does not meet the set adaptation conditions, or when a state switching command is received from the target load, a joint parameter set is synchronously calculated based on the power demand information. This joint parameter set includes the second control loop parameters and the second drive parameters. In this case, the stored second control loop parameters and the second drive parameters can be selected from the preset parameter library as the joint parameter set based on the power demand information.

[0021] S60: Obtain the power switch action sequence, and input the joint parameter set into the digital controller based on the preset silent period associated with the power switch action sequence to replace the first control loop parameters and the first drive parameters; In this embodiment, the power switch action sequence refers to the pulse signal sequence output by the digital controller for controlling the power switch to turn on and off; the preset silent period refers to a specified time during which all power switches are in the off state in the power switch action sequence, and this silent period can be used for parameter updates.

[0022] Specifically, step S60 can be achieved by setting a set time window in the digital controller, for example, by forcing all power switches to turn off for a short period of time every set cycle, and using this set time window as a preset silent period; during the preset silent period, the set of joint parameters can be written into the parameter register of the digital controller through a serial communication interface, such as SPI or I2C.

[0023] S70: Based on the voltage difference between the actual output voltage and the target output voltage of the power conversion device based on the joint parameter set, the joint parameter set is corrected so that the actual output voltage is maintained within the preset allowable range of the target output voltage.

[0024] In this embodiment, the actual output voltage refers to the output voltage actually measured during the operation of the power conversion device; the target output voltage refers to the output voltage value that the power conversion device needs to achieve, as set according to the power demand information.

[0025] Specifically, step S70 can be implemented by a proportional controller, that is, adjusting the second control loop parameter in the joint parameter set according to the voltage difference between the actual output voltage and the target output voltage. For example, if the actual voltage is too low, a certain gain parameter is increased proportionally; if the actual voltage is too high, the parameter is decreased proportionally.

[0026] For example, suppose a power conversion device supplies power to a high-performance computing chip, and the computing chip has complex operating modes, switching multiple times between low-power standby, medium-load processing, and high-power burst computing states. In this case, traditional power management methods may respond after the output voltage has already dropped when the computing chip enters high-power burst computing, resulting in voltage transient distortion and affecting the operation of the computing chip.

[0027] This embodiment first monitors the computing chip, which is then the target load. By using voltage and current sensors connected to the output of the power conversion device, the real-time voltage and current of the computing chip are obtained. At the same time, combined with historical load data such as historical power consumption data and operation logs of the computing chip, a target load behavior sequence of the computing chip is generated. This sequence not only records the current electrical state of the computing chip, but also includes the common power consumption mode switching patterns and operating trends of the computing chip. For example, it may be identified that the computing chip usually has a warm-up phase before performing a specific task, and then enters high-power computing.

[0028] Furthermore, based on the target load behavior sequence, the steady-state operating range and transient switching range (e.g., rapid switching from standby to burst computing) of the computing chip are determined. The steady-state operating range is, for example, a long period of standby or running state, and the transient switching range is, for example, a rapid switching from standby to burst computing. When a transient switching range is identified, the timing of voltage and current changes of the computing chip within that range is extracted, and the trajectories of voltage and current changes over time are recorded.

[0029] Furthermore, based on the timing of voltage and current changes, the next operating state of the computing chip is deduced, and corresponding power demand information is generated. For example, it may be predicted that the computing chip is about to enter a burst computing state requiring response and high current output from the standby state, and power demand information including the target steady-state operating point, maximum allowable current change rate, voltage fluctuation window, and transition time is generated. At the same time, the first control loop parameters and first drive parameters currently used by the power conversion device are obtained, and the current power output capability information of the power conversion device is calculated based on the obtained parameters, such as its maximum output current and response speed. Subsequently, the power demand information is compared with the power output capability information. If the comparison result shows that the current power conversion device cannot meet the upcoming high-power burst computing demand of the computing chip, for example, insufficient response speed may cause the voltage drop to exceed the range, or a state switching command is received from the computing chip, then the parameter reconstruction process is triggered.

[0030] At this point, a set of joint parameters is calculated based on the power demand information. These parameters include the second control loop parameters and the second drive parameters. These parameters are generated to meet the upcoming high-power burst computing demands of the computing chip, aiming to provide faster response speed and more stable output performance.

[0031] Furthermore, to ensure the safety and smoothness of the parameter update process, the power switch action sequence is obtained, namely the pulse drive signal output by the digital controller for controlling the power switches, and the preset silent period during which all power switches in the power switch action sequence are turned off is identified. During the silent period, the calculated set of joint parameters is input into the digital controller to replace the original first control loop parameters and first drive parameters, which can avoid waveform distortion, logic glitches or system oscillations that may be caused by directly modifying parameters during the switching process.

[0032] Finally, after the parameters are updated, the actual output voltage of the power conversion device based on the joint parameter set is monitored and compared with the target output voltage. If there is a voltage difference, the joint parameter set is adjusted and corrected according to the voltage difference to ensure that the actual output voltage can be maintained within the preset allowable range of the target output voltage, thereby providing power supply for the high-performance computing chip.

[0033] Through the above examples, firstly, the solution of this embodiment, by combining real-time voltage, real-time current, and historical load data, constructs an understanding of load behavior, which contrasts with the triggering mode based on voltage feedback thresholds used in existing technologies. Specifically, existing technologies often intervene only after the output voltage has already dropped, which is a delayed post-event remedy. In contrast, the solution of this embodiment can identify the operating trend and behavior of the load, thereby enabling the prediction of changes in power demand. This can solve the problem of delayed load change perception and difficulty in updating configurations at the appropriate time in existing technologies.

[0034] Secondly, the solution in this embodiment extrapolates the next operating state of the target load based on the timing of voltage and current changes and generates power demand information, thereby enabling the management of future power demand. Specifically, through timing analysis of the transient switching interval, the target voltage value, target current value, and dynamic constraints of the transient process when the target load reaches the next steady state can be calculated. This allows the power conversion device to adjust its parameters to adapt to the upcoming load changes, rather than responding only after a sudden load change occurs. This effectively avoids transient distortion caused by response lag in traditional solutions.

[0035] Furthermore, the solution in this embodiment, when the comparison result does not meet the adaptation conditions or a state switching command is received, synchronously calculates the joint parameter set based on the power demand information to optimize the control parameters and drive parameters. Compared with the control laws and drive parameters of traditional analog power supplies or fixed-parameter digital power supplies, it is also superior to existing solutions that only introduce multiple parameter switching mechanisms but lack calculation. After the above processing, the solution in this embodiment can calculate the appropriate second control loop parameters and second drive parameters based on the power demand information, ensuring that the power conversion device can provide performance under various complex operating conditions.

[0036] Finally, the solution in this embodiment obtains the power switch action sequence and inputs the joint parameter set into the digital controller to replace the parameters based on a preset silent period. This can solve the system stability problem during the control logic reconfiguration process. By identifying the silent period of the power switch for parameter updates, it can avoid waveform distortion, logic glitches, or system oscillations that may occur when key parameters are directly rewritten online during uninterrupted power supply operation. Through this smooth and safe switching mechanism, the reliability and safety of the power supply system during parameter reconfiguration can be improved, and the threat to precision loads and the power supply hardware itself can be reduced.

[0037] In one embodiment, such as Figure 2 As shown, step S10 includes: S11: Collect the real-time voltage and real-time current of the target load over several consecutive switching cycles, and obtain the historical control command sequence of the target load in the corresponding time period; In this embodiment, step S11 aims to acquire the basic electrical operation data of the target load and the external control information it receives within a specific time window. Real-time voltage and real-time current refer to the basic data characterizing the instantaneous energy consumption and state of the load. Historical control command sequences can reflect the intervention of the load's internal or external control systems on its behavior, which is crucial for understanding the load's dynamic response and predicting its future state. As one implementation method, high-precision voltage and current sensors can be configured at the output of the power conversion device to continuously collect real-time voltage and current data at a preset sampling frequency. Simultaneously, the digital controller can record or receive historical control command sequences sent from the target load's control interface. Alternatively, the sampling circuit inside the power conversion device can be used to digitize the output voltage and current, and data can be exchanged with the target load's communication module to obtain its historical control command sequences.

[0038] S12: Timestamp-align and associate real-time voltage, real-time current and historical control command sequences to generate a raw data set with associated time sequence information; In this embodiment, step S12 aims to ensure that data from different sources maintain consistency in the time dimension so that subsequent analysis can accurately reflect the causal relationship between the electrical state of the load and the control commands at a specific moment. Timestamp alignment refers to integrating scattered data points into a raw data set with a unified time reference. One implementation method is to assign precise timestamps to all collected real-time voltage and current data, as well as received historical control command sequences. By comparing timestamps, voltage and current data from the same or similar moments are associated with the corresponding control commands to form structured data records. Alternatively, synchronous sampling can be used, i.e., simultaneously collecting voltage, current, and control commands within the same clock cycle to achieve time alignment.

[0039] S13: Perform signal decomposition on the real-time voltage and real-time current in the original data set to separate the fundamental components of the real-time voltage and real-time current as well as at least one harmonic component in a specific frequency band. In this embodiment, step S13 aims to extract information about different frequency components from the original electrical signal. The fundamental component represents the main energy transmission mode of the load, while the harmonic components can reflect the nonlinear characteristics, switching noise, or specific dynamic response of the load. As one implementation method, a fast Fourier transform can be used to perform frequency domain analysis on the real-time voltage and real-time current data to separate the fundamental component and its harmonic components. By setting a frequency window, harmonic components in a specific frequency band can be extracted. Alternatively, a digital filter bank, such as a bandpass filter or a notch filter, can be used to separate the fundamental component and the harmonic components in a preset frequency band.

[0040] S14: Extract the load electrical feature vector based on the amplitude, phase and rate of change of the fundamental and harmonic components, and calculate the frequency of change and entropy of the historical control command sequence within a preset unit time. In this embodiment, step S14 aims to extract features that can effectively represent the load state and behavior from complex electrical signals and control commands. The load electrical feature vector provides important characteristics of the load by quantifying the frequency components of voltage and current and their dynamic changes. The change frequency and command entropy value can quantify the activity and complexity of load control commands from an information theory perspective, thereby helping to identify load mode switching or abnormal behavior. For the load electrical feature vector, the instantaneous amplitude and phase of the fundamental and harmonic components, as well as the rate of change of the instantaneous amplitude and phase, can be calculated and combined. For historical control command sequences, the number of times the command codes change within each preset unit of time can be counted as the change frequency, while the command entropy value can be obtained by calculating the probability of occurrence of the command code and its information entropy. Alternatively, wavelet transform can be used to perform multi-scale analysis on voltage and current signals to extract energy distribution and transient features at different scales as electrical feature vectors. For historical control command sequences, a Markov chain model can be used to analyze the command state transition probability, and the command entropy value can be calculated based on this.

[0041] S15: Combine and encode the load electrical characteristic vector, change frequency, and command entropy value in chronological order to generate the target load behavior sequence.

[0042] In this embodiment, step S15 aims to integrate and structure the previously extracted features to form a unified and time-continuous sequence, thereby comprehensively describing the complete behavioral trajectory of the target load from steady state to transient state. As one implementation method, the load electrical feature vector, change frequency, and command entropy value calculated at each time point can be used as a frame or sample, and then arranged in chronological order to form a time series. The encoding method can be direct storage or encoding through a certain compression algorithm; or, a state machine model can be used to map different feature combinations to predefined working modes and record the time point and duration of working mode switching.

[0043] For example, in the digital controller of a power conversion device, a high-speed analog-to-digital converter (ADC) can be integrated to acquire the real-time voltage and current of the target load. For instance, two 12-bit ADC channels with a sampling rate of up to 1Msps can be configured and connected to the voltage and current sampling circuits respectively. Simultaneously, the digital controller can communicate with the control unit of the target load via a universal asynchronous receiver / transmitter (UART) or a serial peripheral interface, receiving and recording its historical control command sequences, such as an 8-bit command code sent every 10ms. All acquired data is accompanied by a microsecond-level timestamp generated by the controller's internal timer. During data processing, the floating-point arithmetic capabilities of a digital signal processor (DSP) can be utilized. The system performs a Fast Fourier Transform on real-time voltage and current data to separate the 50Hz fundamental component and specific subharmonic components such as 100Hz and 150Hz. For each component, its instantaneous amplitude, phase, and rate of change over the past 100ms are calculated to form a load electrical characteristic vector. Simultaneously, within a preset unit time per second, the number of times the command code changes in the historical control command sequence is counted as the change frequency, and its Shannon entropy is calculated as the command entropy value based on the probability of occurrence of the command code. Finally, the obtained characteristic data are packaged into a data frame and stored in a circular buffer in chronological order to form a target load behavior sequence, for example, a data frame is generated every 10ms.

[0044] By employing the aforementioned technical solutions, real-time voltage, real-time current, and historical control command sequences are acquired and timestamped to ensure data consistency and accuracy. Fundamental and harmonic component decomposition of the electrical signals and extraction of multi-dimensional features enable the capture of the complex linear and nonlinear, steady-state and transient behaviors of the load. Furthermore, by combining the frequency of command sequence changes and entropy values, the activity and complexity of the load's internal control logic are revealed. Ultimately, the target load behavior sequence generated from these data features not only contains the load's electrical characteristics but also information about its control layer. This allows for a more accurate characterization of the load's complete operating mode and transient processes, improving the precision of target load state identification and the accuracy of prediction. This effectively avoids problems such as unstable power output or low efficiency caused by inaccurate load behavior prediction.

[0045] In one embodiment, the historical control instruction sequence includes several instruction codes, and step S14 includes: S141: Set an observation time sliding window within a preset unit time, and count the total number of state jumps in the historical control command sequence within the observation time sliding window as the change frequency. In this embodiment, the preset unit time refers to a fixed time period used for segmented analysis of historical control command sequences. Its function is to divide the continuous command stream into manageable segments to facilitate the calculation of command change characteristics within that time period. For example, it can be set to 1 second, 100 milliseconds, or a specific number of switching cycles, depending on the dynamic response speed of the load behavior and the data acquisition frequency. The observation time sliding window refers to a movable time window used for dynamic analysis of the command sequence within the preset unit time period. By sliding this window in steps within the preset unit time period, the change characteristics of the historical control command sequence in different local time periods can be captured, thereby... The analysis focuses on the local dynamic behavior of the commands. The length and step size of the observation time sliding window can be configured according to actual application requirements. For example, the window length can be set to one-tenth of a preset unit time, and the step size can be one command cycle. The total number of state transitions refers to the number of changes between adjacent commands in the historical control command sequence within the observation time sliding window. Its function is to quantify the activity level or frequency of change of the command sequence, directly reflecting the frequency with which the control system adjusts the load state within a specific time period. For example, a state transition can be determined by comparing whether adjacent command codes are the same; if they are different, the count is incremented by one.

[0046] S142: Calculate the probability of occurrence of each instruction code within the sliding window of the observation time, and calculate the negative sum of the product of the probability of occurrence and its logarithm to obtain the first entropy value; In this embodiment, the probability of occurrence of an instruction code refers to the frequency of occurrence of each specific instruction code within the observation time sliding window. Its function is to provide basic data for calculating the complexity of instruction content and to reflect the usage tendency of different control instructions within a specific time period. For example, it can be obtained by counting the number of occurrences of each instruction code and then dividing by the total number of instructions in the window. The first entropy value is a value obtained by negatively summing the product of the probability of occurrence of each instruction code and its logarithm. It is used to represent the complexity of instruction content. The first entropy value reflects the uniformity and randomness of the distribution of instruction codes in the instruction sequence. The higher the first entropy value, the more diverse or complex the instruction content is, and the more difficult it is to predict. For example, the first entropy value can be obtained by using the Shannon entropy calculation formula.

[0047] S143: Calculate the Hamming distance between each adjacent consecutive instruction within the observation time sliding window, and calculate the probability of occurrence of each Hamming distance value in the total number of state transitions; In this embodiment, the Hamming distance refers to the number of corresponding bits that differ between any two adjacent consecutive instruction codes within the observation time sliding window. Its function is to quantify the distance or magnitude of state changes between adjacent instructions, reflecting the severity of the instruction switching from one state to another. For example, for binary instruction codes, the number of 1s can be counted directly after bitwise XOR. For other encoding methods, corresponding distance metrics can be defined. The second entropy value is a value obtained by negatively summing the product of the probability of occurrence of each Hamming distance in the total number of state transitions and its logarithm. It is used to represent the state transition complexity, which reflects the diversity and unpredictability of state transition patterns in the instruction sequence. The higher the second entropy value, the more complex or irregular the state transition. For example, the second entropy value can also be obtained using the Shannon entropy calculation formula, but its input is the probability distribution of the Hamming distance.

[0048] S144: Calculate the negative cumulative sum of the probability of occurrence and its logarithmic product to obtain the second entropy value, and then perform a weighted summation of the first entropy value and the second entropy value according to the preset weight coefficients to obtain the instruction entropy value.

[0049] In this embodiment, weighted summation refers to linearly combining the first entropy value and the second entropy value according to a preset weight coefficient. Its purpose is to comprehensively consider the complexity of instruction content and the complexity of state transitions in order to obtain a more representative instruction entropy value. The weight coefficient can be adjusted according to the actual application scenario and the degree of attention to different complexities. For example, when more attention is paid to the severity of state transitions, the weight of the second entropy value can be appropriately increased.

[0050] For example, assuming the historical control command sequence consists of several 8-bit binary command codes, a sliding observation window containing 10 command codes can be set within a preset unit time period, and the window slides in steps of 1 command code; when the command sequence within the window is "00000001, 00000010, 00000010, 000000100, 00000001, 00000001, 00000010, 000 When the sequence is "00010, 00000100, 00000001", firstly, count the total number of state transitions. For example, a transition from "00000001" to "00000010" counts as one transition, and a transition from "00000010" to "00000100" also counts as one transition. In this example, if the number of transitions is 6, this value is the frequency of change. Secondly, count the probability of occurrence of each instruction code. For example, the probability of occurrence of "00000001"... The first entropy value is calculated based on the following probabilities: "00000010" appears 4 times, with a probability of 0.4; "00000100" appears 3 times, with a probability of 0.3; "00000100" appears 3 times, also with a probability of 0.3. Further, the Hamming distance between adjacent consecutive instructions is calculated. For example, the Hamming distance between "00000001" and "00000010" is 2 because the 2nd and 1st bits are different; the Hamming distance between "00000010" and "00000100" is also 2 because the 3rd and 2nd bits are different. Further, the probability of all Hamming distances occurring in 6 state transitions is calculated. For example, the probability of a Hamming distance of 2 is 100%. Finally, the second entropy value is calculated by weighting the first and second entropy values ​​according to preset weighting coefficients, such as a weight of 0.6 for the first entropy value and 0.4 for the second entropy value.

[0051] By employing the aforementioned technical solution, a comprehensive instruction entropy value can be obtained by calculating the first entropy value representing the complexity of the instruction content and the second entropy value representing the complexity of state transitions, and then performing a weighted summation. This instruction entropy value not only reflects the diversity of the instruction code itself but also quantifies the intensity and pattern of state transitions between instructions, thereby overcoming the limitation that a single entropy value calculation cannot fully capture the dynamic complexity of the instruction sequence. Through this instruction entropy value calculation method, the representation of the target load behavior sequence can be made more accurate, especially when the load switches from a steady state to a transient state or performs complex operations within the transient range. It can provide richer information, thereby improving the prediction accuracy and response speed of the power conversion device for the target load's power demand, ensuring the timeliness and effectiveness of the power management strategy, and thus better adapting to the dynamic changes of the load.

[0052] In one embodiment, step S20 includes: S21: Perform sliding window analysis on the target load behavior sequence, calculate the standard deviation and mean of the load electrical characteristic vector within each time window, and monitor the changes in the historical control command sequence; In this embodiment, a sliding window analysis is performed on the target load behavior sequence to capture the dynamic characteristics of the load behavior through local observation. This can be achieved by setting a fixed-length time window and sliding it across the behavior sequence with a certain step size, or by using an adaptive window size to dynamically adjust the window length according to the rate of change of the behavior sequence. Calculating the standard deviation and mean of the load electrical characteristic vector within each time window is to quantify the dispersion and central tendency of the load electrical characteristics. The standard deviation reflects the volatility of the data, while the mean reflects the average level of the data. It can be obtained by statistically calculating the values ​​of all electrical characteristic vectors within the window, or by using methods such as exponentially weighted moving average and moving standard deviation. Monitoring changes in historical control command sequences is to obtain the impact of external control on load behavior. This can be determined by comparing whether the command sequence in the current window is consistent with the command sequence in the previous window, or by quickly detecting changes by calculating the hash value or checksum of the command sequence.

[0053] S22: When the standard deviation of the load electrical characteristic vector is consistently lower than a preset first threshold within any time window, its mean remains stable, and the time during which the historical control command sequence remains unchanged exceeds a preset first duration, the interval corresponding to the time window is determined to be a steady-state operating interval. In this embodiment, the preset first threshold is used to define the acceptable range of electrical characteristic fluctuations. It can be set empirically based on the output accuracy requirements of the power conversion device or the inherent characteristics of the load, or obtained through system identification. The preset first duration is used to ensure that the stability of the load behavior is not accidental, but a continuous steady-state performance. It can be set based on the system response time or the typical cycle of load switching.

[0054] S23: When the standard deviation of the load electrical characteristic vector exceeds the preset second threshold, or the rate of change of its mean exceeds the preset third threshold, or a predefined jump occurs in the historical control command sequence within any time window, the time window is determined to be the starting point of the transient switching interval, and the tracking continues until the end point. The interval between the starting point and the end point is determined to be the transient switching interval. The end point is when the standard deviation of the load electrical characteristic vector falls to the preset fourth threshold and its mean is stable. In this embodiment, the preset third threshold is used to identify rapid changes in the mean of electrical characteristics, for example, by calculating the ratio of the change in the mean over a unit of time to the mean itself; the predefined jump refers to the appearance of a specific instruction or instruction combination in the historical control instruction sequence that indicates a significant change in the load state; continuous tracking until the termination point refers to the complete capture of the transient switching process, which can be implemented through state machine logic, that is, once a transient state is entered, it is continuously monitored until the steady-state condition is met; the termination point is when the standard deviation of the load electrical characteristic vector falls to the preset fourth threshold and its mean is stable, and the preset fourth threshold is usually close to or the same as the first threshold, indicating that the load electrical characteristics have recovered to a stable state.

[0055] S24: Within the transient switching interval, record the process of real-time voltage and real-time current changing with time, extract the timing points including voltage change rise time, voltage change fall time, current change rate peak and their corresponding times, and generate voltage change timing and current change timing.

[0056] In this embodiment, recording the changes in real-time voltage and current over time within the transient switching interval is to obtain detailed dynamic information about the transient process. This can be achieved through a high-sampling-rate data acquisition terminal or information acquisition system, allowing for complete storage of the voltage and current waveforms during the transient period. Extracting time series points, including the rise time of voltage change, the fall time of voltage change, the peak value of current change rate, and their corresponding moments, is to extract key dynamic features from the recorded waveforms. For example, the rise time and fall time can be determined by detecting significant changes in the slope of the voltage waveform; the peak value of the current change rate can be obtained by differentiating the current waveform and finding its maximum value. Generating voltage change time series and current change time series refers to organizing key time series points and their corresponding values ​​into structured time series data for processing and analysis.

[0057] For example, suppose the power conversion device is a digital power supply used to power a high-performance processor, which is the target load. During operation, it will frequently switch between different working modes, such as switching from a low-power standby mode to a high-load computing mode. First, the digital controller will continuously sample the processor's supply voltage and current at high frequency, and combine them with the processor's internal power management instructions, i.e., the historical control instruction sequence, to generate the target load behavior sequence.

[0058] Furthermore, a sliding window analysis is performed on the target load behavior sequence. For example, a window containing 100 sampling points is set and slides in steps of 50 sampling points. Within each window, the standard deviation and mean of the load electrical characteristic vector are calculated. The load electrical characteristic vector is, for example, composed of the fundamental voltage amplitude, fundamental current amplitude, and harmonic content. Simultaneously, changes in the processor's power management instructions are monitored within the window. When the processor is in low-power standby mode, the standard deviation of its supply voltage and current electrical characteristic vectors will consistently be below a preset first threshold (e.g., 0.005V), the mean will remain at a low and stable level, and the power management instruction sequence will not change within a preset first duration (e.g., 50 milliseconds). At this time, the interval corresponding to this time window is determined to be the steady-state operating interval. When the processor suddenly switches from standby mode to high-load computing mode, its supply current will increase rapidly, causing a momentary voltage drop. At this time, a negative voltage drop is detected. If the standard deviation of the load electrical characteristic vector rapidly exceeds a preset second threshold (e.g., 0.05V), or the rate of change of its mean exceeds a preset third threshold (e.g., 0.1V / ms), or a predefined jump command to enter high load mode is directly received from the processor, the time window is immediately determined as the starting point of the transient switching interval. Subsequently, tracking continues until voltage and current fluctuations stabilize, i.e., the standard deviation of the load electrical characteristic vector falls back to a preset fourth threshold (e.g., 0.01V) and the mean is stable. At this point, the transient switching interval is determined as the ending point. Simultaneously, throughout the entire transient switching interval, the waveform data of real-time voltage and real-time current are continuously recorded. From the recorded data, the rise time required for the voltage to rise from a drop to a stable state, the fall time of the voltage drop, and the maximum peak value of the rate of change of current during the transient process and its occurrence time can be extracted as key timing points, and voltage change timing and current change timing can be generated.

[0059] Through the above technical solution, this application can identify the steady-state operating range and transient switching range of the target load, and extract key dynamic information in the transient process. This enables the power management system to adjust the control parameters in a timely and accurate manner based on accurate load status information, thereby effectively responding to rapid load changes and avoiding overshoot or undershoot of output voltage caused by inaccurate status identification. It has the effect of improving the output stability, response speed and overall efficiency of the power conversion device under dynamic load conditions.

[0060] In one embodiment, step S30 includes: S31: Perform differentiation and integration on the voltage change timing and current change timing to calculate the instantaneous power change curve, power change rate curve and load impedance change trend curve of the target load; In this embodiment, differentiation processing can capture the instantaneous rate of change of a signal and reveal its dynamic characteristics, while integration processing can accumulate the influence of the signal and reflect its total amount or trend over a period of time. The derived features obtained through differentiation and integration processing can more comprehensively and precisely describe the energy demand and electrical behavior of the load during transient switching. As one implementation method, discrete differentiation and integration algorithms can be executed by the arithmetic logic unit in a digital signal processor or microcontroller, such as using the difference method for differentiation and the accumulation method for integration. Alternatively, a field-programmable gate array can be used to implement hardware-accelerated differentiator and integrator modules, and the instantaneous power change curve, power change rate curve, and load impedance change trend curve of the target load can be calculated in real time in a parallel processing manner, thereby improving processing speed and response capability.

[0061] S32: Based on the power change rate curve and the load impedance change trend curve, an adaptive filtering algorithm is used to extract the main trend component of the change, and the main trend component is pushed forward to the preset prediction time domain to obtain the predicted power trajectory and the predicted impedance trajectory. In this embodiment, since the power change rate curve and load impedance change trend curve may contain noise and minor fluctuations, the adaptive filtering algorithm can dynamically adjust the filtering parameters according to the statistical characteristics of the signal, effectively filter out noise, and extract the main trend component reflecting the core behavior of the load. The main trend component is then extrapolated to a preset prediction time domain to predict the power and impedance change path of the load in the future, providing forward-looking information for the power management system. The adaptive filtering algorithm can use Kalman filtering or extended Kalman filtering to estimate the system state in real time and predict the future state based on the system model and the statistical characteristics of the measurement noise. Alternatively, it can use adaptive digital filters such as the least mean square algorithm or recursive least squares algorithm to iteratively adjust the filter coefficients so that the output is as close as possible to the desired signal, thereby extracting the main trend of the signal. The extrapolation can be achieved through linear regression, polynomial fitting, or a prediction model based on historical trends.

[0062] S33: Based on the predicted power trajectory and the predicted impedance trajectory, calculate the target voltage and target current values ​​when the target load reaches the next steady state, and extract the key dynamic constraints in the transition process; In this embodiment, the predicted power trajectory and predicted impedance trajectory can provide predictive information about the future behavior of the load. Based on the predicted information, it can be inferred that the load will stabilize in a certain electrical state after completing the transient switching, i.e., its target voltage and target current values. At the same time, during the transition from the current state to the next steady state, the load has specific requirements for the power supply's response speed, allowable voltage drop or overshoot, current change rate, etc., i.e., key dynamic constraints, which are important indicators to ensure stable operation of the load and reliable power supply. The target voltage and target current values ​​can be calculated by combining the predicted steady-state power and impedance with Ohm's law and the power formula. Key dynamic constraints can be preset according to the load type and application scenario. For example, for digital logic circuits, the voltage fluctuation window may be very narrow; for motor loads, the current change rate may need to be limited to avoid overcurrent impact; or, they can also be calculated by lookup table method or based on machine learning model.

[0063] S34: Based on the target voltage value, target current value, and key dynamic constraints, generate power demand information, which includes the target steady-state operating point, the maximum allowable rate of change of current, the voltage fluctuation window, and the expected transition time.

[0064] In this embodiment, the target voltage value, target current value, and key dynamic constraints are integrated to form structured power demand information. This clearly defines the final state that the power supply needs to achieve and the performance indicators that must be met during the transition process, thereby guiding the power controller to adjust parameters. The power demand information can be encapsulated as a data structure or message package containing the aforementioned parameters. For example, the target steady-state operating point can be represented by a pair of target voltage and target current; the maximum allowable current change rate can be a numerical value; the voltage fluctuation window can be represented by the minimum allowable voltage and the maximum allowable voltage; and the expected transition time can be a time value. These parameters can be directly stored in the registers or memory of the digital controller, or encoded as specific instructions or status words and transmitted to the digital controller through the internal communication bus.

[0065] For example, suppose a power conversion device is supplying power to a high-performance processor. When the processor switches from a low-power standby mode to a high-load operating mode, its voltage and current change drastically. First, the digital controller continuously monitors and records the real-time voltage and current at the processor's output, forming voltage and current change time sequences. To gain a deeper understanding of the processor's power requirements, these time sequences are processed. For instance, by calculating the voltage and current differences between adjacent sampling points, the instantaneous rate of change of voltage and current can be obtained, leading to the calculation of the processor's instantaneous power change curve and power change rate curve. Simultaneously, by summing the voltage and current and applying Ohm's law, the processor's equivalent load impedance change trend curve can be derived. Furthermore, to eliminate the influence of measurement noise and short-term fluctuations, a Kalman filter algorithm can be used to process the power change rate curve and load impedance change trend curve to extract their smoothness. The primary trend component, for example, is estimated and predicted in real time by the Kalman filter based on the preset system dynamic model and measurement noise covariance. Furthermore, the filtered primary trend component is extrapolated forward; for example, a linear prediction model can be used to predict the processor's power and impedance change trajectories within the next 100 microseconds. Based on the predicted power and impedance trajectories, the target voltage and current values ​​required for the processor to reach a high-load steady state can be calculated. Simultaneously, based on the processor's electrical characteristics, key dynamic constraints during the transition process can be extracted; for example, the voltage fluctuation window may be limited to ±50mV, the maximum allowable current change rate may be 10A / microsecond, and the expected transition time is 50 microseconds. Finally, the calculated target voltage, target current, and key dynamic constraints are integrated to generate power demand information, which explicitly includes the target steady-state operating point, the maximum allowable current change rate, the voltage fluctuation window, and the expected transition time.

[0066] Through the above technical solution, this application can extract instantaneous power change curves, power change rate curves, and load impedance change trend curves from the original voltage change time series and current change time series through differentiation and integration processing, thereby more comprehensively depicting the dynamic behavior of the load. Based on the extracted derived curves, an adaptive filtering algorithm is used to extract the main trend component and extrapolate it forward, which can achieve accurate prediction of the future power demand of the load, rather than just a delayed response. Through this forward prediction, the power management system can calculate in advance the target voltage and target current values ​​when the target load reaches the next steady state, and extract the key dynamic constraints in the transition process, thereby generating power demand information containing detailed information such as the target steady-state operating point, the maximum allowable current change rate, the voltage fluctuation window, and the expected transition time. This has the effect of improving the accuracy and completeness of power demand information, enabling the power conversion device to adapt to the transient changes of the target load more quickly, smoothly, and accurately, effectively avoiding voltage drops, overshoots, or system instability caused by inaccurate predictions or delayed responses, thereby improving the dynamic response performance and power supply quality of the power system.

[0067] In one embodiment, step S40 includes: S41: Based on the parameters of the first control loop, calculate the current stable output boundary and small-signal frequency response characteristics of the control loop of the power conversion device; In this embodiment, the first control loop parameter refers to the control loop parameter used to adjust the output voltage or current of the power conversion device, such as the proportional gain, integral time constant, and derivative time constant in a PID controller, or the feedback gain matrix in a state-space controller. The first control loop parameter can directly determine the dynamic response characteristics and steady-state accuracy of the control loop. The first control loop parameter can be pre-stored in the memory of the digital controller or obtained through an online identification algorithm.

[0068] Furthermore, the stable output boundary can be characterized by the maximum allowable bandwidth determined by the loop gain and phase margin, and the small-signal frequency response characteristics can include the open-loop crossover frequency and the resonant peak amplitude.

[0069] S42: Calculate the current maximum safe operating boundary of the power switch based on the first driving parameters; In this embodiment, the first driving parameter refers to the parameter used to control the operation of the power switch, such as the switching frequency, dead time, gate drive voltage or current of the pulse width modulation signal, etc. The first driving parameter can directly affect the switching loss, conduction loss and transient current stress of the power switch. The first driving parameter can be generated by the digital controller according to the preset strategy or set through the external configuration interface.

[0070] Furthermore, the maximum safe operating boundary can be jointly defined by the maximum permissible duty cycle, the minimum conduction time, and the safe area for switching losses.

[0071] S43: By combining the stable output boundary, small signal frequency response characteristics, and maximum safe operating boundary, the power output capability information is obtained. The power output capability information includes the steady-state adjustable output range, the maximum safe output current, the load step amplitude, and the load step rate.

[0072] In this embodiment, the power output capability information refers to a comprehensive representation of the power output range and dynamic response capability that the power conversion device can provide under the current configuration. For example, it may include maximum output power, maximum output current, voltage regulation range, transient response time, etc. The comparison refers to comparing the power demand information of the target load with the power output capability information of the power conversion device to determine whether the power conversion device can meet the power demand of the target load. For example, it can be done by comparing whether the target steady-state operating point in the power demand information falls within the steady-state adjustable output range in the power output capability information, or by comparing whether the maximum allowable current change rate in the power demand information is less than the load step rate in the power output capability information.

[0073] For example, assuming the power conversion device uses a digital PID controller, its first control loop parameters include proportional gain Kp, integral time Ki, and derivative time Kd, and the power switch drive parameters include PWM switching frequency fsw and dead time td. First, by analyzing the mathematical model of the power conversion device and combining it with the current Kp, Ki, and Kd, the open-loop transfer function of the control loop can be calculated, and then a Bode plot can be drawn. From this plot, the phase margin, gain margin, open-loop crossover frequency, and resonant peak amplitude can be extracted. For example, if the phase margin is lower than the preset value, it indicates insufficient system stability, and its maximum allowable bandwidth will be limited. Simultaneously, based on the power switch model and the current drive parameters fsw and td, combined with its datasheet, the maximum allowable duty cycle, minimum on-time, and switching loss safety zone at different operating points can be calculated. For example, if the switching frequency is too high, the switching loss may exceed the safety zone. In summary... The calculation results yield the steady-state adjustable output voltage range of the power conversion device under the current parameters (e.g., the output voltage can be adjusted between 10V and 15V while ensuring stability), the maximum safe output current (e.g., the maximum output current is 5A without damaging the power switch), and the response capability to load steps (e.g., it can respond to a 2A current step within 100ns with voltage fluctuations not exceeding 50mV). Subsequently, the power output capability information is compared with the power demand information of the target load. For example, if the power demand of the target load is an output voltage of 12V, a maximum output current of 6A, and a requirement to respond to a 3A current step within 50ns, then the comparison reveals that the maximum safe output current of the current power conversion device, 5A, is less than the target load demand of 6A, and the load step response capability also fails to meet the requirements. In this case, the comparison result does not meet the preset adaptation conditions, triggering the parameter adjustment process.

[0074] Through the above technical solutions, the power conversion device can accurately assess its own power output capacity, avoiding parameter adjustments when it does not have the capacity to meet load requirements. This improves the accuracy and effectiveness of parameter adjustments, helps prevent system overload, oscillation, or performance degradation, and ensures that the power conversion device can operate stably and efficiently under various load conditions.

[0075] In one embodiment, step S50 includes: S51: Using power demand information as joint constraints, establish the system dynamic equations, including the power level topology of the power conversion device, the output filter model, and the equivalent model of the target load. In this embodiment, power demand information refers to the specific power requirements of the target load in the next operating state, such as the target steady-state operating point, maximum allowable rate of change of current, voltage fluctuation window, and expected transition time. This information can set clear performance boundaries and dynamic response requirements for the operation of the power conversion device. Joint constraints refer to the power demand information serving as necessary boundary conditions and performance indicators during parameter calculation, guiding the optimization direction and range of parameters. The power stage topology of the power conversion device refers to the core circuit structure that realizes power conversion in the power conversion device, such as buck, boost, buck-boost, flyback, or forward converters. Its model describes the relationship between physical quantities such as voltage, current, and switching states of the power stage circuit. The output filter model... A filter circuit model is used to smooth the output voltage and current of a power conversion device. It is usually composed of inductors and capacitors and describes the filter's ability to suppress high-frequency ripples and its impact on the system's dynamic response. A target load equivalent model is a mathematical model that abstracts the electrical characteristics of the target load, such as a purely resistive load, a constant current load, a constant power load, or a more complex dynamic load model, to simulate the load's impact on the power conversion device's output. The system dynamic equation is a mathematical expression that describes the changes of the internal state variables of the power conversion device over time under different operating conditions. It can be established using methods such as Kirchhoff's laws, state-space averaging, or switching function methods, and is used to comprehensively characterize the dynamic behavior of the power conversion device under given topology, filter, and load conditions.

[0076] S52: Based on the system dynamic equations, and with the phase margin, maximum allowable bandwidth, and peak output impedance of the control loop of the power conversion device as optimization objectives, the parameters of the second control loop are obtained in parallel under joint constraints. In this embodiment, the phase margin of the control loop is an important indicator for measuring the stability of the control system. It represents the phase margin from the unstable state at the gain crossover frequency. Generally, a higher phase margin represents better stability. The maximum allowable bandwidth characterizes the highest frequency range in which the control loop can effectively respond. It is usually represented by the maximum allowable bandwidth determined by the loop gain and phase margin. Generally, a larger bandwidth represents a faster dynamic response speed. The peak output impedance refers to the maximum equivalent impedance at the output of the power conversion device at a specific frequency. It reflects the power conversion device's ability to suppress load changes. Generally, a lower peak output impedance represents a better load regulation rate and smaller output voltage fluctuations. The optimization objective refers to the performance indicators that need to be considered and optimized simultaneously when solving for the second control loop parameters. Parallel solution means finding the optimal second control loop parameters simultaneously or collaboratively under the premise of satisfying joint constraints, so that the above multiple optimization objectives achieve the best balance. The second control loop parameters refer to the control algorithm parameters used by the digital controller to adjust the output voltage or current under new power demand, such as the Kp, Ki, and Kd parameters of the PID controller, or the feedback gain matrix in the state-space controller.

[0077] S53: Taking the transient current stress, switching loss and conduction loss of the power switch as internal optimization objectives, the second driving parameters are obtained by solving under joint constraints. In this embodiment, transient current stress of the power switch refers to the instantaneous large current surge that the power switch experiences during the switching process. Excessive transient current stress may damage the power switch or shorten its lifespan. Switching loss refers to the energy loss caused by the overlap of voltage and current during the power switch's conduction and turn-off processes, which mainly occurs during the switching transition. Conduction loss refers to the energy loss caused by the power switch's on-resistance when it is in the conduction state. The internal optimization objective refers to the performance indicators of the power switch itself that need to be considered and optimized when solving for the second driving parameters. Solving refers to determining the optimal second driving parameters by means of calculation or simulation, under the premise of satisfying the joint constraints. The second driving parameters refer to the parameters used to control the conduction and turn-off behavior of the power switch, such as gate drive voltage, drive current, dead time, switching frequency, or the rise and fall rates of the pulse width modulation signal.

[0078] S54: Encapsulate and associate the second control loop parameters with the second drive parameters to generate a joint parameter set.

[0079] In this embodiment, encapsulation and association refer to organizing the calculated second control loop parameters and second drive parameters into a logical unit or data structure so that the digital controller can perform unified management and updates; the joint parameter set refers to a complete dataset including the second control loop parameters and second drive parameters, which represents all configurable parameters required for the power conversion device to achieve optimal performance under new power demand.

[0080] Furthermore, the solution process for the joint parameter set is based on the small-signal model of the power conversion device. The control loop parameters and drive parameters are used as joint optimization variables. The weighted combination of output voltage overshoot, dynamic response time and output impedance peak value is used as the optimization objective. At the same time, phase margin, amplitude margin and control bandwidth range are used as constraints to form a constrained parameter optimization problem.

[0081] Regarding convergence, for basic topologies such as Buck and Boost, their small-signal models are minimum-phase systems, and the objective function has good convexity in the stable region. The sequential quadratic programming algorithm can guarantee convergence to a local optimum. For non-minimum-phase topologies such as flyback converters, the particle swarm optimization algorithm is used for global search, and a maximum iteration limit is set to ensure real-time constraints and avoid the solution process from continuing indefinitely.

[0082] In terms of real-time performance, the above solution process can be executed asynchronously in the background task, completely decoupled from the foreground control interrupt, and does not occupy real-time control resources. Taking a typical digital signal processor with a main frequency of 200MHz as an example, when the control parameter dimension is 6 to 8, the single solution time does not exceed 500 microseconds, which is much smaller than the typical time constant of load state switching, thus meeting the real-time requirements. If the solution does not converge within the specified time, the update is abandoned and the current parameters remain unchanged, and the system security is not affected.

[0083] For example, suppose the power conversion device is a digitally controlled synchronous buck converter, whose target load switches from light load to heavy load, requiring the output voltage to stabilize at 5V, the maximum allowable current change rate to be 10A / microsecond, the voltage fluctuation window to be less than 50mV, and the expected transition time to be 10 microseconds. First, using the power demand information as joint constraints, the system dynamic equations of the synchronous buck converter are established, including its power stage topology (such as high-side MOSFETs, low-side MOSFETs, inductors, and output capacitors), the output LC filter model, and the equivalent model of the target load under heavy load conditions. (For example, a low-impedance constant power load model), the dynamic equations of this system can be modeled using the state-space averaging method, resulting in a set of differential equations describing the dynamic behavior of the inductor current and output capacitor voltage. Furthermore, based on these dynamic equations, the parameters of the second control loop and the second drive parameters are solved in parallel. For the second control loop parameters, the optimization objectives can be set as follows: a phase margin of not less than 60 degrees, a gain crossover frequency between 1 / 10 and 1 / 5 of the switching frequency, and a peak output impedance of less than 0.1 ohms. Under the constraint of satisfying the above power demand information, the following can be adopted: Optimization algorithms such as genetic algorithms or particle swarm optimization, or frequency domain shaping techniques, are used to iteratively calculate the Kp, Ki, and Kd parameters of the PID controller. Simultaneously, for the second driving parameter, the optimization objective can be set as: the junction temperature of the power switch not exceeding a preset maximum value, minimizing switching losses, and minimizing conduction losses. Under the constraint of meeting power demand information, simulation can be performed using the SPICE model of the power switch, or by consulting a pre-established lookup table of driving parameters, losses, and stress relationships, the optimal gate drive resistor value, dead time, and the rise and fall rates of the PWM signal can be solved. For example, to reduce transient current stress, the gate drive resistor may need to be adjusted appropriately to slow down the switching speed; to reduce switching losses, the dead time may need to be optimized. Finally, the calculated PID parameters are used as the second control loop parameters, and the gate drive resistor value and dead time are used as the second driving parameters, encapsulated into a data structure to generate a joint parameter set, such as a C language structure containing multiple fields. This joint parameter set contains all the configuration information required for the digital controller to achieve optimal performance under heavy load conditions.

[0084] Through the above technical solution, the power conversion device can calculate the matching second control loop parameters and second drive parameters according to the real-time power demand of the target load. This method based on system dynamic equations and multi-objective collaborative optimization can overcome the blindness and local optima problems that may exist in traditional parameter adjustment methods. It can ensure that when the load condition changes drastically or a fast response is required, the power conversion device can not only maintain excellent output voltage stability, fast dynamic response and low output impedance, but also optimize the operating efficiency of power switches and reduce transient current stress, thereby extending the service life of power switches and improving the adaptability, reliability and overall performance of software reconfigurable digital power management, enabling the power conversion device to cope with complex load conditions more efficiently.

[0085] In one embodiment, step S60 includes: S61: Monitor the pulse drive signal output by the digital controller for controlling the power switch to obtain the power switch action sequence; In this embodiment, the pulse drive signal refers to the control command issued by the digital controller to the power switch. Its high and low level changes directly determine the power switch's on and off states. By monitoring the pulse drive signal, key timing information such as the power switch's on-time, off-time, and dead time in each switching cycle can be understood, thereby constructing a complete power switch action sequence. For example, the level of the PWM output pin can be directly sampled through the digital controller's general-purpose GPIO port, or the current PWM signal state can be obtained by reading the status register of the PWM module inside the digital controller.

[0086] S62: Based on the power switch action sequence, determine the silent period in each switching cycle when all power switches are in the off state, and preload the joint parameter set into the parameter buffer of the digital controller before the silent period begins. In this embodiment, the silent period refers to the time window during one or more switching cycles of the power conversion device when all power switches are in the off state. In typical switching converters, to avoid short circuits caused by simultaneous conduction of the upper and lower bridge arm power switches, a short dead time is usually inserted within a switching cycle so that neither power switch will conduct simultaneously during switching state transitions. This dead time is a common silent period. The purpose of determining the silent period is to provide a safety window for parameter updates, avoiding parameter modifications during the transient process of power switch conduction or turn-off, which could lead to system oscillations or output anomalies. Furthermore, in addition to the dead time, power switching can also be analyzed... The rate switch action sequence identifies specific time periods when all switches are in a non-conducting state; the parameter buffer refers to the memory area inside the digital controller used to temporarily store parameters to be updated; the preload operation refers to writing the pre-calculated set of joint parameters into the parameter buffer before the actual parameter update action occurs, so as to ensure that the required new parameters are ready when the parameter update is needed, without the need for real-time calculation or reading from external memory, thereby shortening the time required for the actual update operation and improving the real-time performance of the update; the preload can be written directly to memory by the main processor of the digital controller, or the data can be transferred from the main memory to the parameter buffer through the direct memory access controller.

[0087] S63: At the beginning of the silent period, the set of joint parameters in the parameter buffer is synchronously updated to the running register group of the digital controller so that the second control loop parameters and the second drive parameters replace the first control loop parameters and the first drive parameters.

[0088] In this embodiment, the running register group refers to the set of hardware registers in the digital controller that actually control the operation of the power conversion device, such as the coefficient register of the PID controller, the duty cycle register of the PWM generator, and the dead time register. Synchronous update refers to writing the new parameters preloaded in the parameter buffer into the running register group at the precise start time of the silent period to ensure that all relevant parameters take effect at the same time and avoid transient mismatch problems that may be caused by batch updates of parameters. For example, the transmission of parameters from the buffer to the running register group can be triggered by hardware interrupts, timer events, or specific synchronization signals of the PWM module to ensure the precise timing of the update. The ultimate goal of parameter update is to replace the first control loop parameters and the first drive parameters with the second control loop parameters and the second drive parameters, that is, to replace the currently used first control loop parameters and the first drive parameters with the new or next working state second control loop parameters and the second drive parameters, so that the control strategy and power switch driving mode of the power conversion device are switched to the new configuration, thereby enabling it to better respond to changes in the power demand of the target load and achieve better performance and stability.

[0089] For example, a high-performance digital signal processor (DSP) can be used as the digital controller. This DSP typically integrates multiple enhanced pulse width modulation (PWM) modules to generate pulse drive signals for controlling the power switch. To obtain the power switch action sequence, the DSP can monitor the rising and falling edges of the PWM signal output by the PWM modules in real time through its internal event manager or capture module, or directly read the counter and compare register values ​​of the PWM modules to infer the current PWM signal state. When determining the silent period, the dead-time generator function built into the PWM module can be used. This function automatically inserts a preset dead time between complementary PWM signals; this dead time is a typical silent period. The DSP firmware can calculate the precise start and end times of the dead time within each switching cycle based on the configuration of the PWM modules. When updating parameters, such as switching from the first control loop parameters and the first drive parameters to the second control loop parameters and the second drive parameters, the central processing unit or direct memory access controller of the digital signal processor will write a pre-calculated set of joint parameters (e.g., new PID controller coefficients, new PWM duty cycle limits, new dead time settings, etc.) into the random access memory area inside the digital signal processor as a parameter buffer before the start of the silent period. Subsequently, at the start of the next silent period, an enhanced pulse width modulation module or an independent timer can be configured to trigger a hardware event. This hardware event will copy the data in the parameter buffer to the running register group of the enhanced pulse width modulation module, ADC module, or control loop calculation module in one go. For example, the zero-point event or period matching event of enhanced pulse width modulation can be used as the trigger point for synchronous update to ensure that the parameter switching is completed when the power switch is in the off state.

[0090] By identifying the silent period of the power switch and preloading and synchronously updating the parameters during the silent period, the transient impact, abnormal fluctuations in output voltage or current, and system instability that may be caused by modifying parameters during the transient process of the power switch being turned on or off can be effectively avoided. This enables the power conversion device to seamlessly switch to the optimal control strategy and drive parameters according to the real-time changes of the target load, thereby ensuring the continuity, stability and high quality of the power output, and improving the adaptability and reliability when dealing with dynamic load changes.

[0091] In one embodiment, step S70 includes: (a) Continuously monitor the actual output voltage of the power conversion device based on the joint parameter set, and calculate the real-time voltage difference and its changing trend between the actual output voltage and the target output voltage; (b) Based on the real-time voltage difference, perform integral compensation on the second control loop parameters in the joint parameter set to generate preliminary correction parameters; (c) Based on the real-time voltage difference change trend, the preliminary correction parameters are adjusted by feedforward to generate the final correction parameters; (d) Update the final corrected parameters to the run register set of the digital controller to replace the second control loop parameters; (e) Repeat steps (a), (b), (c), and (d) in sequence until the real-time voltage difference remains stable within the preset allowable range.

[0092] In this embodiment, step (a) aims to acquire the deviation information between the output state and the desired state of the power conversion device in real time. Specifically, the digital controller can periodically sample the actual output voltage of the power conversion device through an analog-to-digital converter and compare the sampled value with the preset target output voltage to calculate the real-time voltage difference. Simultaneously, by analyzing the continuous real-time voltage difference data, such as calculating its first-order difference or using a moving average algorithm, the trend of voltage difference changes can be obtained, which helps predict the future trend of the voltage. Step (b) aims to eliminate the steady-state error of the output voltage of the power conversion device. Specifically, when there is a continuous non-zero real-time voltage difference between the actual output voltage and the target output voltage, the integral compensation mechanism accumulates this error and adjusts the parameters of the second control loop according to the accumulated amount, thereby gradually driving the actual output voltage to converge towards the target output voltage until the steady-state error is eliminated. Step (c) aims to improve the dynamic response speed and stability of the power conversion device to cope with rapidly changing loads or input disturbances. Specifically, by analyzing the trend of real-time voltage difference changes, the future voltage deviation can be predicted, and a feedforward amount can be calculated based on this. The corrected parameters are superimposed on the initial correction parameters after integral compensation, thereby pre-adjusting before the error actually occurs and effectively suppressing transient voltage fluctuations. Step (d) aims to apply the corrected control parameters to the control algorithm of the digital controller. Specifically, the digital controller writes the calculated final correction parameters into its internal running register group for executing the control algorithm through an internal bus or a specific data transmission mechanism. Once the running register group is updated, the digital controller will immediately start using the new parameters to adjust the action of the power switch, thereby changing the output characteristics of the power conversion device. Steps (a), (b), (c), and (d) are repeated in sequence until the real-time voltage difference remains stable within the preset allowable range. This step (e) constitutes a closed-loop adaptive correction process to ensure that the power conversion device can converge to and maintain the desired stable state after dynamic changes. Specifically, the cycle of voltage monitoring, error calculation, integral compensation, feedforward adjustment, and parameter updating is continuously executed until the real-time voltage difference between the actual output voltage and the target output voltage is detected to remain within the preset minimum allowable range for a period of time. At this point, it indicates that the system has reached a stable and accurate output state.

[0093] For example, as a specific implementation, when the power conversion device (e.g., a digitally controlled DC-DC converter) receives a state switching command from the target load, it has already calculated and updated the joint parameter set based on the power demand information. At this time, to further optimize the accuracy and stability of the output voltage, the digital controller initiates the following correction process: The digital controller continuously acquires the actual output voltage of the power conversion device at a microsecond-level sampling frequency through its built-in high-speed analog-to-digital converter. For example, if the target output voltage is set to 1.0V, and the actual output voltage is 1.005V, then the real-time voltage difference is +5mV. The digital controller simultaneously calculates the continuous rate of change of this real-time voltage difference to determine whether the voltage tends to rise or fall. Based on this real-time voltage difference, the digital controller calls its internal integral compensation algorithm, for example, by... The real-time voltage difference is accumulated and multiplied by an integral gain to generate a correction amount for the second control loop parameters. Simultaneously, if a rapid increase in the voltage difference is detected, for example, if the trend shows that the voltage is continuously rising, the digital controller will calculate a feedforward adjustment amount based on a preset feedforward model or lookup table. This feedforward adjustment amount can reduce the proportional gain in the second control loop parameters in advance to suppress further voltage rise. The two correction amounts are superimposed to form the final correction parameters, which are written to the operating register group responsible for voltage loop control in the digital controller via the internal bus, replacing the current second control loop parameters. After that, the digital controller will continue to monitor the output voltage and repeat the above correction steps until the actual output voltage is continuously stable within the allowable range of ±2mV of the target voltage of 1.0V, for example, fluctuating between 0.998V and 1.002V.

[0094] Furthermore, Table 1 compares the key parameters of the output voltage dynamic response under load change conditions. It shows the comparison of the key parameters of the output voltage dynamic response between the prior art and the present application under load change conditions. As can be seen from Table 1, compared with the prior art, the voltage drop amplitude of the present application is reduced by about 60% to 70%, the lowest point of the drop occurs about 1ms earlier, the overshoot amplitude is reduced by more than 75%, the overall adjustment time is shortened by about 60%, and the adjustment process is smooth without oscillation. The above parameter improvements are all due to the predictive feedforward regulation adopted by the present application. By completing the pre-calculation and disturbance-free switching of control parameters before the load change occurs, the dynamic performance degradation problem caused by response delay in the passive compensation strategy is avoided.

[0095] Table 1: Comparison of key parameters of output voltage dynamic response between this application and prior art under load change conditions. Furthermore, Figure 3A schematic diagram comparing the dynamic response of the output voltage of the prior art and the control method of this application under load change conditions is shown; where the horizontal axis is time t, the vertical axis is the output voltage Vout, the horizontal dashed line represents the rated voltage Vref, and the vertical dashed line represents the moment when the load change occurs, t=3ms.

[0096] Before the load change occurs, the output voltage of both the prior art and this application is stably maintained near the rated voltage, and the response curves basically overlap.

[0097] For the existing technology, specifically the part with the thin solid line, the control strategy is passive compensation. That is, after a sudden load change occurs, the controller detects the output voltage deviation and then initiates the adjustment action. Due to the inherent delay in the detection, calculation, and execution stages, the output voltage drops significantly after the load change, with a drop of approximately 4% to 5% of the rated voltage. The lowest point of the drop occurs about 1.5ms after the change. After that, the controller gradually intervenes to adjust, and the output voltage slowly recovers, accompanied by a significant overshoot phenomenon, with an overshoot of approximately 2% of the rated voltage. The overall adjustment time is approximately 5ms, during which the output voltage oscillates continuously, resulting in poor dynamic performance.

[0098] For this application, specifically the part indicated by the thick solid line, the control strategy is based on an adaptive filtering algorithm to extract and extrapolate the load's power demand in real time. It pre-calculates the joint parameter set before load surges occur and performs seamless switching of control parameters during quiet periods. Therefore, when a load surge actually occurs, the controller is already configured with parameters matching the new load state, eliminating the need to wait for accumulated deviations before initiating adjustment. Figure 3 As shown, the output voltage drop of this application is only about 1% to 1.5% of the rated voltage. The lowest drop point occurs about 0.5ms after the sudden change, and then it quickly and smoothly recovers to near the rated voltage without obvious overshoot. The overall adjustment time is about 2ms, and the dynamic response performance is significantly better than the prior art.

[0099] The above comparison shows that, through predictive feedforward adjustment, this application advances the switching timing of control parameters from after the load change to before the load change, eliminating the voltage drop and oscillation problems caused by response delay in the passive compensation strategy, thereby effectively improving the output voltage stability of the power conversion device under dynamic operating conditions.

[0100] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0101] In one embodiment, a software-reconfigurable digital power management system is provided, which corresponds one-to-one with the software-reconfigurable digital power management method described in the above embodiments. The software-reconfigurable digital power management system includes: The behavior sequence generation module is used to monitor the target load associated with the output terminal of the power conversion device, obtain the real-time voltage and real-time current of the target load, and generate the target load behavior sequence by combining the historical load data. The timing sequence extraction module is used to determine the steady-state operating range and transient switching range of the target load based on the target load behavior sequence, and to extract the voltage change timing sequence and current change timing sequence of the transient switching range. The demand information generation module is used to deduce the next working state of the target load and generate corresponding power demand information based on the voltage change time sequence and the current change time sequence. The power information comparison module is used to obtain the current first control loop parameters of the power conversion device and the current first drive parameters of the power switch, obtain the power output capacity information based on the first control loop parameters and the first drive parameters, and compare the power demand information with the power output capacity information. The joint parameter calculation module is used to synchronously calculate a set of joint parameters based on power demand information when the comparison result does not meet the preset adaptation conditions or when a state switching command is received from the target load. The set of joint parameters includes the second control loop parameters and the second drive parameters. The joint parameter loading module is used to obtain the power switch action sequence and input the joint parameter set into the digital controller based on the preset silent period associated with the power switch action sequence, so as to replace the first control loop parameters and the first drive parameters. The joint parameter correction module is used to correct the joint parameter set based on the voltage difference between the actual output voltage and the target output voltage of the power conversion device, so that the actual output voltage is maintained within the preset allowable range of the target output voltage.

[0102] For specific limitations regarding a software-reconfigurable digital power management system, please refer to the limitations of a software-reconfigurable digital power management method described above, which will not be repeated here. Each module in the aforementioned software-reconfigurable digital power management system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0103] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A software-reconfigurable digital power management method, applied to a power conversion device including a power switch and a digital controller, characterized in that, Including the following steps: The target load associated with the output terminal of the power conversion device is monitored to obtain the real-time voltage and real-time current of the target load, and the target load behavior sequence is generated by combining the historical load data. The steady-state operating range and transient switching range of the target load are determined based on the target load behavior sequence, and the voltage change time sequence and current change time sequence of the transient switching range are extracted. Based on the timing of voltage and current changes, the next operating state of the target load is deduced and the corresponding power demand information is generated. The first control loop parameters of the power conversion device and the first drive parameters of the power switch are obtained. Based on the first control loop parameters and the first drive parameters, the power output capacity information is obtained, and the power demand information is compared with the power output capacity information. When the comparison result does not meet the preset adaptation conditions, or when a state switching command is received from the target load, a joint parameter set is synchronously calculated based on the power demand information. The joint parameter set includes the second control loop parameters and the second drive parameters. Obtain the power switch action sequence, and input the joint parameter set into the digital controller based on the preset silent period associated with the power switch action sequence to replace the first control loop parameters and the first drive parameters. Based on the voltage difference between the actual output voltage and the target output voltage of the power conversion device using a set of combined parameters, the set of combined parameters is corrected to keep the actual output voltage within the preset allowable range of the target output voltage.

2. The software-reconfigurable digital power management method according to claim 1, characterized in that: The steps of monitoring the target load associated with the output terminal of the power conversion device, obtaining the real-time voltage and real-time current of the target load, and generating a target load behavior sequence by combining historical load data include: Collect the real-time voltage and real-time current of the target load over several consecutive switching cycles, and obtain the historical control command sequence of the target load in the corresponding time period; Real-time voltage and real-time current are timestamped and correlated with historical control command sequences to generate a raw data set with associated time sequence information. The real-time voltage and real-time current in the original dataset are decomposed to separate the fundamental components of the real-time voltage and real-time current as well as at least one harmonic component in a specific frequency band. The load electrical feature vector is extracted based on the amplitude, phase, and rate of change of the fundamental and harmonic components, and the frequency of change and entropy of the historical control command sequence within a preset unit time are calculated. The load electrical characteristic vector, change frequency, and command entropy value are combined and encoded in chronological order to generate a target load behavior sequence.

3. The software-reconfigurable digital power management method according to claim 2, characterized in that: The historical control command sequence includes several command codes. The step of extracting the load electrical feature vector based on the amplitude, phase, and rate of change of the fundamental and harmonic components, and calculating the change frequency and command entropy value of the historical control command sequence within a preset unit time includes: Set a sliding window for observation time within a preset unit of time, and count the total number of state jumps in the historical control command sequence within the observation time sliding window as the change frequency. The probability of occurrence of each instruction code within the sliding window of the observation time is statistically analyzed, and the negative summation of the product of the occurrence probability and its logarithm is calculated to obtain the first entropy value. Calculate the Hamming distance between each consecutive adjacent command within the observation time sliding window, and calculate the probability of occurrence of each Hamming distance value in the total number of state transitions; The negative summation of the probability of occurrence and its logarithmic product is calculated to obtain the second entropy value. Then, the first entropy value and the second entropy value are weighted and summed according to the preset weight coefficients to obtain the instruction entropy value.

4. The software-reconfigurable digital power management method according to claim 2, characterized in that: The steps of determining the steady-state operating range and transient switching range of the target load based on the target load behavior sequence, and extracting the voltage change timing and current change timing of the transient switching range, include: Sliding window analysis is performed on the target load behavior sequence to calculate the standard deviation and mean of the load electrical characteristic vector within each time window, and the changes in the historical control command sequence are monitored. When the standard deviation of the load electrical characteristic vector is consistently lower than a preset first threshold within any time window, and its mean remains stable and the historical control command sequence remains unchanged for more than a preset first duration, the interval corresponding to the time window is determined to be a steady-state operating interval. When the standard deviation of the load electrical characteristic vector exceeds the preset second threshold, or the rate of change of its mean exceeds the preset third threshold, or a predefined jump occurs in the historical control command sequence within any time window, the time window is determined to be the starting point of the transient switching interval, and the tracking continues until the end point. The interval between the starting point and the end point is determined to be the transient switching interval. The end point is when the standard deviation of the load electrical characteristic vector falls to the preset fourth threshold and its mean is stable. Within the transient switching interval, the process of real-time voltage and real-time current changing with time is recorded, and the timing points including the rise time of voltage change, the fall time of voltage change, the peak value of current change rate and their corresponding moments are extracted to generate voltage change timing and current change timing.

5. The software-reconfigurable digital power management method according to claim 1, characterized in that: The step of deducing the next operating state of the target load and generating corresponding power demand information based on voltage and current change timing includes: Differential and integral processing is performed on the voltage and current change time series to calculate the instantaneous power change curve, power change rate curve, and load impedance change trend curve of the target load; Based on the power change rate curve and the load impedance change trend curve, an adaptive filtering algorithm is used to extract the main trend component of the change, and the main trend component is pushed forward to the preset prediction time domain to obtain the predicted power trajectory and the predicted impedance trajectory. Based on the predicted power trajectory and the predicted impedance trajectory, the target voltage and target current values ​​when the target load reaches the next steady state are calculated, and the key dynamic constraints in the transition process are extracted. Based on the target voltage value, target current value, and key dynamic constraints, power demand information is generated, which includes the target steady-state operating point, the maximum allowable rate of change of current, the voltage fluctuation window, and the expected transition time.

6. The software-reconfigurable digital power management method according to claim 1, characterized in that: The steps of obtaining the current first control loop parameters of the power conversion device and the current first drive parameters of the power switch, obtaining power output capacity information based on the first control loop parameters and the first drive parameters, and comparing the power demand information with the power output capacity information include: Based on the parameters of the first control loop, calculate the current stable output boundary and small-signal frequency response characteristics of the control loop of the power conversion device. Based on the first driving parameters, calculate the current maximum safe operating boundary of the power switch; By combining the stable output boundary, small-signal frequency response characteristics, and maximum safe operating boundary, the power output capability information is obtained. The power output capability information includes the steady-state adjustable output range, the maximum safe output current, the load step amplitude, and the load step rate.

7. The software-reconfigurable digital power management method according to claim 1, characterized in that: The step of synchronously calculating a joint parameter set based on power demand information when the comparison result does not meet the preset adaptation conditions, or when a state switching command is received from the target load, includes: Using power demand information as joint constraints, system dynamic equations are established, including the power stage topology of the power conversion device, the output filter model, and the equivalent model of the target load. Based on the system dynamic equations, and with the phase margin, maximum allowable bandwidth, and peak output impedance of the control loop of the power conversion device as optimization objectives, the parameters of the second control loop are obtained in parallel under joint constraints. The second driving parameters are obtained by solving the problem under joint constraints, with the transient current stress, switching loss and conduction loss of the power switch as the internal optimization objectives. The parameters of the second control loop and the second drive parameters are encapsulated and associated to generate a joint parameter set.

8. The software-reconfigurable digital power management method according to claim 1, characterized in that: The step of obtaining the power switch action sequence and, based on a preset silent period associated with the power switch action sequence, inputting a set of joint parameters into the digital controller to replace the first control loop parameters and the first drive parameters includes: Monitor the pulse drive signal output by the digital controller used to control the power switch to obtain the power switch action sequence; Based on the power switch action sequence, determine the silent period in each switching cycle when all power switches are in the off state, and preload the joint parameter set into the parameter buffer of the digital controller before the silent period begins. At the start of the silent period, the set of joint parameters in the parameter buffer is synchronously updated to the running register group of the digital controller, so that the second control loop parameters and the second drive parameters replace the first control loop parameters and the first drive parameters.

9. The software-reconfigurable digital power management method according to claim 1, characterized in that: The step of correcting the joint parameter set based on the voltage difference between the actual output voltage and the target output voltage of the power conversion device, so that the actual output voltage is maintained within the preset allowable range of the target output voltage, includes: (a) Continuously monitor the actual output voltage of the power conversion device based on the joint parameter set, and calculate the real-time voltage difference and its changing trend between the actual output voltage and the target output voltage; (b) Based on the real-time voltage difference, perform integral compensation on the second control loop parameters in the joint parameter set to generate preliminary correction parameters; (c) Based on the real-time voltage difference change trend, the preliminary correction parameters are adjusted by feedforward to generate the final correction parameters; (d) Update the final corrected parameters to the run register set of the digital controller to replace the second control loop parameters; (e) Repeat steps (a), (b), (c), and (d) in sequence until the real-time voltage difference remains stable within the preset allowable range.

10. A software-reconfigurable digital power management system, characterized in that, include: The behavior sequence generation module is used to monitor the target load associated with the output terminal of the power conversion device, obtain the real-time voltage and real-time current of the target load, and generate the target load behavior sequence by combining the historical load data. The timing sequence extraction module is used to determine the steady-state operating range and transient switching range of the target load based on the target load behavior sequence, and to extract the voltage change timing sequence and current change timing sequence of the transient switching range. The demand information generation module is used to deduce the next working state of the target load and generate corresponding power demand information based on the voltage change time sequence and the current change time sequence. The power information comparison module is used to obtain the current first control loop parameters of the power conversion device and the current first drive parameters of the power switch, obtain the power output capacity information based on the first control loop parameters and the first drive parameters, and compare the power demand information with the power output capacity information. The joint parameter calculation module is used to synchronously calculate a set of joint parameters based on power demand information when the comparison result does not meet the preset adaptation conditions or when a state switching command is received from the target load. The set of joint parameters includes the second control loop parameters and the second drive parameters. The joint parameter loading module is used to obtain the power switch action sequence and input the joint parameter set into the digital controller based on the preset silent period associated with the power switch action sequence, so as to replace the first control loop parameters and the first drive parameters. The joint parameter correction module is used to correct the joint parameter set based on the voltage difference between the actual output voltage and the target output voltage of the power conversion device, so that the actual output voltage is maintained within the preset allowable range of the target output voltage.

Images