Inverter module supporting hot plug and power supply control method thereof

Abstract

This invention relates to the field of hot-swappable inverter power supply control technology, and discloses a hot-swappable inverter power supply module and its power supply control method. The method includes, upon module connection, analyzing the changing patterns of initial electrical disturbance signals to identify the severity of the connection and mapping it to an initial intensity label. Based on this label, the bus energy flow spectrum within the relevant time period is extracted, and the energy extraction lag characteristics of each load branch are analyzed to form a response list with time-series markers. The list is compared with the system power scheduling cycle template to calculate the time-series offset, and an adaptive power compensation time window is constructed for load branches with excessive offsets. Based on this time window, the energy injection timing of the corresponding branch is reshaped in subsequent power allocation, generating and executing the inverter module control strategy. This method enhances the dynamic stability and power supply quality of the system through real-time perception of the intensity of hot-swappable events and adaptive time-series compensation for load responses.

Description

Technical Field

[0001] This invention relates to the field of hot-swappable inverter power supply control technology, specifically to a hot-swappable inverter power supply module and its power supply control method. Background Technology

[0002] In inverter power supply systems that support hot-swapping of modules, the charging of the capacitor at the input end of a new module will generate electrical disturbances on the bus when it is connected to the system bus. Existing control methods typically determine whether a hot-swapping event has occurred by detecting a drop in bus voltage or a sudden change in current, and then immediately activate a predefined buffer start-up or current limiting strategy. These methods treat all connection events as homogeneous shocks, using fixed timing and parameters for power compensation and load adjustment, failing to distinguish the actual differences in the impact of different physical connection behaviors on the system.

[0003] Existing compensation strategies primarily focus on stabilizing bus voltage or limiting inrush current amplitude, and their power dispatch is often based on fixed periodic templates. However, disturbances caused by different hot-plug events can lead to varying response times and recovery curves in downstream loads. A mismatch exists between the fixed compensation timing and the delayed response of the actual energy demand of the loads. This timing misalignment can cause secondary power oscillations or stability issues in multi-load, periodically dispatched systems, limiting the dynamic performance and reliability of the system under frequent hot-plug conditions. A control method is needed that can sense the intensity of connection events and adaptively adjust the power allocation timing based on the actual dynamic response characteristics of the loads. Summary of the Invention

[0004] The purpose of this invention is to provide an inverter power supply module that supports hot-swapping and a power supply control method thereof, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a power supply control method for a hot-swappable inverter power supply module, the method comprising:

[0006] When a physical connection is detected between the inverter power module and the system bus, the initial electrical disturbance signal caused by the connection action is captured. The severity of the connection behavior is identified based on the changing pattern of the initial electrical disturbance signal, and the severity is mapped to the initial intensity label of the hot-plug event.

[0007] Based on the initial strength label, the bus energy flow spectrum within a preset time period before and after the hot-swap connection event is traced back. Power fluctuation segments that are causally related to the initial strength label in time are extracted from the spectrum, and the energy absorption lag characteristics of each load branch within the power fluctuation segment are analyzed to form a load response lag list with time sequence marking.

[0008] The load response lag list is compared with the system's preset power scheduling cycle template. The offset between the lag time and the scheduling cycle boundary is calculated. Load branches with offsets exceeding the tolerance are selected. Based on the duration and frequency of their lag characteristics, an adaptive power compensation time window is constructed for the load branches with offsets exceeding the tolerance.

[0009] Based on the adaptive power compensation time window, in the subsequent power distribution cycle, the energy injection process of the corresponding load branch is reshaped according to a preset timing, generating an inverter module control strategy that includes the reshaped power release trajectory, and driving the power switching devices to turn on and off according to the inverter module control strategy.

[0010] Preferably, the step of forming a load response hysteresis list with time-stamped tags includes:

[0011] Read the bus voltage and total current sampling data corresponding to the time node of the initial intensity label, and draw a dynamic energy spectrum centered on the time node;

[0012] In the dynamic energy spectrum, the power dip region caused by module access is located, and the depth value of the power dip and the time taken to recover to steady state are recorded. The product of the depth value and the recovery time is quantified as the total disturbance energy.

[0013] Based on the magnitude of the total disturbance energy, trace downstream in the dynamic energy spectrum to identify the establishment process of the current in each parallel load branch, and mark the time taken for the current in each load branch to rise from the beginning to reach the steady state value, as the inherent response delay of the load branch.

[0014] The inherent response delays of all load branches are summarized and arranged in the order in which the delays occur. The corresponding load branch identifiers are associated with the total disturbance energy to generate a load response hysteresis list with time-series tags.

[0015] Preferably, the step of constructing the adaptive power compensation time window includes:

[0016] Obtain the load response hysteresis list with time-series markers, and extract the inherent response delay data for each record in the list;

[0017] The system calls a preset power scheduling cycle template, which defines the theoretical start and end times of energy injection for each load branch in a complete power allocation cycle.

[0018] The inherent response delay data is compared with the corresponding theoretical cutoff time in the power scheduling cycle template, and the time difference between the actual energy establishment completion point and the theoretical cutoff point is calculated and recorded as the scheduling offset.

[0019] Set a scheduling offset threshold, filter out load branch records whose scheduling offset exceeds the offset threshold, and determine that the load branch whose scheduling offset exceeds the threshold has a situation where power distribution has not fully converged;

[0020] For each load branch with incomplete convergence, the duration of its inherent response delay and the frequency of its occurrence in multiple consecutive hot-plug events are analyzed. Based on the duration and frequency, an additional energy replenishment period independent of the standard scheduling cycle is calculated, which is the adaptive power compensation time window constructed for the load branch.

[0021] Preferably, the step of generating an inverter module control strategy including the reshaped power release trajectory includes:

[0022] Based on the adaptive power compensation time window, the standard power scheduling cycle template is locally modified. After the theoretical energy injection cutoff time of the corresponding load branch, the adaptive power compensation time window is inserted to form the modified scheduling sequence diagram.

[0023] In the revised scheduling timing diagram, a specific power compensation waveform is defined for each inserted adaptive power compensation time window. The power compensation waveform starts at the original theoretical cutoff time and ends at the end of the compensation time window. Its amplitude is dynamically calculated based on the energy gap of the load branch in the most recent hot-plug event.

[0024] The defined power replenishment waveform is spliced ​​and smoothly transitioned with the original power release waveform within the standard scheduling cycle to ensure that the entire energy injection process is continuous in time and without abrupt changes in amplitude, forming a reshaped power release trajectory for each load branch.

[0025] The power release trajectories of all load branches are integrated, and the power release trajectories of the load branches are converted into pulse width modulation signal sequences of the corresponding power switching devices according to the topology of the inverter power supply module. The pulse width modulation signal sequence is the control strategy of the inverter module.

[0026] Preferably, the method further includes a step of performing closed-loop verification of the power compensation process:

[0027] During the execution of the inverter module control strategy, the current feedback signal of the corresponding load branch within the adaptive power compensation time window is collected in real time.

[0028] The collected current feedback signal is compared point by point with the expected current waveform in the reshaped power release trajectory, and the real-time error between the two in terms of amplitude and trend is calculated.

[0029] If the real-time error continues to exceed the preset allowable range, the amplitude of the currently executing power compensation waveform or the duration of the adaptive power compensation time window will be dynamically adjusted to form a compensation parameter adjustment command.

[0030] Based on the compensation parameter adjustment instruction, the pulse width modulation signal of the corresponding segment in the inverter module control strategy is updated online, and the adjusted parameters are recorded in the historical compensation feature library of the load branch for optimizing the construction of the subsequent adaptive power compensation time window.

[0031] Preferably, the step of constructing the historical compensation feature library and using it for optimization includes:

[0032] An independent compensation record set is established for each load branch to store the amplitude adjustment and duration adjustment included in each compensation parameter adjustment instruction;

[0033] When the adaptive power compensation time window is constructed again for the same load branch, the corresponding compensation record set is queried, and the statistical characteristics of the most recent adjustment amounts are extracted, including the mean and trend of the adjustment amounts.

[0034] The extracted statistical features are used as prior knowledge and applied to the initial calculation of the amplitude of the power compensation waveform and the initial duration setting of the adaptive power compensation time window, so that the initial compensation parameters are closer to the actual needs and the adjustment range and number of times in the subsequent closed-loop verification are reduced.

[0035] Preferably, the step of capturing the initial electrical disturbance signal caused by the connection action further includes a step of identifying the interference signal:

[0036] Upon detecting a physical connection, a high-frequency sampling channel is activated to continuously acquire the voltage differential signal at the bus connection point.

[0037] Morphological analysis was performed on the acquired voltage differential signal to distinguish between brief high-frequency oscillations caused by mechanical contact bounce and continuous unidirectional changes caused by the charging of the internal capacitor of the module.

[0038] Only the continuous unidirectional change signal caused by the charging of the internal capacitor of the module is used as the effective initial electrical disturbance signal for subsequent severity identification, while the oscillation signal caused by mechanical contact bounce is filtered out.

[0039] Preferably, the step of identifying the intensity of the connection behavior specifically includes:

[0040] The effective initial electrical disturbance signal is integrated in the time domain to calculate its accumulated energy value over a preset time period.

[0041] Multiple energy threshold ranges are set, each range corresponding to a level of connection intensity;

[0042] The calculated cumulative energy value is assigned to the corresponding energy threshold interval, thereby mapping the continuous signal energy value to the initial intensity label of the discrete hot-plug event.

[0043] Preferably, the method further includes a conflict resolution step before generating the inverter module control strategy:

[0044] Check whether the multiple adaptive power compensation time windows constructed for different load branches overlap on the time axis;

[0045] If there are overlapping power compensation time windows, calculate the total amount of supplementary power required by each branch during the overlapping period, and determine whether the total amount exceeds the maximum remaining capacity that the inverter power supply module can provide under the current operating conditions.

[0046] If the limit is exceeded, the power compensation waveform amplitude of each branch during the overlapping period will be reduced proportionally according to the priority of each branch or the urgency of the energy gap, or the start time of some power compensation time windows will be finely adjusted to avoid overlap, so as to ensure that the total power demand is within the module capacity limit.

[0047] After the conflict resolution step is completed, the final version of the inverter module control strategy is generated and compiled into executable code, which is then loaded into the digital signal processor of the inverter power supply module. The digital signal processor generates precise pulse width modulation signals based on the code to control the operation of the power switching devices, thereby realizing hot-swappable power supply control.

[0048] Preferably, the present invention further includes a hot-swappable inverter power supply module, the module including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the power supply control method for the hot-swappable inverter power supply module as described above.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] By capturing the initial electrical disturbance signal at the moment of connection and analyzing its changing patterns, the mechanical behavior of the physical connection is quantified into an initial intensity label characterizing the impact severity. This process achieves refined perception of hot-plug events from "presence" to "strength." Subsequent analysis based on this intensity label enables the system to activate corresponding levels of response plans according to the actual impact severity of the event. This provides a precise decision-making basis for implementing differentiated preprocessing and compensation controls that match the severity of the event, avoiding efficiency losses due to overcompensation or risks caused by undercompensation.

[0051] By retrospectively analyzing the bus energy flow map under a specific intensity of disturbance, the energy extraction lag characteristics of each load branch after the event can be accurately extracted, forming a response list with clear timing markers. Comparing this list with the system's inherent power scheduling cycle template identifies load branches with non-negligible offsets between response lag and scheduling timing. By constructing adaptive power compensation time windows for these branches, their power injection timing is reshaped in subsequent scheduling cycles. This technique enables the system to proactively and specifically correct timing deviations caused by disturbances, resynchronizing asynchronous load responses to the system's global scheduling rhythm, suppressing the risk of system oscillations caused by timing mismatches, and improving the stability and efficiency of multi-load collaborative operation. Attached Figure Description

[0052] Figure 1 This is a schematic diagram illustrating the working principle of the hot-swappable inverter power supply module and its power supply control method described in this invention.

[0053] Figure 2 A flowchart for creating a time-stamped list of load response hysteresis;

[0054] Figure 3 A flowchart for generating the inverter module control strategy;

[0055] Figure 4 The error curve for closed-loop verification of the adaptive power compensation process;

[0056] Figure 5 A comparison chart of total power demand before and after conflict resolution within the power compensation time window. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Please see Figure 1This invention provides an inverter power supply module supporting hot-swapping and its power supply control method. The method includes: when the inverter power supply module physically connects to the system bus, the system detects the connection action and captures the resulting initial electrical disturbance signal. The system analyzes the changing pattern of the initial electrical disturbance signal, identifies the severity of the connection behavior based on its morphological characteristics, and maps the severity to an initial intensity label representing the severity level of the hot-swapping event. Based on the initial intensity label, the system traces back the bus energy flow spectrum within a preset time period before and after the hot-swapping connection event, and extracts power fluctuation segments that are temporally causally related to the initial intensity label from the spectrum. The system further analyzes the energy absorption lag characteristics of each load branch within this power fluctuation segment, i.e., the delay in load current establishment, and finally forms a load response lag list with time-series markers. The system compares this list with a preset power scheduling cycle template, calculates the offset between the actual response lag time of each load branch and the theoretical boundary of the scheduling cycle, and filters out load branches whose offset exceeds the allowable tolerance. For these selected load branches, the system constructs an adaptive power compensation time window for each branch, taking into account the duration and historical frequency of its hysteresis characteristics. Based on this adaptive power compensation time window, in subsequent periodic power distribution cycles, the system performs a preset timing reshaping of the energy injection process for the corresponding load branch, generating an inverter module control strategy that includes the precise power release trajectory after reshaping. According to this inverter module control strategy, the system generates corresponding drive signals to precisely control the on and off timing of the power switching devices in the inverter, thereby completing adaptive energy management for hot-plug events.

[0059] Example 1: See Figure 2The system reads the bus voltage and total current sampling data at the time node corresponding to the initial intensity label, and plots an energy map centered on that time node, covering the dynamic changes before and after the event. In this dynamic energy map, the system locates the bus power dip region caused by the inverter power module connection, records the depth of the power dip and the total time it takes for it to recover to the steady-state power level, and multiplies the depth value by the recovery time to quantify the total energy involved in this disturbance. Based on the magnitude of this total disturbance energy, the system traces downstream to the load side in the dynamic energy map, identifies the current establishment process of each parallel load branch, and marks the specific time taken for each load branch current to rise from the start to reach the steady-state rated value. This time is recorded as the inherent response delay of that load branch. The system summarizes the inherent response delay data of all load branches, arranges them in chronological order of the delay occurrence, and associates each delay data with its corresponding load branch identifier and the calculated total disturbance energy, thereby generating a load response hysteresis list with time-series markings. The system retrieves the load response lag list and extracts the inherent response delay data for each record. The system then calls a preset power scheduling cycle template, which explicitly defines the theoretical start and end times of energy injection for each load branch within a complete standard power allocation cycle. The system compares the extracted inherent response delay data with the theoretical energy injection end time of the corresponding load branch in the power scheduling cycle template, calculating the time difference between the actual energy establishment completion point and the theoretical end point. This difference is recorded as the scheduling offset. The system sets a scheduling offset threshold, compares the calculated scheduling offset with the threshold, and filters out load branch records with scheduling offsets exceeding the threshold, determining that these branches experienced a power allocation failure to fully converge within the theoretical cycle during the event. For each load branch that is determined to be incompletely converged, the system analyzes the duration of its inherent response delay and counts the frequency of this lag phenomenon in multiple consecutive hot-plug events. Based on this duration and frequency, the system calculates an additional energy replenishment period that is independent of the standard scheduling cycle. This period is the adaptive power compensation time window that the system constructs for the load branch.

[0060] In practical implementation, after capturing the initial intensity tag generated by the hot-plug event, the system immediately reads high-frequency sampling data of the bus voltage and total current within a specific range before and after the time node corresponding to the initial intensity tag. These data are aligned with the time node corresponding to the initial intensity tag as the zero point. The system uses the aligned data to draw a complete dynamic energy map covering the steady state before the event, the transient state during the event, and the recovery period after the event. The horizontal axis of the dynamic energy map represents time, and the vertical axis represents the instantaneous power of the bus. The instantaneous power is obtained by multiplying the synchronously sampled voltage and current values. In the dynamic energy map, the system uses an algorithm to locate the region where the bus power significantly decreases due to the inverter power module connection. This region is defined as the power dip region. The system records the lowest power value of the power dip region as the depth value and records the complete time it takes for the bus power to recover from the dip start point to the pre-event steady-state power level as the recovery duration. The system multiplies the depth value by the recovery duration, and the product is quantified as the total disturbance energy caused by this hot-plug event. The formula for calculating the total disturbance energy is:

[0061] in: This represents the total amount of disturbance energy. This represents the steady-state power value before the event. Represents the dynamic energy spectrum The actual power value at that moment, and These represent the start time of the power dip region and the time when the power recovers to its original value. At that moment.

[0062] In some embodiments, based on the calculated total disturbance energy, the system traces downstream to the load side in the dynamic energy map to identify all load branches connected in parallel to the bus. The system synchronously retrieves independent current sampling data for each load branch within the same time window and analyzes the current build-up process of each load branch after a power trough occurs. The system marks the precise time interval taken for each load branch current to reach its steady-state rated current value from the start of its rise; this time interval is defined as the inherent response delay of the corresponding load branch. The system summarizes the inherent response delay data of all identified load branches and sorts all records according to the chronological order of their occurrence. In each sorted record, the system associates its corresponding unique load branch identifier with the previously calculated total disturbance energy value, ultimately generating a structured, time-stamped list of load response hysteresis. This time-stamped list of load response hysteresis serves as direct input for subsequent analysis.

[0063] In practical implementation, after obtaining a list of load response hysteresis records with time-series tags, the system extracts inherent response delay data line by line. Simultaneously, the system calls a power scheduling cycle template pre-stored in the controller. This template defines the theoretical energy injection start and end times for each load branch within a standardized, periodic power allocation cycle. The system compares the extracted inherent response delay data with the theoretical energy injection end times for the corresponding load branches in the power scheduling cycle template, calculating the time difference between the actual energy establishment completion time and the theoretical energy injection end time. This difference is recorded as the scheduling offset. The system sets a preset scheduling offset threshold and compares all calculated scheduling offsets with this threshold, filtering out load branch records where the scheduling offset exceeds the threshold. These load branches are considered to have failed to fully converge power allocation within the theoretical cycle during this hot-plug event. Optionally, the scheduling offset threshold can be configured according to the overall system stability requirements.

[0064] In some embodiments, for each load branch determined to have incomplete convergence, the system analyzes the duration of its inherent response delay and queries historical event records to count the frequency of the inherent response delay characteristic of that load branch in multiple consecutive hot-plug events. Based on the duration and frequency of the inherent response delay, the system calculates an additional time period according to predefined rules. This time period is independent of the standard power scheduling cycle and is specifically used for energy replenishment of the load branch. This calculated additional time period is the adaptive power compensation time window constructed by the system for that load branch. The construction logic of the adaptive power compensation time window makes the compensation more targeted.

[0065] Example 2: See Figure 3The system locally modifies the standard power scheduling cycle template based on the adaptive power compensation time windows constructed for each load branch. The modification operation involves inserting the calculated adaptive power compensation time window for the corresponding load branch after its theoretical energy injection cutoff time, thus forming a new, modified scheduling sequence diagram that includes the compensation period. In this modified scheduling sequence diagram, the system defines a specific power compensation waveform for each inserted adaptive power compensation time window. This waveform starts at the original theoretical energy injection cutoff time and ends at the end of the compensation time window, with its amplitude dynamically calculated based on the energy gap measured in the most recent hot-plug event for that load branch. The system then splices the defined power compensation waveform with the original power release waveform within the standard scheduling cycle and performs smooth transition processing at the splicing points to ensure that the entire energy injection process is continuous in the time dimension and without abrupt changes in power amplitude, thereby forming a reshaped power release trajectory for each load branch. The system integrates the reshaped power release trajectories of all load branches and, based on the specific topology and control principle of the inverter power supply module, converts the power release trajectory of each path into a pulse width modulation signal sequence that drives the corresponding power switching device. This complete pulse width modulation signal sequence constitutes the final executable inverter module control strategy.

[0066] In implementation, the system performs a local correction operation on the pre-stored power scheduling cycle template based on the adaptive power compensation time window constructed for each load branch. This correction operation is exemplified for the load branch with the load branch identifier L3. The theoretical energy injection cutoff time for the load branch with the load branch identifier L3 in the power scheduling cycle template is recorded as T_end_L3, and the duration of the adaptive power compensation time window constructed for the load branch with the load branch identifier L3 is recorded as T_comp_L3. After the theoretical energy injection cutoff time T_end_L3 for the load branch with the load branch identifier L3, the system inserts a time period of duration T_comp_L3. This time period is the adaptive power compensation time window corresponding to the load branch with the load branch identifier L3. After performing the same operation on all load branches requiring compensation, the system forms a corrected scheduling sequence diagram containing multiple compensation time periods. It can be understood that the scheduling sequence diagram is a mapping relationship between the time axis and the expected power state of each load branch.

[0067] In some embodiments, in the modified scheduling timing diagram, the system defines a specific power compensation waveform for each inserted adaptive power compensation time window. For the load branch with load branch identifier L3, its power compensation waveform starts at the original theoretical energy injection cutoff time T_end_L3 and ends at the end time of the compensation time window T_end_L3+T_comp_L3. The amplitude of the power compensation waveform is dynamically calculated by the system based on the energy gap of the load branch with load branch identifier L3 in the most recent hot-plug event. The energy gap is obtained by comparing the difference between the steady-state current value I_steady_L3 and the actual current value I_actual_L3 of the load branch with load branch identifier L3 after the event and combining it with the bus voltage. The formula for calculating the amplitude of the power compensation waveform is:

[0068]

[0069] in: This represents the power compensation waveform amplitude calculated for the load branch with load branch identifier L3. I_steady_L3 represents the steady-state current value of the load branch with load branch identifier L3. I_actual_L3 represents the actual current value I_actual_L3 reached by the load branch with load branch identifier L3 after the event. This represents the nominal value of the bus voltage. The amplitude adjustment coefficient is preset for the system. The specific shape of the power compensation waveform can be rectangular, ramp, or a specific curve, preset by the system according to the load characteristics.

[0070] In implementation, the system splices the predefined power compensation waveform with the original power release waveform within the standard power scheduling cycle. For the load branch with load branch identifier L3, the splicing point is located at the theoretical energy injection cutoff time T_end_L3. The system applies a smooth transition algorithm to the splicing point. This algorithm introduces a short time interval before and after the splicing point, within which the waveform amplitude changes according to a specific function, ensuring that the entire energy injection process is continuous in the time dimension and has no abrupt changes in power amplitude. The complete waveform after smooth transition processing is the reshaped power release trajectory for the load branch with load branch identifier L3. The system performs the same waveform splicing and smooth transition processing on all load branches that need compensation, generating the reshaped power release trajectory for each load branch. Optionally, the smooth transition algorithm can be implemented using linear interpolation, a sine function, or a polynomial function.

[0071] In some embodiments, the system integrates the reshaped power release trajectories of all load branches to form a multidimensional power-time command set. Based on the full-bridge topology of the inverter power supply module, the system converts this multidimensional power-time command set into a pulse-width modulation (PWM) signal sequence to drive the corresponding power switching devices. The conversion process, based on the space vector pulse-width modulation (SVM) principle, maps the sum of power demands of each load branch at each moment to the voltage vector of the inverter bridge arm, and then decomposes it into the on and off times of the corresponding upper and lower bridge arm power switching devices. These time-sequential on and off time commands constitute a complete and executable inverter module control strategy. It can be understood that the inverter module control strategy ultimately manifests as a series of precisely timed logic level signals.

[0072] Example 3: During the execution of the inverter module control strategy, the system collects the current feedback signal of the corresponding load branch within the adaptive power compensation time window in real time. The system compares the collected real-time current feedback signal with the expected current waveform of that period in the reshaped power release trajectory point by point, and calculates the real-time error in amplitude and trend. If the calculated real-time error continues to exceed the preset allowable range, the system dynamically adjusts the amplitude of the currently executing power compensation waveform or adjusts the duration of the adaptive power compensation time window, forming a compensation parameter adjustment command. Based on the compensation parameter adjustment command, the system updates the pulse width modulation signal parameters of the corresponding period in the inverter module control strategy online, and records the amplitude adjustment and duration adjustment amount in the dedicated historical compensation feature library for that load branch. When the system constructs an adaptive power compensation time window for the same load branch again, it queries the corresponding historical compensation feature library to extract the statistical characteristics of the most recent adjustments, including the mean and trend of the adjustments. The system uses the extracted statistical features as prior knowledge and applies them to the initial amplitude calculation of the power compensation waveform and the initial duration setting of the adaptive power compensation time window, so that the initial compensation parameters are closer to the actual requirements.

[0073] In practical implementation, during the execution of the inverter module control strategy, within the adaptive power compensation time window defined for the load branch identified as L3, the system uses a current sensor on the load branch identified as L3 to collect the current feedback signal of the load branch identified as L3 in real time. The system then compares the waveform of the collected real-time current feedback signal with the expected current waveform in the reshaped power release trajectory generated for the load branch identified as L3 within the adaptive power compensation time window, comparing the current amplitude at the sampling points at the same time. It can be understood that the current feedback signal directly reflects the actual response. The system calculates the real-time error between the real-time current feedback signal and the expected current waveform in terms of amplitude and trend. The real-time error includes the instantaneous amplitude difference and the root mean square error within a time window. In practical implementation, if the calculated real-time error continuously exceeds the preset allowable range for more than three control cycles, the system dynamically adjusts the amplitude of the power compensation waveform currently being executed for the load branch with load branch identifier L3, or adjusts the duration of the adaptive power compensation time window constructed for the load branch with load branch identifier L3, forming a compensation parameter adjustment instruction containing the load branch identifier, adjustment type, and adjustment amount. Based on the compensation parameter adjustment instruction, the system updates the pulse width modulation signal parameters of the corresponding time period and the corresponding load branch with load branch identifier L3 in the inverter module control strategy online. The system records the amplitude adjustment amount and duration adjustment amount included in this compensation parameter adjustment instruction to the dedicated historical compensation feature library for the load branch with load branch identifier L3. Optionally, the allowable range can be set to ±5% of the steady-state current value according to the system's current accuracy requirements.

[0074] In some embodiments, the system establishes an independent compensation record set for each load branch. This compensation record set serves as the storage structure for a historical compensation feature library, storing the amplitude and duration adjustments included in each compensation parameter adjustment instruction in chronological order. When the system constructs an adaptive power compensation time window for the same load branch with identifier L3, it queries the compensation record set corresponding to that load branch. The system extracts statistical features from the compensation record set for the most recent adjustments. These features include the mean and trend of the last five amplitude adjustments, and the mean and trend of the last five duration adjustments. The system uses these extracted statistical features as prior knowledge, applying them to the initial amplitude calculation of the power compensation waveform for the load branch with identifier L3 and the initial duration setting of the adaptive power compensation time window for that load branch. The calculation method involves superimposing 70% of the historical adjustment mean onto the base calculated value. In specific implementations, the process of optimizing the initial compensation parameters based on historical records can shorten the adjustment time required for closed-loop verification. The system generates an initial power compensation waveform and an initial adaptive power compensation time window for the load branch with load branch identifier L3 based on the superimposed parameters. This makes the initial compensation parameters closer to the actual requirements of the load branch with load branch identifier L3, which may reduce the adjustment magnitude and number of adjustments in subsequent closed-loop verification. Optionally, the specific number of recent adjustments can be configured to five or ten, depending on the system's reliance on historical data.

[0075] Example 4: Upon detecting a physical connection between the inverter power module and the system bus, the system activates a high-frequency sampling channel to continuously acquire the voltage differential signal at the bus connection point. The system performs morphological analysis on the acquired voltage differential signal to distinguish between brief high-frequency oscillations caused by mechanical contact bounce and continuous unidirectional changes caused by capacitor charging within the module. The system only uses the identified continuous unidirectional changes caused by capacitor charging within the module as a valid initial electrical disturbance signal for subsequent severity identification, while filtering out the oscillation signal caused by mechanical contact bounce. The system performs time-domain integration on the valid initial electrical disturbance signal to calculate its accumulated energy value within a preset time window. The system pre-sets multiple energy threshold intervals, each corresponding to a connection severity level. The system assigns the calculated accumulated energy value to the corresponding energy threshold interval, thus mapping continuous energy values ​​to an initial intensity label for a discrete hot-plug event.

[0076] In practical implementation, when the system detects a physical connection event between the inverter power module and the system bus, it simultaneously activates an independent high-frequency sampling channel. This channel continuously acquires the instantaneous voltage value sequence at the bus connection point at a sampling rate of no less than 1MHz. The system calculates the voltage deviation signal from this sequence. The voltage deviation signal is obtained by subtracting the system's steady-state voltage reference value from the real-time instantaneous voltage value at the bus connection point. This voltage deviation signal reflects the bus voltage offset caused by the connection event. The system performs morphological analysis on the voltage deviation signal, including identifying the oscillation frequency, amplitude envelope, and polarity maintenance characteristics of the signal waveform. Based on the morphological analysis results, the system distinguishes between brief high-frequency oscillations caused by mechanical contact bounce and continuous unidirectional changes caused by capacitor charging within the module. Brief high-frequency oscillations caused by mechanical contact bounce are characterized by rapid fluctuations around zero within milliseconds, with amplitude decay; continuous unidirectional changes caused by capacitor charging within the module are characterized by a waveform where the voltage deviation signal maintains a single polarity and its amplitude decays slowly over a timescale of tens of milliseconds. The system establishes a discrimination rule based on duration and polarity preservation. Only signal segments with a duration exceeding a preset threshold and whose waveforms conform to continuous unidirectional change characteristics are considered valid initial electrical disturbance signals for subsequent severity identification. Signal segments with short durations and conforming to brief high-frequency oscillation characteristics are filtered out as interference signals. It can be understood that the core of the discrimination rule is a joint criterion of time scale and waveform characteristics.

[0077] In some embodiments, the system performs a time-domain integration operation on the valid initial electrical disturbance signal to calculate the cumulative absolute energy value of the valid initial electrical disturbance signal within a preset time window T_w. The cumulative absolute energy value is obtained by integrating the absolute value of the valid initial electrical disturbance signal, and its calculation formula is as follows:

[0078]

[0079] in: This represents the calculated cumulative absolute energy value, and its unit is volt-second (V·s). The value of the effective initial electrical disturbance signal at time t (i.e., voltage deviation). This represents the preset integration time window length, for example, 20 milliseconds. The system has multiple preset discrete absolute energy accumulation threshold ranges, each corresponding to a level of connectivity severity. The mapping relationship between the absolute energy accumulation threshold ranges and connectivity severity levels is defined in the system configuration file, see Table 1:

[0080] Table 1: Mapping Table of Absolute Energy Accumulation Range and Connection Severity Level

[0081]

[0082] In practical implementation, the system will calculate the cumulative absolute energy value. The absolute energy accumulation value is compared with a predefined threshold range. The values ​​are categorized into the corresponding absolute energy accumulation threshold range. Based on the mapping table, the system assigns consecutive absolute energy accumulation values... The numerical value is mapped to a discrete symbolic label, which serves as the initial intensity label for the hot-plug event. This initial intensity label serves as the starting point for backtracking analysis and strategy adjustments in subsequent steps. Optionally, the boundary values ​​of the absolute energy accumulation threshold range can be adjusted through the system calibration procedure to accommodate different bus impedances and module input capacitance parameters.

[0083] In some embodiments, the opening and closing of the high-frequency sampling channel is triggered by a digital signal emitted by the connection detection circuit, ensuring that the acquisition window accurately covers the physical process of the connection action. Optionally, morphological analysis can employ wavelet transform to more accurately distinguish high-frequency oscillations and low-frequency trend components by analyzing the signal components at different scales. The polarity of the continuously unidirectional changing signal caused by the charging of the internal capacitors of the module depends on the relative levels of the bus voltage and the initial voltage of the module when it is connected.

[0084] See Figure 4 In the design and verification of the adaptive power compensation time window, this figure presents the closed-loop verification error characteristics of the compensation process. Specifically, the figure uses compensation time (in ms) as the horizontal axis and current error (in A) as the vertical axis, simultaneously showing the changes in the error before adjustment, the error after adjustment, and the upper and lower limits of the allowable error (±0.25A). In the actual verification process, the error before adjustment (red line) shows significant fluctuations, with its amplitude exceeding the allowable error range multiple times, reflecting the instability of power distribution without adaptive compensation. The error after adjustment (green line), through timing reshaping and dynamic parameter adjustment of the adaptive power compensation time window, has its fluctuation amplitude effectively constrained within the allowable error range, and the overall trend is towards convergence. This result verifies the effectiveness of the adaptive power compensation time window: by timing reshaping and closed-loop parameter adjustment of the load branch energy injection process, the real-time error between the current feedback signal and the expected power release trajectory is controlled within the preset tolerance, achieving stable convergence of power distribution in hot-swappable scenarios. The time dimension characteristics of the curve (26ms to 34ms) correspond to the dynamic response stage of the compensation process. The rapid convergence of the adjusted error also reflects the optimization effect of the historical compensation feature library on the initial compensation parameters. By introducing prior knowledge, the adjustment range and number of adjustments in the closed-loop verification are reduced, and the real-time performance of the compensation strategy is improved.

[0085] Example 5: The system checks whether the multiple adaptive power compensation time windows constructed for different load branches overlap on the system time axis. If overlapping power compensation time windows exist, the system calculates the total supplementary power required by all relevant branches during the overlapping period. The system determines whether this total power exceeds the maximum remaining capacity that the inverter module can provide under the current operating conditions. If the total demand power exceeds the module capacity limit, the system proportionally reduces the amplitude of the power compensation waveform of each branch during the overlapping period, or fine-tunes the start time of some power compensation time windows to avoid overlap, based on the preset priority of each branch or the urgency of its energy gap, thereby ensuring that the total demand power is within the module capacity limit. After the conflict resolution step is completed, the system generates the final version of the inverter module control strategy and compiles it into executable code, which is then loaded into the digital signal processor of the inverter module. The digital signal processor generates precise pulse width modulation signals based on this code to control the operation of the power switching devices.

[0086] In practical implementation, before generating the inverter module control strategy based on the adaptive power compensation time window, the system performs a conflict resolution step. The system checks the distribution of multiple adaptive power compensation time windows constructed for different load branches on the system's global time axis. The system traverses all records of adaptive power compensation time windows, extracts the start and end time parameters from each record, and determines whether any two adaptive power compensation time windows overlap by comparing the time interval relationships. For example, if the adaptive power compensation time window constructed for the load branch with load branch identifier L3 has a time interval of [T3_start, T3_end], and the adaptive power compensation time window constructed for the load branch with load branch identifier L5 has a time interval of [T5_start, T5_end], the system determines that if the intervals [T3_start, T3_end] and [T5_start, T5_end] intersect, then the adaptive power compensation time windows corresponding to the load branch with load branch identifier L3 and the load branch with load branch identifier L5 overlap in time. It is understandable that time overlap means that multiple load branches may require additional power during the same time period.

[0087] In some embodiments, if the system detects an adaptive power compensation time window with time overlap, the system calculates the sum of the supplementary power required by each relevant branch during the overlapping period. Based on the defined power compensation waveform amplitude and waveform function of each load branch, the system calculates the sum of the instantaneous values ​​of the supplementary power required by each branch at each moment during the overlapping period. The system compares the calculated total power demand during the overlapping period with the maximum available remaining capacity of the inverter power supply module under the current operating conditions. The maximum available remaining capacity is obtained by subtracting the power value currently allocated to other non-compensated loads from the rated capacity of the inverter power supply module. The system then determines whether the total power demand exceeds the capacity limit of the inverter power supply module. The judgment logic is based on an inequality, expressed as follows:

[0088] in: Represents the overlapping time points At this point, the sum of the instantaneous values ​​of the supplementary power required by all relevant branches. This represents the set of identifiers for all load branches where time windows overlap. The load branch with load branch identifier i represents the load branch at time i in its power compensation waveform. The power amplitude, This represents the maximum remaining capacity that the inverter power supply module can provide under the current operating conditions. The system calculates in real time the capacity at each sampling moment within the overlapping time period. The value is checked to see if it consistently exceeds [a certain value]. .

[0089] In practical implementation, if the system determines that the total power demand exceeds the capacity limit of the inverter power supply module, then there is a problem. In such cases, the system adjusts the power allocation during overlapping periods based on predefined priority parameters for each branch or on the urgency parameters represented by the energy gap values ​​calculated for each branch in recent events. Adjustment methods include proportionally reducing the amplitude of the power compensation waveform for each load branch during the overlapping period, with the reduction ratio inversely or directly proportional to the priority or energy gap value of each load branch; or fine-tuning the start time of some adaptive power compensation time windows to avoid overlap. For example, the start time of the adaptive power compensation time window for the load branch identified as L5 is delayed from T5_start to T5_start+Δt, thus preventing it from overlapping with the time window of the load branch identified as L3. The system ensures that after adjustment, the total power demand at any given time is within the inverter module capacity limit. Within. Optional, the priority parameter is set by the configuration file during system initialization.

[0090] In some embodiments, after the conflict resolution step is completed, the system generates a final version of the inverter module control strategy. This final version integrates all adaptive power compensation time windows and the adjustment results from conflict resolution. The system compiles the final version of the inverter module control strategy into target code executable by a digital signal processor (DSP) and loads the target code into the DSP of the inverter power supply module. Based on the loaded target code, the DSP generates a pulse width modulation (PWM) signal with precise timing in each control cycle. The PWM signal directly controls the on / off operation of the power switching devices on each phase arm of the inverter circuit, thereby achieving hot-swappable power supply control. It can be understood that the DSP is the final execution carrier of the control strategy. The changes in the operating state of the power switching devices directly determine the voltage and current waveforms output from the inverter power supply module to the bus, thereby achieving precise reshaping of the energy injection process of the load branch in terms of timing and amplitude.

[0091] See Figure 5 During the conflict resolution process of the inverter power module's power compensation time window, the relationship between the original total power demand (red solid line), the module's maximum remaining capacity (black dashed line, 300W), and the total power demand after conflict resolution (green solid line) is presented with time (ms) as the horizontal axis and power demand (W) as the vertical axis. Specifically, the original total power demand exhibits a stepped fluctuation at different times, with higher power peaks in some periods. Through conflict resolution strategies (such as proportional reduction of power amplitude and fine-tuning of the time window start time), the waveform of the total power demand after conflict resolution is reshaped. Its power value throughout the entire time period is within the module's maximum remaining capacity limit, and a reasonable allocation of power demand is achieved in terms of timing. This demonstrates the constraint and optimization effect of conflict resolution on the total power demand, ensuring the power supply stability of the inverter power module in hot-swappable scenarios.

[0092] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0093] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A power supply control method for a hot-swappable inverter power module, characterized in that, include: When a physical connection is detected between the inverter power module and the system bus, the initial electrical disturbance signal caused by the connection action is captured. The severity of the connection behavior is identified based on the changing pattern of the initial electrical disturbance signal, and the severity is mapped to the initial intensity label of the hot-plug event. Based on the initial strength label, the bus energy flow spectrum within a preset time period before and after the hot-swap connection event is traced back. Power fluctuation segments that are causally related to the initial strength label in time are extracted from the spectrum, and the energy absorption lag characteristics of each load branch within the power fluctuation segment are analyzed to form a load response lag list with time sequence marking. The load response lag list is compared with the system's preset power scheduling cycle template. The offset between the lag time and the scheduling cycle boundary is calculated. Load branches with offsets exceeding the tolerance are selected. Based on the duration and frequency of their lag characteristics, an adaptive power compensation time window is constructed for the load branches with offsets exceeding the tolerance. Based on the adaptive power compensation time window, in the subsequent power distribution cycle, the energy injection process of the corresponding load branch is reshaped according to a preset timing, generating an inverter module control strategy that includes the reshaped power release trajectory, and driving the power switching devices to turn on and off according to the inverter module control strategy.

2. The power supply control method for a hot-swappable inverter power module according to claim 1, characterized in that, The step of forming a load response hysteresis list with time-stamped tags includes: Read the bus voltage and total current sampling data corresponding to the time node of the initial intensity label, and draw a dynamic energy spectrum centered on the time node; In the dynamic energy spectrum, the power dip region caused by module access is located, and the depth value of the power dip and the time taken to recover to steady state are recorded. The product of the depth value and the recovery time is quantified as the total disturbance energy. Based on the magnitude of the total disturbance energy, trace downstream in the dynamic energy spectrum to identify the establishment process of the current in each parallel load branch, and mark the time taken for the current in each load branch to rise from the beginning to reach the steady state value, as the inherent response delay of the load branch. The inherent response delays of all load branches are summarized and arranged in the order in which the delays occur. The corresponding load branch identifiers are associated with the total disturbance energy to generate a load response hysteresis list with time-series tags.

3. The power supply control method for a hot-swappable inverter power module according to claim 2, characterized in that, The steps for constructing the adaptive power compensation time window include: Obtain the load response hysteresis list with time-series markers, and extract the inherent response delay data for each record in the list; The system calls a preset power scheduling cycle template, which defines the theoretical start and end times of energy injection for each load branch in a complete power allocation cycle. The inherent response delay data is compared with the corresponding theoretical cutoff time in the power scheduling cycle template, and the time difference between the actual energy establishment completion point and the theoretical cutoff point is calculated and recorded as the scheduling offset. Set a scheduling offset threshold, filter out load branch records whose scheduling offset exceeds the offset threshold, and determine that the load branch whose scheduling offset exceeds the threshold has a situation where power distribution has not fully converged; For each load branch with incomplete convergence, the duration of its inherent response delay and the frequency of its occurrence in multiple consecutive hot-plug events are analyzed. Based on the duration and frequency, an additional energy replenishment period independent of the standard scheduling cycle is calculated, which is the adaptive power compensation time window constructed for the load branch.

4. The power supply control method for a hot-swappable inverter power module according to claim 3, characterized in that, The step of generating an inverter module control strategy that includes the reshaped power release trajectory includes: Based on the adaptive power compensation time window, the standard power scheduling cycle template is locally modified. After the theoretical energy injection cutoff time of the corresponding load branch, the adaptive power compensation time window is inserted to form the modified scheduling sequence diagram. In the revised scheduling timing diagram, a specific power compensation waveform is defined for each inserted adaptive power compensation time window. The power compensation waveform starts at the original theoretical cutoff time and ends at the end of the compensation time window. Its amplitude is dynamically calculated based on the energy gap of the load branch in the most recent hot-plug event. The defined power replenishment waveform is spliced ​​and smoothly transitioned with the original power release waveform within the standard scheduling cycle to ensure that the entire energy injection process is continuous in time and without abrupt changes in amplitude, forming a reshaped power release trajectory for each load branch. The power release trajectories of all load branches are integrated, and the power release trajectories of the load branches are converted into pulse width modulation signal sequences of the corresponding power switching devices according to the topology of the inverter power supply module. The pulse width modulation signal sequence is the control strategy of the inverter module.

5. The power supply control method for a hot-swappable inverter power module according to claim 4, characterized in that, The method also includes a step of closed-loop verification of the power compensation process: During the execution of the inverter module control strategy, the current feedback signal of the corresponding load branch within the adaptive power compensation time window is collected in real time. The collected current feedback signal is compared point by point with the expected current waveform in the reshaped power release trajectory, and the real-time error between the two in terms of amplitude and trend is calculated. If the real-time error continues to exceed the preset allowable range, the amplitude of the currently executing power compensation waveform or the duration of the adaptive power compensation time window will be dynamically adjusted to form a compensation parameter adjustment command. Based on the compensation parameter adjustment instruction, the pulse width modulation signal of the corresponding segment in the inverter module control strategy is updated online, and the adjusted parameters are recorded in the historical compensation feature library of the load branch for optimizing the construction of the subsequent adaptive power compensation time window.

6. The power supply control method for a hot-swappable inverter power module according to claim 5, characterized in that, The step of constructing a historical compensation feature library and using it for optimization includes: An independent compensation record set is established for each load branch to store the amplitude adjustment and duration adjustment included in each compensation parameter adjustment instruction; When the adaptive power compensation time window is constructed again for the same load branch, the corresponding compensation record set is queried, and the statistical characteristics of the most recent adjustment amounts are extracted, including the mean and trend of the adjustment amounts. The extracted statistical features are used as prior knowledge and applied to the initial calculation of the amplitude of the power compensation waveform and the initial duration setting of the adaptive power compensation time window, so that the initial compensation parameters are closer to the actual needs and the adjustment range and number of times in the subsequent closed-loop verification are reduced.

7. The power supply control method for a hot-swappable inverter power module according to claim 1, characterized in that, The process of capturing the initial electrical disturbance signal caused by the connection action also includes a step of identifying the interference signal: Upon detecting a physical connection, a high-frequency sampling channel is activated to continuously acquire the voltage differential signal at the bus connection point. Morphological analysis was performed on the acquired voltage differential signal to distinguish between brief high-frequency oscillations caused by mechanical contact bounce and continuous unidirectional changes caused by the charging of the internal capacitor of the module. Only the continuous unidirectional change signal caused by the charging of the internal capacitor of the module is used as the effective initial electrical disturbance signal for subsequent severity identification, while the oscillation signal caused by mechanical contact bounce is filtered out.

8. The power supply control method for a hot-swappable inverter power module according to claim 7, characterized in that, The steps for identifying the intensity of connection behavior specifically include: The effective initial electrical disturbance signal is integrated in the time domain to calculate its accumulated energy value over a preset time period. Multiple energy threshold ranges are set, each range corresponding to a level of connection intensity; The calculated cumulative energy value is assigned to the corresponding energy threshold interval, thereby mapping the continuous signal energy value to the initial intensity label of the discrete hot-plug event.

9. The power supply control method for a hot-swappable inverter power module according to claim 1, characterized in that, The method further includes a conflict resolution step before generating the inverter module control strategy: Check whether the multiple adaptive power compensation time windows constructed for different load branches overlap on the time axis; If there are overlapping power compensation time windows, calculate the total amount of supplementary power required by each branch during the overlapping period, and determine whether the total amount exceeds the maximum remaining capacity that the inverter power supply module can provide under the current operating conditions. If the limit is exceeded, the power compensation waveform amplitude of each branch during the overlapping period will be reduced proportionally according to the priority of each branch or the urgency of the energy gap, or the start time of some power compensation time windows will be finely adjusted to avoid overlap, so as to ensure that the total power demand is within the module capacity limit. After the conflict resolution step is completed, the final version of the inverter module control strategy is generated and compiled into executable code, which is then loaded into the digital signal processor of the inverter power supply module. The digital signal processor generates precise pulse width modulation signals based on the code to control the operation of the power switching devices, thereby realizing hot-swappable power supply control.

10. A hot-swappable inverter power supply module, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the power supply control method for the hot-swappable inverter power module as described in any one of claims 1 to 9.

Images