An inspired all-in-one machine mode switching method

Abstract

This invention discloses a method for switching modes in an integrated heuristic control system, belonging to the field of motor control technology. The method includes: Step 1, constructing a dual-mode shared state variable set, incorporating the state variables of the starting mode and the control quantities of the generation mode into the same extended state vector to form a unified state space; Step 2, at the end of the starting phase, when the estimated rotor electric angular velocity reaches a preset switching threshold, using the unified state space as input, while maintaining the operation of the main control channel in the starting mode, simultaneously calculating the output command of the generation mode controller in the current state to form a shadow channel; Step 3, based on the difference between the output command of the main control channel and the output command of the shadow channel. This invention can effectively suppress the step disturbance of the control quantity being loaded from zero during mode switching, alleviate the switching impact problem caused by the dual-track separation of the two control systems, and effectively improve the smoothness of operating condition switching.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, and more specifically, to a method for switching modes in an integrated heuristic device. Background Technology

[0002] With the rapid development of my country's aviation industry, especially the continuous maturation of more-electric aircraft technology, there is ample room for the lightweighting and integration of airborne electrical equipment. Currently, in aircraft engine accessory systems, traditional independent starter motors and generators occupy a large space and account for a high proportion of weight, and will gradually be replaced by integrated starter-generator designs. At the same time, with the advancement of power electronics technology and motor control algorithms, high-power-density controllers are increasingly widely used in the aviation field. Currently, existing integrated starter-generator motor control schemes generally adopt a hard-switching strategy based on speed thresholds. When the motor speed reaches a preset threshold, the power topology and control algorithm are directly switched. However, this switching method has significant technical defects. During the mode switching process, severe current and voltage surges and torque fluctuations are generated, which can easily cause DC bus voltage drops and system oscillations. In severe cases, it can trigger electrical protection mechanisms, leading to operating condition switching failure and failing to meet the high stability and high reliability requirements of aero-engines.

[0003] The root cause lies in the disconnect between the power topology and control algorithm in the existing solution. The starting mode uses a three-phase full-bridge inverter with a sensorless FOC algorithm, relying on the coupled state equations of motor current, flux linkage, and speed to achieve torque and speed control. The generation mode uses a Vienna rectifier topology with dual closed-loop voltage and current control, relying on the rectifier power balance equation to achieve voltage stabilization and power control. Switching between these two independent physical-control systems inevitably generates system disturbances. Secondly, sensorless FOC relies on EKF recursive state estimation. The existing switching method abruptly resets the EKF state variables, disrupting the algorithm's recursive continuity and causing drastic fluctuations in rotor position and speed estimations, further exacerbating control output oscillations. Finally, the definitions of the core control variables in the two modes are fundamentally conflicting. The starting mode uses torque current components, excitation current components, and current components as the control core, while the generation mode uses bus voltage and power factor. Switching between operating conditions requires simultaneous changes to the controlled object and control strategy, resulting in highly volatile control architecture.

[0004] In summary, existing technologies have inherent technical contradictions. The need for smooth switching with low disturbance and no obvious interruption is incompatible with the problem of state continuity disruption caused by control architecture switching. Hard switching sacrifices stability, and delayed transition sacrifices response speed. It is difficult to simultaneously meet the dual stringent requirements of "fast response, high stability, and zero failure" for motor condition switching in aviation scenarios. Therefore, it is urgent to optimize and inspire a heuristic integrated machine mode switching method. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this invention is to provide a method for switching modes of an integrated machine that can effectively suppress the step disturbance of the control quantity being loaded from zero during mode switching, alleviate the switching shock caused by the dual-track separation of the two control systems, and effectively improve the smoothness of operating condition switching.

[0006] To solve the above problems, the present invention adopts the following technical solution:

[0007] A method for switching modes of an integrated machine, the method includes: Step 1, constructing a dual-mode shared state variable set, incorporating the state variables in the start-up mode and the control quantities in the power generation mode into the same extended state vector to form a unified state space;

[0008] Step 2: At the end of the start-up phase, when the estimated value of the rotor electric angular velocity reaches the preset switching threshold, the unified state space is used as input. While maintaining the operation of the main control channel of the start-up mode, the output command of the power generation mode controller in the current state is calculated in parallel to form a shadow channel.

[0009] Step 3: Calculate the state consistency index value based on the difference between the output command of the main control channel and the output command of the shadow channel, and the deviation between the actual state variable and the corresponding reference state variable in the unified state space; when the state consistency index value is less than the preset threshold for state consistency and the estimated value of the rotor electric angular velocity reaches the preset switching threshold, trigger the switch from the start-up mode to the power generation mode.

[0010] Step 4: When switching is triggered, perform a mapping operation on the error covariance matrix of the state estimator to map the state error covariance under the current control model to the state error covariance under the target control model.

[0011] Step 5: During the power topology switching process, the switching period carrier phases of the two power conversion topologies are locked, and the drive signals of the two power conversion topologies are exchanged at the phase corresponding to the zero crossing of the phase current in the carrier period to complete the power topology switching.

[0012] Step 6: After the power topology switching is completed, the output commands of the main control channel and the output commands of the shadow channel are weighted and merged according to preset weights to generate the final control output. The preset weights change monotonically from the first weight value representing the dominance of the main control channel to the second weight value representing the dominance of the shadow channel within multiple control cycles after the switching is completed.

[0013] Further, step 1 includes:

[0014] Step 11: Using the two-phase stationary coordinate system as the reference coordinate system, the stator current vector component, rotor flux vector component, and rotor electric angular velocity in the starting mode are used as the first set of state variables to form the starting state vector; the DC bus voltage and AC side input current vector component in the power generation mode are used as the second set of state variables to form the power generation state vector.

[0015] Step 12: Merge the starting state vector and the power generation state vector to form an extended state vector, and use the extended state vector as a unified state space.

[0016] Further, step 2 includes:

[0017] Step 21: Monitor the estimated value of the rotor electric angular velocity in the unified state space. When the estimated value of the rotor electric angular velocity reaches the set proportion of the preset switching threshold, start the parallel calculation preparation process for the output command of the power generation mode controller. The preset switching threshold is determined according to the minimum operating speed required by the power generation mode, and the set proportion is a positive number less than 1.

[0018] Step 22: Using the measured value of DC bus voltage in the unified state space, the measured value of AC side input current vector component, the estimated value of rotor electric angular velocity, and the preset target value of DC bus voltage in the power generation mode as input quantities, calculate the output command according to the control law of the power generation mode; the output command includes the switching timing parameters of the power conversion topology in the power generation mode; the calculation is performed in parallel with the calculation of the output command of the main control channel in the start-up mode.

[0019] Step 23: The output command calculated in step 22 is temporarily stored in the shadow channel register, and the calculation is repeated in each subsequent control cycle to continuously update the output command content in the shadow channel register. At the same time, in each control cycle, the updated output command in the shadow channel register is associated with the rotor electric angular velocity estimate and stored to form an output command sequence with a speed tag.

[0020] Further, step 3 includes:

[0021] Step 31: Obtain the output command of the main control channel in the current control cycle, and read the output command stored in association with the current rotor electric angular velocity estimate from the shadow channel register, and calculate the difference between the two.

[0022] Step 32: Obtain the measured values ​​of each state variable in the unified state space and the corresponding reference values ​​of each state variable. Subtract each measured value from each reference value one by one and take the absolute value to obtain the absolute value of the deviation of each state variable.

[0023] Furthermore, step 3 also includes:

[0024] Step 33: The difference is used as the first dimension deviation, and the absolute values ​​of each deviation are multiplied by their respective preset weight coefficients and then summed to obtain the second dimension comprehensive deviation; the first dimension deviation and the second dimension comprehensive deviation are normalized to a preset numerical range; the normalized first dimension deviation value and the normalized second dimension deviation value are added to obtain the state consistency index value; the state consistency index value is compared with the state consistency preset threshold. When the state consistency index value is less than the state consistency preset threshold and the rotor electric angular velocity estimate reaches the preset switching threshold, the switch from the start-up mode to the power generation mode is triggered.

[0025] Further, step 4 includes:

[0026] Step 41: Obtain the first Jacobian matrix of the starting mode state estimator, which is constructed based on the state transition function of the starting mode at the state estimate at the current time; Obtain the second Jacobian matrix of the power generation mode state estimator, which is constructed based on the state transition function of the power generation mode at the state estimate at the current time.

[0027] Step 42: Using the first Jacobian matrix and the second Jacobian matrix as input, multiply the second Jacobian matrix on the left by the inverse of the first Jacobian matrix to obtain the transformation matrix.

[0028] Furthermore, step 4 also includes:

[0029] Step 43: Obtain the current error covariance matrix of the starting mode state estimator; multiply the current error covariance matrix on the left by the transformation matrix, and then multiply it on the right by the transpose of the transformation matrix to obtain the mapped error covariance matrix;

[0030] Step 44: Perform a symmetry operation on the mapped error covariance matrix to obtain a symmetric matrix; perform a positive definiteness check on the symmetric matrix. If the positive definiteness check passes, the symmetric matrix is ​​used as the error covariance matrix of the power generation mode state estimator; if the positive definiteness check fails, the diagonal elements of the symmetric matrix are modified, and the modified symmetric matrix is ​​used as the error covariance matrix of the power generation mode state estimator.

[0031] Further, step 5 includes:

[0032] Step 51: Align the carrier period of the startup mode power conversion topology with the carrier period of the generation mode power conversion topology to the same carrier frequency, and lock the starting phase of both to the same phase angle.

[0033] Step 52: Using the phase-locked carrier signal as the time reference, detect the instantaneous values ​​of the three-phase currents, and record the carrier phase angle corresponding to the instantaneous value of each phase current crossing the zero point, as the zero-crossing phase angle of each phase current.

[0034] Furthermore, step 5 also includes:

[0035] Step 53: Based on the zero-crossing phase angle of the current in each phase, at the zero-crossing phase angle of the current in each phase, generate a first switch control signal for turning off the drive signal of the power conversion topology switch in the starting mode of that phase, and a second switch control signal for turning on the drive signal of the power conversion topology switch in the generation mode of that phase, so that the output time of the first switch control signal and the second switch control signal are both locked at the same zero-crossing phase angle of the current; and execute the drive signal handover operation of the three phases in sequence according to the corresponding zero-crossing phase angle of the current in each of the three phases.

[0036] Further, step 6 includes:

[0037] Step 61: Take the moment when the power topology switching is completed as the start time of the weight change, take the preset total number of change cycles as the change duration, and determine the end time of the change; set the weight value of the control cycle where the start time is located as the first weight value, and the weight value of the control cycle where the end time is located as the second weight value; in each control cycle during the change period, calculate the current weight value according to the progress of the current control cycle in the total number of change cycles, according to the preset weight function curve.

[0038] Step 62: In each control cycle, obtain the first output command of the main control channel and the second output command of the shadow channel, multiply the first output command by the current weight value calculated in step 61, multiply the second output command by one minus the current weight value, and add the two products to obtain the final control output of the current control cycle.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] (1) This scheme constructs a unified state space shared by the two modes and starts the parallel calculation of the shadow channel of the power generation mode in advance at the end of the start-up period, so that the power generation control output tracks the convergence of the current working condition in advance. This can effectively suppress the step disturbance of the control quantity being loaded from zero at the moment of mode switching, alleviate the switching shock problem caused by the dual-track separation of the two control systems, and effectively improve the stability of the working condition switching.

[0041] (2) This scheme adopts a state consistency index that integrates output difference and state deviation, and triggers mode switching together with speed threshold, replacing the traditional hard switching strategy of single speed threshold. This ensures that the switching action occurs when the operating conditions are stable and the two control outputs are highly matched, reducing the probability of switching failure and improving the operational reliability of the aviation electrical system.

[0042] (3) This scheme adopts phase-by-phase switching of phase current under carrier phase-locked loop to realize power topology switching. Combined with the smooth transition of control output with multi-cycle weight change after switching, it suppresses current and voltage impact and torque fluctuation from two layers of power path and control law, ensuring the stability of DC bus voltage and adapting to the high stability and high reliability switching requirements of aviation scenarios. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0044] Figure 1 This is a flowchart illustrating the overall process architecture for switching modes in the all-in-one machine according to the present invention.

[0045] Figure 2 This is a diagram illustrating the unified space construction of dual-mode shared state variables in this invention.

[0046] Figure 3 This is a storage diagram showing the shadow channel parallel computing and rotational speed correlation of the present invention;

[0047] Figure 4 This is a diagram showing the phase-by-phase transition between carrier phase-locked loop and zero-crossing phase current of the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0049] Please see Figures 1 to 4 A method for switching modes in an all-in-one machine, the method comprising:

[0050] Step 1: Construct a dual-mode shared state variable set, incorporating the state variables in the start-up mode and the control quantities in the power generation mode into the same extended state vector to form a unified state space. The specific operations are as follows:

[0051] The process of constructing a shared state variable set for dual-mode operation involves incorporating the state variables and control-related variables required for controller operation in both modes into the same data structure, forming a unified state space that can be read by both controllers. In the actual operation of the integrated aerospace heuristic, the start-up mode controller and the power generation mode controller originally relied on independent state acquisition and estimation links. During the start-up phase, only the motor-side state variables were updated, and during the power generation phase, only the rectifier and bus-side state variables were updated. State variable jumps and omissions occurred during mode switching. By constructing a shared state variable set, the state variables required for the power generation mode can be included in the real-time update range at the end of the start-up phase, allowing the power generation mode controller to complete pre-calculation based on the actual operating state before the formal switch. The unified state space also serves as a common basis for the state estimator mapping, enabling smooth transfer of state estimation results between the two modes during switching and avoiding state fluctuations caused by estimator resets.

[0052] Step 1 also includes the following steps:

[0053] Step 11: Using the two-phase stationary coordinate system as the reference coordinate system, the stator current vector component, rotor flux linkage vector component, and rotor electric angular velocity in the starting mode are used as the first set of state variables to form the starting state vector; the DC bus voltage and AC side input current vector component in the power generation mode are used as the second set of state variables to form the power generation state vector. The specific operation is as follows:

[0054] A two-phase stationary coordinate system is selected as a unified reference coordinate system. The vector description in this coordinate system can be adapted to both the field-oriented control in the starting mode and the rectifier control in the power generation mode, reducing the calculation errors and phase shifts introduced by the conversion between different coordinate systems. The first set of state variables constitutes the starting state vector. The stator current vector component includes the stator current components along the α and β axes in the two-phase stationary coordinate system. This component is obtained by collecting the instantaneous three-phase currents from Hall current sensors located at the output terminals of the three-phase stator windings and then calculating using the Clarke transform. The rotor flux linkage vector component includes the rotor flux linkage components along the α and β axes in the two-phase stationary coordinate system. This component is estimated in real time by the extended Kalman filter state estimator of the starting mode based on the stator voltage equation and the flux linkage recursive relationship. The rotor electric angular velocity is the instantaneous value of the rotor electric angular frequency, obtained by differentiating the rotor position angle output by the state estimator, or directly by the recursive output of the state observer. The above three categories of variables together constitute the complete state input required for the operation of the starting mode controller, which can be directly used for torque and speed regulation in sensorless field-oriented control.

[0055] The second set of state variables constitutes the generation state vector. The DC bus voltage is directly sampled by a resistive voltage divider sensor placed between the positive and negative terminals of the DC bus, serving as the main feedback quantity for the generation mode voltage regulation control. The AC side input current vector component includes the AC side input current components along the α and β axes in a two-phase stationary coordinate system. This component is calculated by collecting the three-phase instantaneous current from a Hall current sensor placed at the AC input terminal of the rectifier and then performing a Clarke transform, serving as the feedback quantity for the generation mode current loop control. The two sets of state variables have a natural coupling relationship at the electrical level. The motor stator current and the rectifier AC side current have corresponding conduction paths, and the rotor electric angular velocity is proportional to the amplitude of the AC side induced electromotive force. Therefore, the two sets of variables have a physical basis for merging into a single state system.

[0056] Step 12: Merge the starting state vector and the power generation state vector to form an extended state vector, and use the extended state vector as the unified state space. The specific operation is as follows:

[0057] The starting state vector and the generating state vector are concatenated sequentially to form a one-dimensional extended state vector. The total number of elements in this extended state vector is the sum of the number of elements in the two sets of state vectors. During the concatenation process, the elements of the starting state vector are arranged in the order of stator current α component, stator current β component, rotor flux linkage α component, rotor flux linkage β component, and rotor electric angular velocity. The elements of the generating state vector are arranged in the order of DC bus voltage, AC side current α component, and AC side current β component, ultimately forming an extended state vector in column vector form. The unified state space is a linear space based on this extended state vector. Any point in the space corresponds to a set of instantaneous states that simultaneously contain operating information from both the motor side and the rectifier side.

[0058] Within each control cycle, all sampled variables in the extended state vector are synchronously acquired by the processor's analog-to-digital converter at the same sampling moment. The component values ​​in the corresponding coordinate system are then calculated through coordinate transformation. Estimated variables are recursively updated based on the sampled data at the same moment, ensuring that the time base of all elements in the vector is consistent and avoiding state deviations caused by time asynchrony. A unified state space serves as the common input interface for all subsequent control operations. Both the main control channel in start-up mode and the shadow channel in power generation mode can directly read all required state quantities from this state space without needing to build independent sampling and estimation links. This ensures the continuity of the controller input system before and after mode switching and reduces control output oscillations caused by abrupt changes in the state input structure.

[0059] In a preferred embodiment of the present invention, step 2 is further included: at the end of the start-up phase, when the estimated rotor electric angular velocity reaches a preset switching threshold ratio, using a unified state space as input, while maintaining the operation of the main control channel of the start-up mode, the output command of the power generation mode controller in the current state is calculated in parallel to form a shadow channel. The specific operation is as follows:

[0060] During the startup process of the integrated aero-inspired generator, the main control channel continuously runs the startup mode control algorithm, continuously outputting torque to drive the engine rotor to accelerate until the engine completes ignition and enters a self-sustaining operating state. If the generator mode controller calculation is only started when the speed reaches the switching threshold, the controller output needs to undergo an adjustment process from the initial value to the steady-state value. During this period, large fluctuations in the control quantity will occur and be transmitted to the power topology, causing current and voltage surges. By starting the shadow channel for parallel calculation in advance, the generator mode controller can continuously iterate and output based on the real-time operating state for multiple control cycles before the formal switch, so that the control command gradually converges to a steady-state value that matches the current operating condition. When the switching conditions are met, the outputs of the two channels are already compatible. All inputs of the shadow channel are taken from a unified state space and share the same set of state sampling and estimation data with the main control channel, ensuring that the state reference of the two calculations is completely consistent and avoiding output deviations caused by differences in data sources.

[0061] Step 2 also includes the following steps:

[0062] Step 21: Monitor the estimated rotor electric angular velocity in the unified state space. When the estimated rotor electric angular velocity reaches a preset percentage of the switching threshold, initiate the parallel calculation preparation process for the generator mode controller to output commands. The preset switching threshold is determined based on the minimum operating speed required for the generator mode, and the set percentage is a positive number less than 1. The specific operation is as follows:

[0063] The rotor electric angular velocity estimate is recursively output by the state estimator of the starting mode, updated once per control cycle and written into a unified state space, which can be directly read and called. The preset switching threshold corresponds to the minimum operating speed at which the DC bus voltage can be stably established in the power generation mode. This threshold is jointly determined by the operating characteristics of the power conversion topology, the motor back EMF coefficient, and the rated DC bus voltage. During the tuning process, it is necessary to ensure that when the speed reaches this threshold, the peak value of the induced electromotive force on the stator side of the motor is higher than the DC bus voltage, so that the rectifier topology has the conditions for energy feedback and voltage regulation.

[0064] The set ratio is a positive number less than 1, used to advance the start time of the shadow channel to before the switching threshold is reached, reserving sufficient pre-adjustment period. In engineering applications, the set ratio can be set to 0.8, corresponding to starting parallel calculation when the speed reaches 80% of the switching threshold, which can typically cover the pre-running time of tens to hundreds of control cycles. Within each control cycle, the control program reads the estimated value of the current rotor electric angular velocity from the unified state space, compares it with the product of the preset switching threshold and the set ratio, and when the estimated value is greater than or equal to the product result for the first time, the initialization operation of generating the shadow channel is triggered, and the parallel calculation preparation process for the output command of the power generation mode controller is started.

[0065] Step 22: Using the measured DC bus voltage, the measured AC input current vector component, the estimated rotor electric angular velocity, and the preset target DC bus voltage for the power generation mode as inputs, calculate the output command according to the control law of the power generation mode. The output command includes the switching timing parameters of the power conversion topology in the power generation mode. The calculation is performed in parallel with the calculation of the output command of the main control channel in the start-up mode. The specific operation is as follows:

[0066] The measured DC bus voltage in the input quantities is obtained by analog-to-digital conversion from the resistor voltage divider sampling circuit arranged between the positive and negative poles of the DC bus, reflecting the actual voltage level of the current DC bus; the AC input current vector component is obtained by sampling the three-phase current sensor at the AC input terminal of the rectifier and converting it to a two-phase stationary coordinate system through Clarke transformation, which is used for current closed-loop regulation; the rotor electric angular velocity estimate is taken from the state estimation result of the unified state space, which is used for feedforward compensation and phase synchronization calculation; the DC bus voltage target value in the power generation mode is the rated operating voltage preset by the system, which serves as the control reference for the voltage outer loop.

[0067] The power generation mode control law adopts a voltage and current dual closed-loop structure. The outer loop calculates the AC side current reference value based on the bus voltage deviation, and the inner loop calculates the modulation wave signal based on the deviation between the current reference value and the measured current. Finally, it outputs the conduction timing parameters of each switch in the corresponding power conversion topology, including the conduction and turn-off times of each bridge arm switch. Parallel computing can be achieved through the multi-task scheduling mechanism of a digital signal processor. Independent computation time slices are allocated within a single control cycle to execute two sets of control algorithms, or a multi-core processor can be used to run the main control channel and shadow channel programs separately. This ensures that the two calculations are based on the state sampling data at the same time, and that the output priority of the main control channel is higher than that of the shadow channel, thus maintaining control over the power topology at all times. The calculation results of the shadow channel are only stored internally and are not directly output to the drive circuit to avoid interference with the current startup process.

[0068] Step 23: The output command calculated in Step 22 is temporarily stored in the shadow channel register, and the calculation is repeated in each subsequent control cycle to continuously update the output command content in the shadow channel register. Simultaneously, in each control cycle, the updated output command in the shadow channel register is associated with the rotor electric angular velocity estimate and stored to form an output command sequence with a speed tag. The specific operations are as follows:

[0069] The shadow channel register is a dedicated storage unit within the processor, used to store the calculation results of the output instructions of the power generation mode controller in the current control cycle. After the power generation mode control law calculation is completed in each control cycle, the new output instruction will overwrite the old data of the previous cycle in the register, ensuring that the register always retains the latest calculation results. While updating the register data, the program uses the estimated rotor electric angular velocity of the current control cycle as a speed tag, and binds it to the set of output instructions for storage, forming a set of instruction data with operating condition tags.

[0070] Multiple sets of tagged instruction data in consecutive cycles are stored sequentially in a ring buffer of preset depth, forming an output instruction sequence with speed tags. The depth of the ring buffer can be set according to the speed increase rate and control cycle duration, and can typically hold dozens to hundreds of sets of instruction data, covering the entire operating cycle from the start of the shadow channel to the switching trigger. The speed tag is used to establish a mapping relationship between the output instruction and the corresponding operating condition. When calculating the deviation between the two outputs, the shadow channel instruction under the corresponding operating condition can be matched according to the current speed, eliminating the inherent differences in instructions caused by dynamic speed changes and improving the accuracy of state consistency judgment. The continuously updated storage mechanism ensures that the shadow channel output always follows the real-time operating condition evolution, ensuring that the instruction data retrieved at the switching trigger moment is completely adapted to the current operating state.

[0071] In a preferred embodiment of the present invention, step 3 is further included: calculating a state consistency index value based on the difference between the output command of the main control channel and the output command of the shadow channel, and the deviation between the actual state variable and the corresponding reference state variable in the unified state space; when the state consistency index value is less than the state consistency preset threshold and the rotor electric angular velocity estimate reaches the preset switching threshold, the switching from the start-up mode to the power generation mode is triggered, and the specific operation is as follows:

[0072] Traditional switching strategies based on a single speed threshold use speed as the sole criterion, failing to identify whether the outputs of the two control systems are compatible under current operating conditions or whether the system is in a steady state. This can easily lead to switching during periods of significant state fluctuation, causing electrical shocks. This step assesses switching suitability from both control output and system state perspectives. The output command difference reflects the similarity of control quantities between the two controllers under the same operating conditions, while the state variable deviation reflects the stability of the current system operation. The state consistency index, formed by combining these two types of information, comprehensively characterizes the maturity of switching conditions. Switching triggers must simultaneously meet both the speed threshold and the consistency threshold, ensuring that the power generation mode has a stable physical speed foundation and that both the control quantity and system state are within a stable range at the moment of switching, thus reducing disturbances caused by mode switching at the triggering timing level. In the actual operation of the aviation heuristic integrated machine, this mechanism can avoid erroneous switching when the speed just reaches the threshold but the bus voltage is still oscillating and the control output has not yet converged. This ensures that the switching action always occurs when the operating conditions are stable and the matching degree of the two control outputs is high, thereby improving the reliability and stability of the switching process. The state consistency index is calculated in real time once in each control cycle to continuously monitor the changes in the system state and output matching degree until the triggering conditions are met.

[0073] Step 3 also includes the following steps:

[0074] Step 31: Obtain the output command of the main control channel within the current control cycle, and read the output command associated with the current rotor electric angular velocity estimate from the shadow channel register, and calculate the difference between the two. The specific operation is as follows:

[0075] The output command of the main control channel is taken from the calculation result of the starting mode control algorithm within the current control cycle. This result is directly used to generate the switching transistor drive signal of the starting power topology, including the conduction duration and phase timing parameters of each bridge arm switching transistor. The output command of the shadow channel is obtained by retrieving the output command sequence with speed tag by the current rotor electric angular velocity estimate. The retrieval process uses the current speed value as the index and matches the set of command data with the closest speed tag in the sequence to ensure that the two sets of commands participating in the difference calculation correspond to the exact same speed condition, eliminating the interference of dynamic speed changes on the difference calculation result. The output command is an ordered set containing multiple timing parameters. When calculating the difference, the parameters with one-to-one physical meaning in the two sets of commands are subtracted one by one. The absolute values ​​of all component differences are taken and then summed to obtain the difference result that characterizes the overall difference between the two outputs. This difference result can directly reflect the degree of deviation of the output control quantity of the two control algorithms under the same operating condition. The smaller the difference, the better the connection between the two outputs and the lower the control quantity jump amplitude generated at the moment of switching.

[0076] Step 32: Obtain the measured values ​​of each state variable in the unified state space and the corresponding reference values. Subtract each measured value from each reference value one by one and take the absolute value to obtain the absolute value of the deviation of each state variable. The specific operation is as follows:

[0077] The unified state space encompasses all key operating parameters on both the starting and generating sides. The stator current vector component and the AC input current vector component are obtained by collecting three-phase instantaneous values ​​from Hall current sensors at corresponding locations and then performing a Clarke transform. The DC bus voltage is obtained from a resistive voltage divider sensor via analog-to-digital conversion. The rotor flux vector component and rotor electric angular velocity are recursively output by the state estimator. Reference values ​​for each state variable are generated based on the control objective of the current operating mode. In starting mode, the current, flux, and speed reference values ​​are output from the outer loop calculation of the starting controller. In generating mode, the DC bus voltage reference value is the system's preset rated target voltage, and the AC current reference value is output from the voltage outer loop regulator. During the calculation process, for each state variable, the measured or estimated value is subtracted from the corresponding reference value, and the absolute value of the result is taken to obtain the absolute deviation value for that state variable. After all state variables are calculated, a set of absolute deviation values ​​corresponding to the state vector dimensions is formed. The larger the absolute deviation value of each parameter, the further the parameter deviates from the target state, and the worse the system's steady-state performance.

[0078] Step 33: The difference is used as the first dimension deviation, and the absolute values ​​of each deviation are multiplied by their respective preset weight coefficients and summed to obtain the second dimension comprehensive deviation; the first dimension deviation and the second dimension comprehensive deviation are normalized to a preset numerical range; the normalized first dimension deviation value and the normalized second dimension deviation value are added to obtain the state consistency index value; the state consistency index value is compared with the state consistency preset threshold. When the state consistency index value is less than the state consistency preset threshold and the rotor electric angular velocity estimate reaches the preset switching threshold, the switch from the start-up mode to the power generation mode is triggered. The specific operation is as follows:

[0079] The output command difference obtained in step 31 is used as the first-dimensional deviation. This deviation reflects the matching degree of the two channels from the control output level. The absolute values ​​of the deviations of each state variable obtained in step 32 are multiplied by their respective preset weight coefficients and then summed to obtain the second-dimensional comprehensive deviation. The weight coefficients are pre-tuned according to the influence of each state variable on the smoothness of switching. Parameters that have a greater impact on the switching process can be set with higher weight coefficients. This deviation reflects the steady-state degree of operation from the system state level. To eliminate the differences in dimensions and numerical ranges between the two types of deviations, normalization is performed on the first-dimensional deviation and the second-dimensional comprehensive deviation. The normalization adopts a linear mapping method, mapping the actual value range of the deviation to a preset numerical range of 0 to 1. The mapping ratio is determined by the maximum allowable fluctuation range of the corresponding deviation.

[0080] To clarify the calculation relationship, the state consistency index can be expressed by the following mathematical relationship. This formula, based on the design logic of a two-dimensional evaluation, first aggregates the deviations of multiple state variables into a single-dimensional comprehensive deviation through weighted summation. Then, it unifies the dimensions and numerical scales of the two types of deviations through linear normalization. Finally, it superimposes these deviations to obtain a comprehensive scalar that can be directly used for threshold comparison, facilitating engineering implementation and parameter tuning. The state consistency index satisfies: , where S is the state consistency index value, which is a dimensionless scalar, and the smaller the value, the better the state consistency; The first dimension deviation is the difference between the output commands of the master control channel and the shadow channel. This is the normalization coefficient for the first dimension bias; This is the normalized coefficient for the overall bias in the second dimension; The preset weight coefficient corresponding to the i-th state variable is a positive real number; For the measured or estimated value of the i-th state variable in the unified state space; is the reference value corresponding to the i-th state variable; n is the total number of state variables involved in the calculation.

[0081] Among them, the normalization coefficient The value is determined by the maximum allowable fluctuation range of the output instruction difference, specifically satisfying... ,in This is the preset maximum allowable difference in output commands, corresponding to the maximum allowable deviation between the two output commands during the switching process. In engineering, it can be taken as 10% to 20% of the output command amplitude under rated operating conditions. This coefficient can be used to linearly map the first dimension deviation Δu to the dimensionless interval [0,1]. When Δu equals At that time, the deviation value of the first dimension after normalization is 1.

[0082] Normalization coefficient The value is determined by the maximum permissible fluctuation range of the overall state deviation, specifically satisfying... ,in The second-dimensional comprehensive deviation is the preset maximum allowable value, corresponding to the maximum allowable state deviation of the system before the switch. In engineering, it can be taken as 5% to 15% of the weighted sum of the rated baseline values ​​of each state variable. This coefficient can be used to linearly map the second-dimensional comprehensive deviation to the dimensionless interval [0,1]. When the second-dimensional comprehensive deviation is equal to... At that time, the normalized second dimension bias value is 1.

[0083] During engineering tuning, the system's permissible switching impact level can be determined first. and Then, the above relationship is used to calculate and obtain and The specific values. Taking an aviation-grade integrated heuristic with a rated power of 10kW as an example, One can take 15% of the rated duty cycle amplitude. 10% of the weighted sum of the rated states can be taken, corresponding to and They can be determined as 6.67 and 10 respectively.

[0084] After obtaining the state consistency index value, it is compared with the preset state consistency threshold, and it is also determined whether the rotor electric angular velocity estimate has reached the preset switching threshold. When both conditions are met, a mode switching trigger signal is output to start the subsequent state estimator mapping, power topology handover and control weight transition process. The preset state consistency threshold is tuned according to the maximum allowable switching impact level of the system. The smaller the threshold is set, the higher the requirement for the state stability before switching and the smaller the switching impact, but the waiting time for switching trigger may be correspondingly extended.

[0085] In a preferred embodiment of the present invention, step 4 is further included: when switching triggers, a mapping operation is performed on the error covariance matrix of the state estimator to map the state error covariance under the current control model to the state error covariance under the target control model. The specific operation is as follows:

[0086] In the sensorless control architecture of the aerospace heuristic, independent state estimators are configured for the start-up mode and the power generation mode. These two estimators construct state transition equations based on different physical models, and their corresponding error covariance matrices represent the confidence distribution of the state estimates under their respective models. If the error covariance matrix of the power generation mode state estimator is directly reset to its initial default value during switching, it will cause a step change in the estimator gain, resulting in drastic fluctuations in state estimates such as rotor position and speed, which will then be transmitted to the control loop, causing output oscillations. This step uses a linear mapping method to transform the converged error covariance matrix in the start-up mode to the state space corresponding to the power generation model. This allows the power generation mode estimator after switching to directly inherit the error statistics before switching, maintaining the continuity of recursive operations. The mapping operation is performed only once at the switching trigger moment. After completion, the power generation mode state estimator can enter the normal recursive operation state based on the mapped covariance matrix without requiring an additional convergence transition process.

[0087] Step 4 also includes the following steps:

[0088] Step 41: Obtain the first Jacobian matrix of the starting mode state estimator, which is constructed based on the state transition function of the starting mode at the current state estimate; obtain the second Jacobian matrix of the power generation mode state estimator, which is constructed based on the state transition function of the power generation mode at the current state estimate. The specific operations are as follows:

[0089] The first Jacobian matrix corresponds to the state estimator for the starting mode. Its construction is based on the discrete state transition function of the starting mode. Partial derivatives are calculated for each state variable in the state vector, and all partial derivatives are arranged in matrix form to form the Jacobian matrix. During calculation, the current state estimate is substituted into the partial derivative expressions to obtain the numerical matrix under the current operating condition. The state transition function for the starting mode originates from the discretized forms of the motor stator voltage balance equation, rotor flux linkage dynamic equation, and mechanical motion equation. The Jacobian matrix represents the linear amplification relationship of small state increments near the state estimate after one-step transition and is a necessary parameter for the extended Kalman filter covariance prediction step. The second Jacobian matrix corresponds to the state estimator for the power generation mode. Its construction is also based on the discrete state transition function of the power generation mode. Similarly, partial derivatives are calculated for each state variable in the power generation state vector and arranged in matrix form. The current state estimate is then substituted into the matrix to obtain the numerical result. The state transition function of the power generation mode is derived from the discretized form of the rectifier power balance equation and the DC bus voltage dynamic equation, which characterizes the recursive propagation characteristics of the rectifier side state variables. Both Jacobian matrices are calculated at the state point at the same switching time to ensure the time consistency of the mapping reference.

[0090] Step 42: Using the first and second Jacobian matrices as input, multiply the second Jacobian matrix on the left by the inverse of the first Jacobian matrix to obtain the transformation matrix. The specific operation is as follows:

[0091] Using the first and second Jacobian submatrices as input, the invertibility of the first Jacobian submatrice is first checked: the determinant of the first Jacobian submatrice is calculated. If the absolute value of the determinant is greater than a preset singularity threshold, the matrix is ​​determined to be invertible, and the second Jacobian submatric is left-multiplied by the inverse of the first Jacobian submatric to obtain the transformation matrix. If the absolute value of the determinant is less than or equal to the singularity threshold, the Moore-Penrose pseudo-inverse is used instead of the inverse matrix, and the second Jacobian submatric is left-multiplied by the pseudo-inverse of the first Jacobian submatric to obtain the transformation matrix. The singularity threshold is tuned according to the numerical calculation accuracy of the system and can be set to a value suitable for engineering applications. To avoid numerical divergence caused by ill-conditioned matrix inversion, pseudo-inverse operation can be achieved through singular value decomposition, which improves the numerical robustness of the algorithm while ensuring mapping accuracy.

[0092] The physical meaning of the transformation matrix is ​​to linearly transform the error vector in the state space of the starting model into the corresponding error vector in the state space of the power generation model, achieving an equivalent transformation of error distribution under different state coordinate systems. This computational logic originates from the basis transformation concept of linear spaces. The inverse of the first Jacobian matrix is ​​used to map the state error under the starting model back to the general state increment space, while the second Jacobian matrix projects the general state increment onto the state space of the power generation model. The two are cascaded to form a complete error transformation channel between models. To clarify the computational relationship, the mathematical expression of the transformation matrix can be expressed as: ;

[0093] in The transformation matrix has rows corresponding to the dimension of the power generation mode state vector and columns corresponding to the dimension of the start-up mode state vector. is the first Jacobian matrix, which corresponds to the partial derivative matrix of the start-up mode state transition function at the current state estimate; is the second Jacobian matrix, which corresponds to the partial derivative matrix of the power generation mode state transition function at the current state estimate; This indicates an inverse operation or a pseudo-inverse operation, selected based on the matrix invertibility verification result.

[0094] Step 43: Obtain the current error covariance matrix of the startup mode state estimator; multiply the current error covariance matrix on the left by the transformation matrix, and then on the right by the transpose of the transformation matrix to obtain the mapped error covariance matrix. The specific operation is as follows:

[0095] The calculation process first reads the current error covariance matrix of the start-up mode state estimator at the switching moment. This matrix records the error variance and covariance information of each state estimator in the start-up mode, and is the result of long-term recursive convergence of the estimator. The current error covariance matrix is ​​then multiplied left by the transformation matrix and right by the transpose of the transformation matrix. The result is the mapped error covariance matrix. This operation follows the covariance propagation law of linear transformation of random vectors. If a random vector is transformed by a linear transformation matrix to obtain a new random vector, the covariance matrix of the new vector is equal to the transformation matrix multiplied left by the original covariance matrix and right by the transpose of the transformation matrix. Through this operation, the error statistical distribution in the start-up mode can be equivalently transmitted to the state space of the power generation mode, enabling the initial covariance of the power generation mode estimator to accurately reflect the state confidence level under the current operating condition, avoiding estimation oscillations caused by the mismatch between the default initial value and the actual error level.

[0096] The corresponding mathematical expression is: ;in This is the mapped power generation mode error covariance matrix; The current error covariance matrix of the start-up mode state estimator at the switching time; The transformation matrix obtained in step 42; This is the transpose of the transformation matrix.

[0097] Step 44: Perform a symmetry operation on the mapped error covariance matrix to obtain a symmetric matrix; perform a positive definiteness check on the symmetric matrix. If the positive definiteness check passes, the symmetric matrix is ​​used as the error covariance matrix of the power generation mode state estimator; if the positive definiteness check fails, a correction amount is added to the diagonal elements of the symmetric matrix, and the corrected symmetric matrix is ​​used as the error covariance matrix of the power generation mode state estimator. The specific operations are as follows:

[0098] The symmetry operation is performed by adding the mapped matrix to its transpose and then dividing by two to eliminate the asymmetric components of the matrix introduced by numerical calculation errors, thus obtaining the symmetric matrix. The error covariance matrix must be a symmetric positive definite matrix mathematically. Symmetry processing ensures the symmetry of the matrix. The positive definiteness check is used to verify whether the matrix satisfies the positive definiteness condition. The check can be achieved by calculating whether all the principal minors of the matrix are greater than zero, or by judging whether the Cholesky decomposition can be completed normally.

[0099] If the positive definiteness check passes, it means that the mapped matrix has a valid covariance attribute, and the symmetric matrix can be directly assigned to the power generation mode state estimator as its initial error covariance matrix. If the positive definiteness check fails, it means that numerical errors or model differences cause non-positive definite components in the mapping result. In this case, a preset positive correction amount is uniformly added to all diagonal elements of the symmetric matrix. The correction amount is a small positive number. By increasing the variance benchmark value of each state quantity, the matrix is ​​forced to meet the positive definiteness condition, and then the corrected matrix is ​​used as the error covariance matrix of the power generation mode state estimator. The value of the correction amount needs to balance positive definiteness and estimation accuracy. If the value is too large, it will lead to a low initial gain of the estimator and slower convergence. If the value is too small, positive definiteness cannot be guaranteed. In engineering, it is usually taken on the order of one-hundredth to one-thousandth of the variance of the corresponding state quantity under rated conditions.

[0100] In a preferred embodiment of the present invention, step 5 is further included: during the power topology switching process, the switching period carrier phases of the two power conversion topologies are locked, and the driving signals of the two power conversion topologies are exchanged at the phase corresponding to the zero crossing of the phase current in the carrier period to complete the power topology switching. The specific operation is as follows:

[0101] The two power conversion topologies of the aviation-grade heuristic unit are connected to the three-phase stator winding ports of the motor. During the startup phase, the three-phase full-bridge inverter topology handles power transmission, while during the power generation phase, the rectifier topology handles power transmission. The switching transistor drive timings of the two topologies are independent. If the drive signal switching is performed directly at any time, the switching action will be superimposed on the non-zero current, resulting in a high current change rate and induced overvoltage, which can easily cause DC bus oscillation and excessive stress on the switching devices. This step first unifies the carrier time reference of the two topologies, and then uses the phase current zero-crossing point as the time anchor point for drive signal handover, so that the switching action of the transistor occurs at the moment when the phase current instantaneously reaches zero. At this time, the switching loss and current surge of the transistor are at a low level. The three phases complete the drive handover at their respective current zero-crossing phases, completing the topology switching phase by phase, avoiding the power surge caused by simultaneous switching of the three phases, and further smoothing out the electrical disturbances during the switching process.

[0102] Step 5 also includes the following steps:

[0103] Step 51: Align the carrier period of the startup mode power conversion topology with the carrier period of the generation mode power conversion topology to the same carrier frequency, and lock the starting phase of both to the same phase angle. The specific operation is as follows:

[0104] The starting mode power conversion topology and the generating mode power conversion topology originally each had independent pulse width modulation carriers. These carriers were generated by the timer peripheral of the digital signal processor, and the carrier period corresponded to the single switching duration of the switching transistor. During alignment, the period registers of the two carriers were first configured with the same value, ensuring that the switching frequencies of the two topologies were completely identical. In engineering applications, the same frequency value within the range of 10kHz to 20kHz can be selected to balance switching losses and current ripple performance. Phase locking is achieved through a synchronous trigger mechanism within the processor. A reset trigger signal is simultaneously sent to the timers of both carriers, causing the two carrier counters to increment from zero at the same moment, ensuring that the initial phases of the carriers are completely coincident. After phase locking is completed, the carrier periods of the two topologies are completely aligned in time, and the carrier phase angles at any given time are exactly the same. Subsequent current zero-crossing phase detection and drive signal generation can be performed based on the same time coordinate system, eliminating timing deviations caused by carrier asynchrony.

[0105] Step 52: Using the phase-locked carrier signal as the time reference, detect the instantaneous values ​​of the three-phase currents, and record the carrier phase angle corresponding to when the instantaneous value of each phase current crosses zero, as the zero-crossing phase angle of each phase current. The specific operation is as follows:

[0106] The instantaneous values ​​of the three-phase current are sampled in real time by Hall current sensors located at the stator winding output terminals. The sampling frequency is synchronized with the carrier frequency, and a single three-phase current synchronous sampling is completed within each carrier cycle. Using the phase-locked carrier signal as the time reference, the time length of a single carrier cycle is mapped to a phase interval of 0 to 2π, with each sampling moment corresponding to a specific carrier phase angle. Zero-crossing detection is determined by the change in the sign of the current between two adjacent sampling cycles. When a phase current changes from positive to negative or from negative to positive in two adjacent samplings, it is determined that the phase current has crossed the zero point within the two sampling intervals. To improve the detection accuracy at the zero-crossing moment, a linear interpolation method is used to calculate the accurate zero-crossing phase angle. This method is based on the approximate assumption of linear current change in a short time and uses known quantities from two adjacent sampling points to estimate the phase value at the zero-crossing point. It is suitable for engineering scenarios where the sampling frequency is much higher than the power frequency and can improve the detection accuracy to within a single sampling interval. The corresponding calculation relationship can be expressed as: ;

[0107] in The zero-crossing phase angle of the calculated phase current, in radians; The carrier phase angle corresponding to the (k-1)th sampling time is expressed in radians. The instantaneous value of the phase current obtained from the (k-1)th sampling is expressed in amperes. The carrier phase angle corresponding to the kth sampling time is expressed in radians. The instantaneous value of the phase current obtained from the kth sampling is expressed in amperes. Each phase current has two zero-crossing moments within each power frequency cycle, corresponding to the zero-crossing on the rising edge and the zero-crossing on the falling edge. The corresponding carrier phase angle can be calculated using the above method.

[0108] Step 53: Based on the zero-crossing phase angle of the current in each phase, at each phase's current zero-crossing phase angle, generate a first switching control signal for turning off the drive signal of the power conversion topology switch in the starting mode of that phase, and a second switching control signal for turning on the drive signal of the power conversion topology switch in the generation mode of that phase, so that the output times of the first switching control signal and the second switching control signal are both locked at the same current zero-crossing phase angle; and according to the corresponding current zero-crossing phase angles of the three phases, sequentially perform the drive signal handover operation of the three phases, as follows:

[0109] For each phase circuit, two switching control signals are generated. The first switching control signal is used to turn off the drive signal of the switching transistor for that phase in the starting mode power conversion topology, and the second switching control signal is used to turn on the drive signal of the switching transistor for that phase in the generating mode power conversion topology. The output times of both control signals are locked at the corresponding current zero-crossing phase angle, so that the switching action of the transistor occurs exactly at the moment when the instantaneous value of the phase current is zero. Since the current is zero at the switching moment, no sudden current component is generated during the switching process, which can effectively suppress induced overvoltage and circulating current impact. The three-phase drive signal handover is executed sequentially according to the order of their respective current zero-crossing phase angles. The phase that reaches the zero-crossing phase first completes the switching first, and the remaining phases wait for their own zero-crossing moment to arrive before performing the handover, until all three phases have completed the drive signal switching. The phase-by-phase handover method can disperse the power switching time nodes, avoid the instantaneous power jump caused by the simultaneous switching of the three-phase power paths, make the power topology switching process smooth, and reduce the electrical stress on the DC bus and motor windings.

[0110] In a preferred embodiment of the present invention, step 6 is further included: after the power topology switching is completed, the output command of the main control channel and the output command of the shadow channel are weighted and merged according to a preset weight to generate the final control output; the preset weight changes monotonically from a first weight value representing the dominance of the main control channel to a second weight value representing the dominance of the shadow channel within multiple control cycles after the switching is completed, and the specific operation is as follows:

[0111] After the physical handover of the power topology is completed, there may still be slight deviations in the output commands of the two control channels. If the system is directly switched to a single control channel in the generation mode, the jump in control quantity will be transmitted to the motor windings and DC bus through the power topology, causing current fluctuations and voltage oscillations. Through a multi-cycle weighted transition mechanism, the final control output can continuously transition from the control characteristics of the start-up mode to the control characteristics of the generation mode, allowing the closed-loop regulation process of the system to be smoothly carried out without undergoing a transition process of control loop re-convergence. This mechanism, together with the previous shadow channel pre-calculation, state consistency triggering, and current zero-crossing topology switching, constitutes a complete smooth switching system. It suppresses switching shocks throughout the entire process from control pre-preparation, switching timing selection, power path handover to control law transition, adapting to the requirements of aviation electrical systems for the smoothness and reliability of operating condition switching.

[0112] Step 6 also includes the following steps:

[0113] Step 61: The start time of weight change is determined by setting the completion time of the power topology switch as the start time and the preset total number of change cycles as the change duration. The end time of the change is then determined. The weight value of the control cycle at the start time is set as the first weight value, and the weight value of the control cycle at the end time is set as the second weight value. During each control cycle within the change period, the current weight value is calculated according to the progress of the current control cycle within the total number of change cycles, based on the preset weight function curve. The specific operation is as follows:

[0114] The weight transition begins at the moment the power topology switch is completed, corresponding to the start of the first control cycle after all three-phase drive signals have been switched over. This ensures the timing of the weight transition and topology switch. The preset total number of transition cycles is a positive integer, adjusted according to the system's trade-off requirements for transition response speed and stability. In engineering applications, it can be set to 10 to 100 control cycles. When the control cycle is 100 microseconds, the corresponding transition duration is 2 to 10 milliseconds. The first weight value is the weight coefficient of the main control channel at the start of the transition, with a value greater than 0.5, indicating that the main control channel output dominates the synthesized result. It is usually set to a value of 1, so that the final output completely follows the main control channel command in the initial stage. The second weight value is the weight coefficient of the main control channel at the end of the transition, with a value less than 0.5, indicating that the shadow channel output dominates the synthesized result. It is usually set to a value of zero, so that the final output completely follows the shadow channel command after the transition ends. During the transition period, the weight values ​​show a monotonically changing trend, and the changing law is determined by the preset weight function curve. The linear function is the weight curve form commonly used in engineering. The derivation of linear weight variation starts with the weight boundary conditions at the beginning and end of the variation. Using the progress percentage of the current cycle as the independent variable, the current weight value is obtained through linear interpolation, achieving uniform weight change over time. This simplifies parameter tuning and ensures a smooth transition without abrupt changes. The corresponding calculation is expressed as follows: ;

[0115] In the formula, w(k) is the current weight value of the kth incremental control cycle, and is a dimensionless coefficient; This is the first weight value, corresponding to the weight of the main control channel at the start of the change; The second weight value corresponds to the weight of the main control channel at the end of the change; k is the sequence number of the current change cycle, which starts counting from zero and ranges from 0 to N; N is the preset total number of change cycles, which is a positive integer; within each control cycle, the sequence number of the current cycle is read first, its progress ratio in the total number of cycles is calculated, and then substituted into the weight function to obtain the weight value of the current cycle.

[0116] Step 62: In each control cycle, obtain the first output command of the main control channel and the second output command of the shadow channel. Multiply the first output command by the current weight value calculated in step 61, and multiply the second output command by one minus the current weight value. Add the two products together to obtain the final control output of the current control cycle. The specific operation is as follows:

[0117] Within each control cycle, the first output command of the main control channel and the second output command of the shadow channel are read respectively. The two sets of commands contain control parameters with consistent dimensions and correspond one-to-one, covering the conduction timing and modulation parameters of the switching transistors in each bridge arm of the power conversion topology. Weighted calculations are performed on the control parameters at each corresponding position. The corresponding parameter of the first output command is multiplied by the current weight value, and the corresponding parameter of the second output command is multiplied by the difference between the current weight value and the first weight value. The two products are added together to obtain the final output value of that parameter. After all parameters are calculated, a complete final control output is formed and sent to the power drive circuit for execution. The derivation of the weighted synthesis is based on the fundamental idea of ​​linear transition. The smooth connection between the two control outputs is achieved through continuous changes in the weight coefficients, allowing the characteristics of the final control output to gradually shift with the transition process, ensuring smooth changes in the gain and dynamic characteristics of the control loop, and avoiding system oscillations. The corresponding calculation relationship is as follows: ;

[0118] In the formula, is the final control output command for the k-th control cycle; w(k) is the current weight value for the k-th control cycle; This is the first output command of the main control channel in the kth control cycle; This is the second output command of the shadow channel in the k-th control cycle. During the iteration process, the weight value changes monotonically, the proportion of the main control channel gradually decreases, and the proportion of the shadow channel gradually increases until the end of the iteration cycle, when the control is completely transferred to the power generation mode shadow channel, and the entire mode switching process is completed.

[0119] In addition, this invention also sets up corresponding fault-tolerant processing logic for various abnormal operating conditions that may occur during mode switching to ensure the reliability of system operation, specifically including:

[0120] 1. Handling of Status Consistency Timeout Anomalies: When the estimated rotor electrical angular velocity reaches the preset switching threshold, if the status consistency index value remains greater than the preset status consistency threshold within the preset timeout judgment period, failing to meet the switching triggering conditions, it is determined to be a status consistency timeout anomaly. At this time, the system suspends the switching process, maintains startup mode operation, and simultaneously sends a switching pending alarm signal to the airborne control system; subsequently, it gradually increases the pre-adjustment range of the DC bus voltage target value in the power generation mode, expands the output adjustment range of the shadow channel, and re-triggers the switching after the status consistency index meets the threshold; if the conditions are still not met after exceeding the maximum waiting period, a degraded switching process is executed, the consistency threshold is appropriately relaxed, the switching is completed, and the anomaly log is recorded.

[0121] 2. Handling of Shadow Channel Calculation Anomalies: Each control cycle performs amplitude and rationality checks on the output commands of the shadow channel. If the output command exceeds the preset reasonable range or anomalies such as calculation overflow occur, it is determined to be a shadow channel calculation anomaly. In this case, the shadow channel calculation result of the current cycle is discarded, and the valid output command of the previous cycle is used. Simultaneously, the control parameter self-check process is triggered to investigate the cause of anomalies in the state sampling and control law calculation stages. If anomalies occur in multiple consecutive control cycles, the shadow channel update is paused, the current switching process is terminated, and stable operation in the start-up mode is maintained. After the operating conditions are reset, the switching preparation stage is re-entered.

[0122] 3. Handling Power Topology Switching Anomalies: During the carrier phase-locking (PLC) phase-locking (PLL) phase, if phase locking of two carriers cannot be completed within the preset PLC waiting period, it is determined to be a carrier PLC anomaly. In this case, the topology switching process is terminated, the current power topology operation is maintained, and the carrier synchronization module is restarted. Switching is performed again after successful PLC locking. During the drive signal handover process, if an anomaly such as current amplitude exceeding the limit or bridge arm shoot-through risk is detected after a phase current crosses zero, the two drive signals of the corresponding phase are immediately blocked, triggering the hardware protection mechanism. At the same time, the topology state is rolled back to the previous stage to prevent the fault from escalating.

[0123] 4. Handling of Covariance Mapping Anomalies: If the mapped error covariance matrix still fails the positive definiteness check after symmetry and diagonal correction, it is determined to be an anomaly in the covariance mapping. In this case, the mapping result is discarded, and the default initial error covariance matrix of the power generation mode state estimator is adopted. The gain coefficient of the state estimator is appropriately limited to reduce the estimation fluctuation in the initial stage. At the same time, the fast convergence mechanism of the state estimator is activated, and the estimation convergence is accelerated by increasing the observation noise weight. Within several control cycles, the estimator enters a stable operating state, avoiding the estimation oscillations from being transmitted to the control loop.

[0124] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and its improved concepts, should be covered within the scope of protection of the present invention.

Claims

1. A method for switching modes in an all-in-one machine, characterized in that, include: Step 1: Construct a dual-mode shared state variable set, incorporating the state variables in the start-up mode and the control quantities in the power generation mode into the same extended state vector to form a unified state space; Step 2: At the end of the start-up phase, when the estimated value of the rotor electric angular velocity reaches the preset switching threshold, the unified state space is used as input. While maintaining the operation of the main control channel of the start-up mode, the output command of the power generation mode controller in the current state is calculated in parallel to form a shadow channel. Step 3: Calculate the state consistency index value based on the difference between the output command of the master control channel and the output command of the shadow channel, as well as the deviation between the actual state variables and the corresponding reference state variables in the unified state space. When the state consistency index value is less than the preset state consistency threshold and the rotor electric angular velocity estimate reaches the preset switching threshold, the switch from the start-up mode to the power generation mode is triggered. Step 4: When switching is triggered, perform a mapping operation on the error covariance matrix of the state estimator to map the state error covariance under the current control model to the state error covariance under the target control model. Step 5: During the power topology switching process, the switching period carrier phases of the two power conversion topologies are locked, and the drive signals of the two power conversion topologies are exchanged at the phase corresponding to the zero crossing of the phase current in the carrier period to complete the power topology switching. Step 6: After the power topology switching is completed, the output commands of the main control channel and the output commands of the shadow channel are weighted and merged according to preset weights to generate the final control output. The preset weights change monotonically from the first weight value representing the dominance of the master control channel to the second weight value representing the dominance of the shadow channel within multiple control cycles after the switch is completed.

2. The method for switching modes of an all-in-one machine according to claim 1, characterized in that, Step 1 includes: Step 11: Using the two-phase stationary coordinate system as the reference coordinate system, the stator current vector component, rotor flux vector component, and rotor electric angular velocity in the starting mode are used as the first set of state variables to form the starting state vector; the DC bus voltage and AC side input current vector component in the power generation mode are used as the second set of state variables to form the power generation state vector. Step 12: Merge the starting state vector and the power generation state vector to form an extended state vector, and use the extended state vector as a unified state space.

3. The method for switching modes of an all-in-one machine according to claim 2, characterized in that, Step 2 includes: Step 21: Monitor the estimated value of the rotor electric angular velocity in the unified state space. When the estimated value of the rotor electric angular velocity reaches the set proportion of the preset switching threshold, start the parallel calculation preparation process for the output command of the power generation mode controller. The preset switching threshold is determined according to the minimum operating speed required by the power generation mode, and the set proportion is a positive number less than 1. Step 22: Using the measured value of DC bus voltage in the unified state space, the measured value of AC side input current vector component, the estimated value of rotor electric angular velocity, and the preset target value of DC bus voltage in the power generation mode as input quantities, calculate the output command according to the control law of the power generation mode; the output command includes the switching timing parameters of the power conversion topology in the power generation mode; the calculation is performed in parallel with the calculation of the output command of the main control channel in the start-up mode. Step 23: The output command calculated in step 22 is temporarily stored in the shadow channel register, and the calculation is repeated in each subsequent control cycle to continuously update the output command content in the shadow channel register. At the same time, in each control cycle, the updated output command in the shadow channel register is associated with the rotor electric angular velocity estimate and stored to form an output command sequence with a speed tag.

4. The method for switching modes of an all-in-one machine according to claim 3, characterized in that, Step 3 includes: Step 31: Obtain the output command of the main control channel in the current control cycle, and read the output command stored in association with the current rotor electric angular velocity estimate from the shadow channel register, and calculate the difference between the two. Step 32: Obtain the measured values ​​of each state variable in the unified state space and the corresponding reference values ​​of each state variable. Subtract each measured value from each reference value one by one and take the absolute value to obtain the absolute value of the deviation of each state variable.

5. The method for switching modes of an all-in-one machine according to claim 4, characterized in that, Step 3 also includes: Step 33: The difference is used as the first dimension deviation, and the absolute values ​​of each deviation are multiplied by their respective preset weight coefficients and then summed to obtain the second dimension comprehensive deviation; the first dimension deviation and the second dimension comprehensive deviation are normalized to a preset numerical range; the normalized first dimension deviation value and the normalized second dimension deviation value are added to obtain the state consistency index value; the state consistency index value is compared with the state consistency preset threshold. When the state consistency index value is less than the state consistency preset threshold and the rotor electric angular velocity estimate reaches the preset switching threshold, the switch from the start-up mode to the power generation mode is triggered.

6. The method for switching modes of an all-in-one machine according to claim 5, characterized in that, Step 4 includes: Step 41: Obtain the first Jacobian matrix of the starting mode state estimator, which is constructed based on the state transition function of the starting mode at the state estimate at the current time; Obtain the second Jacobian matrix of the power generation mode state estimator, which is constructed based on the state transition function of the power generation mode at the state estimate at the current time. Step 42: Using the first Jacobian matrix and the second Jacobian matrix as input, multiply the second Jacobian matrix on the left by the inverse of the first Jacobian matrix to obtain the transformation matrix.

7. The method for switching modes of an all-in-one machine according to claim 6, characterized in that, Step 4 also includes: Step 43: Obtain the current error covariance matrix of the starting mode state estimator; multiply the current error covariance matrix on the left by the transformation matrix, and then multiply it on the right by the transpose of the transformation matrix to obtain the mapped error covariance matrix; Step 44: Perform a symmetry operation on the mapped error covariance matrix to obtain a symmetric matrix; perform a positive definiteness check on the symmetric matrix. If the positive definiteness check passes, the symmetric matrix is ​​used as the error covariance matrix of the power generation mode state estimator; if the positive definiteness check fails, the diagonal elements of the symmetric matrix are modified, and the modified symmetric matrix is ​​used as the error covariance matrix of the power generation mode state estimator.

8. The method for switching modes of an all-in-one machine according to claim 7, characterized in that, Step 5 includes: Step 51: Align the carrier period of the startup mode power conversion topology with the carrier period of the generation mode power conversion topology to the same carrier frequency, and lock the starting phase of both to the same phase angle. Step 52: Using the phase-locked carrier signal as the time reference, detect the instantaneous values ​​of the three-phase currents, and record the carrier phase angle corresponding to the instantaneous value of each phase current crossing the zero point, as the zero-crossing phase angle of each phase current.

9. The method for switching modes of an all-in-one machine according to claim 8, characterized in that, Step 5 further includes: Step 53: Based on the zero-crossing phase angle of the current in each phase, at the zero-crossing phase angle of the current in each phase, generate a first switch control signal for turning off the drive signal of the power conversion topology switch in the starting mode of that phase, and a second switch control signal for turning on the drive signal of the power conversion topology switch in the generation mode of that phase, so that the output time of the first switch control signal and the second switch control signal are both locked at the same zero-crossing phase angle of the current; and execute the drive signal handover operation of the three phases in sequence according to the corresponding zero-crossing phase angle of the current in each of the three phases.

10. A method for switching modes of an all-in-one machine according to claim 9, characterized in that, Step 6 includes: Step 61: Take the moment when the power topology switching is completed as the start time of the weight change, take the preset total number of change cycles as the change duration, and determine the end time of the change; set the weight value of the control cycle where the start time is located as the first weight value, and the weight value of the control cycle where the end time is located as the second weight value; in each control cycle during the change period, calculate the current weight value according to the progress of the current control cycle in the total number of change cycles, according to the preset weight function curve. Step 62: In each control cycle, obtain the first output command of the main control channel and the second output command of the shadow channel, multiply the first output command by the current weight value calculated in step 61, multiply the second output command by one minus the current weight value, and add the two products to obtain the final control output of the current control cycle.

Images